Stay Ahead, Stay ONMINE

Generative AI Is Declarative

ChatGPT launched in 2022 and kicked off the Generative Ai boom. In the two years since, academics, technologists, and armchair experts have written libraries worth of articles on the technical underpinnings of generative AI and about the potential capabilities of both current and future generative AI models. Surprisingly little has been written about how we interact with these tools—the human-AI interface. The point where we interact with AI models is at least as important as the algorithms and data that create them. “There is no success where there is no possibility of failure, no art without the resistance of the medium” (Raymond Chandler). In that vein, it’s useful to examine human-AI interaction and the strengths and weaknesses inherent in that interaction. If we understand the “resistance in the medium” then product managers can make smarter decisions about how to incorporate generative AI into their products. Executives can make smarter decisions about what capabilities to invest in. Engineers and designers can build around the tools’ limitations and showcase their strengths. Everyday people can know when to use generative AI and when not to. Imagine walking into a restaurant and ordering a cheeseburger. You don’t tell the chef how to grind the beef, how hot to set the grill, or how long to toast the bun. Instead, you simply describe what you want: “I’d like a cheeseburger, medium rare, with lettuce and tomato.” The chef interprets your request, handles the implementation, and delivers the desired outcome. This is the essence of declarative interaction—focusing on the what rather than the how. Now, imagine interacting with a Large Language Model (LLM) like ChatGPT. You don’t have to provide step-by-step instructions for how to generate a response. Instead, you describe the result you’re looking for: “A user story that lets us implement A/B testing for the Buy button on our website.” The LLM interprets your prompt, fills in the missing details, and delivers a response. Just like ordering a cheeseburger, this is a declarative mode of interaction. Explaining the steps to make a cheeseburger is an imperative interaction. Our LLM prompts sometimes feel imperative. We might phrase our prompts like a question: ”What is the tallest mountain on earth?” This is equivalent to describing “the answer to the question ‘What is the tallest mountain on earth?’” We might phrase our prompt as a series of instructions: ”Write a summary of the attached report, then read it as if you are a product manager, then type up some feedback on the report.” But, again, we’re describing the result of a process with some context for what that process is. In this case, it is a sequence of descriptive results—the report then the feedback. This is a more useful way to think about LLMs and generative AI. In some ways it is more accurate; the neural network model behind the curtain doesn’t explain why or how it produced one output instead of another. More importantly though, the limitations and strengths of generative AI make more sense and become more predictable when we think of these models as declarative. LLMs as a declarative mode of interaction Computer scientists use the term “declarative” to describe coding languages. SQL is one of the most common. The code describes the output table and the procedures in the database figure out how to retrieve and combine the data to produce the result. LLMs share many of the benefits of declarative languages like SQL or declarative interactions like ordering a cheeseburger. Focus on desired outcome: Just as you describe the cheeseburger you want, you describe the output you want from the LLM. For example, “Summarize this article in three bullet points” focuses on the result, not the process. Abstraction of implementation: When you order a cheeseburger, you don’t need to know how the chef prepares it. When submitting SQL code to a server, the server figures out where the data lives, how to fetch it, and how to aggregate it based on your description. You as the user don’t need to know how. With LLMs, you don’t need to know how the model generates the response. The underlying mechanisms are abstracted away. Filling in missing details: If you don’t specify onions on your cheeseburger, the chef won’t include them. If you don’t specify a field in your SQL code, it won’t show up in the output table. This is where LLMs differ slightly from declarative coding languages like SQL. If you ask ChatGPT to create an image of “a cheeseburger with lettuce and tomato” it may also show the burger on a sesame seed bun or include pickles, even if that wasn’t in your description. The details you omit are inferred by the LLM using the “average” or “most likely” detail depending on the context, with a bit of randomness thrown in. Ask for the cheeseburger image six times; it may show you three burgers with cheddar cheese, two with Swiss, and one with pepper jack. Like other forms of declarative interaction, LLMs share one key limitation. If your description is vague, ambiguous, or lacks enough detail, then the result may not be what you hoped to see. It is up to the user to describe the result with sufficient detail. This explains why we often iterate to get what we’re looking for when using LLMs and generative AI. Going back to our cheeseburger analogy, the process to generate a cheeseburger from an LLM may look like this. “Make me a cheeseburger, medium rare, with lettuce and tomatoes.” The result also has pickles and uses cheddar cheese. The bun is toasted. There’s mayo on the top bun. “Make the same thing but this time no pickles, use pepper jack cheese, and a sriracha mayo instead of plain mayo.” The result now has pepper jack, no pickles. The sriracha mayo is applied to the bottom bun and the bun is no longer toasted. “Make the same thing again, but this time, put the sriracha mayo on the top bun. The buns should be toasted.” Finally, you have the cheeseburger you’re looking for. This example demonstrates one of the main points of friction with human-AI interaction. Human beings are really bad at describing what they want with sufficient detail on the first attempt. When we asked for a cheeseburger, we had to refine our description to be more specific (the type of cheese). In the second generation, some of the inferred details (whether the bun was toasted) changed from one iteration to the next, so then we had to add that specificity to our description as well. Iteration is an important part of AI-human generation. Insight: When using generative AI, we need to design an iterative human-AI interaction loop that enables people to discover the details of what they want and refine their descriptions accordingly. To iterate, we need to evaluate the results. Evaluation is extremely important with generative AI. Say you’re using an LLM to write code. You can evaluate the code quality if you know enough to understand it or if you can execute it and inspect the results. On the other hand, hypothetical questions can’t be tested. Say you ask ChatGPT, “What if we raise our product prices by 5 percent?” A seasoned expert could read the output and know from experience if a recommendation doesn’t take into account important details. If your product is property insurance, then increasing premiums by 5 percent may mean pushback from regulators, something an experienced veteran of the industry would know. For non-experts in a topic, there’s no way to tell if the “average” details inferred by the model make sense for your specific use case. You can’t test and iterate. Insight: LLMs work best when the user can evaluate the result quickly, whether through execution or through prior knowledge. The examples so far involve general knowledge. We all know what a cheeseburger is. When you start asking about non-general information—like when you can make dinner reservations next week—you delve into new points of friction. In the next section we’ll think about different types of information, what we can expect the AI to “know”, and how this impacts human-AI interaction. What did the AI know, and when did it know it? Above, I explained how generative AI is a declarative mode of interaction and how that helps understand its strengths and weaknesses. Here, I’ll identify how different types of information create better or worse human-AI interactions. Understanding the information available When we describe what we want to an LLM, and when it infers missing details from our description, it draws from different sources of information. Understanding these sources of information is important. Here’s a useful taxonomy for information types: General information used to train the base model. Non-general information that the base model is not aware of. Fresh information that is new or changes rapidly, like stock prices or current events. Non-public information, like facts about you and where you live or about your company, its employees, its processes, or its codebase. General information vs. non-general information LLMs are built on a massive corpus of written word data. A large part of GPT-3 was trained on a combination of books, journals, Wikipedia, Reddit, and CommonCrawl (an open-source repository of web crawl data). You can think of the models as a highly compressed version of that data, organized in a gestalt manner—all the like things are close together. When we submit a prompt, the model takes the words we use (and any words added to the prompt behind the scenes) and finds the closest set of related words based on how those things appear in the data corpus. So when we say “cheeseburger” it knows that word is related to “bun” and “tomato” and “lettuce” and “pickles” because they all occur in the same context throughout many data sources. Even when we don’t specify pickles, it uses this gestalt approach to fill in the blanks. This training information is general information, and a good rule of thumb is this: if it was in Wikipedia a year ago then the LLM “knows” about it. There could be new articles on Wikipedia, but that didn’t exist when the model was trained. The LLM doesn’t know about that unless told. Now, say you’re a company using an LLM to write a product requirements document for a new web app feature. Your company, like most companies, is full of its own lingo. It has its own lore and history scattered across thousands of Slack messages, emails, documents, and some tenured employees who remember that one meeting in Q1 last year. The LLM doesn’t know any of that. It will infer any missing details from general information. You need to supply everything else. If it wasn’t in Wikipedia a year ago, the LLM doesn’t know about it. The resulting product requirements document may be full of general facts about your industry and product but could lack important details specific to your firm. This is non-general information. This includes personal info, anything kept behind a log-in or paywall, and non-digital information. This non-general information permeates our lives, and incorporating it is another source of friction when working with generative AI. Non-general information can be incorporated into a generative AI application in three ways: Through model fine-tuning (supplying a large corpus to the base model to expand its reference data). Retrieved and fed it to the model at query time (e.g., the retrieval augmented generation or “RAG” technique). Supplied by the user in the prompt. Insight: When designing any human-AI interactions, you should think about what non-general information is required, where you will get it, and how you will expose it to the AI. Fresh information Any information that changes in real-time or is new can be called fresh information. This includes new facts like current events but also frequently changing facts like your bank account balance. If the fresh information is available in a database or some searchable source, then it needs to be retrieved and incorporated into the application. To retrieve the information from a database, the LLM must create a query, which may require specific details that the user didn’t include. Here’s an example. I have a chatbot that gives information on the stock market. You, the user, type the following: “What is the current price of Apple? Has it been increasing or decreasing recently?” The LLM doesn’t have the current price of Apple in its training data. This is fresh, non-general information. So, we need to retrieve it from a database. The LLM can read “Apple”, know that you’re talking about the computer company, and that the ticker symbol is AAPL. This is all general information. What about the “increasing or decreasing” part of the prompt? You did not specify over what period—increasing in the past day, month, year? In order to construct a database query, we need more detail. LLMs are bad at knowing when to ask for detail and when to fill it in. The application could easily pull the wrong data and provide an unexpected or inaccurate answer. Only you know what these details should be, depending on your intent. You must be more specific in your prompt. A designer of this LLM application can improve the user experience by specifying required parameters for expected queries. We can ask the user to explicitly input the time range or design the chatbot to ask for more specific details if not provided. In either case, we need to have a specific type of query in mind and explicitly design how to handle it. The LLM will not know how to do this unassisted. Insight: If a user is expecting a more specific type of output, you need to explicitly ask for enough detail. Too little detail could produce a poor quality output. Non-public information Incorporating non-public information into an LLM prompt can be done if that information can be accessed in a database. This introduces privacy issues (should the LLM be able to access my medical records?) and complexity when incorporating multiple non-public sources of information. Let’s say I have a chatbot that helps you make dinner reservations. You, the user, type the following: “Help me make dinner reservations somewhere with good Neapolitan pizza.” The LLM knows what a Neapolitan pizza is and can infer that “dinner” means this is for an evening meal. To do this task well, it needs information about your location, the restaurants near you and their booking status, or even personal details like dietary restrictions. Assuming all that non-public information is available in databases, bringing them all together into the prompt takes a lot of engineering work. Even if the LLM could find the “best” restaurant for you and book the reservation, can you be confident it has done that correctly? You never specified how many people you need a reservation for. Since only you know this information, the application needs to ask for it upfront. If you’re designing this LLM-based application, you can make some thoughtful choices to help with these problems. We could ask about a user’s dietary restrictions when they sign up for the app. Other information, like the user’s schedule that evening, can be given in a prompting tip or by showing the default prompt option “show me reservations for two for tomorrow at 7PM”. Promoting tips may not feel as automagical as a bot that does it all, but they are a straightforward way to collect and integrate the non-public information. Some non-public information is large and can’t be quickly collected and processed when the prompt is given. These need to be fine-tuned in batch or retrieved at prompt time and incorporated. A chatbot that answers information about a company’s HR policies can obtain this information from a corpus of non-public HR documents. You can fine-tune the model ahead of time by feeding it the corpus. Or you can implement a retrieval augmented generation technique, searching a corpus for relevant documents and summarizing the results. Either way, the response will only be as accurate and up-to-date as the corpus itself. Insight: When designing an AI application, you need to be aware of non-public information and how to retrieve it. Some of that information can be pulled from databases. Some needs to come from the user, which may require prompt suggestions or explicitly asking. If you understand the types of information and treat human-AI interaction as declarative, you can more easily predict which AI applications will work and which ones won’t. In the next section we’ll look at OpenAI’s Operator and deep research products. Using this framework, we can see where these applications fall short, where they work well, and why. Critiquing OpenAI’s Operator and deep research through a declarative lens I have now explained how thinking of generative AI as declarative helps us understand its strengths and weaknesses. I also identified how different types of information create better or worse human-AI interactions. Now I’ll apply these ideas by critiquing two recent products from OpenAI—Operator and deep research. It’s important to be honest about the shortcomings of AI applications. Bigger models trained on more data or using new techniques might one day solve some issues with generative AI. But other issues arise from the human-AI interaction itself and can only be addressed by making appropriate design and product choices. These critiques demonstrate how the framework can help identify where the limitations are and how to address them. The limitations of Operator Journalist Casey Newton of Platformer reviewed Operator in an article that was largely positive. Newton has covered AI extensively and optimistically. Still, Newton couldn’t help but point out some of Operator’s frustrating limitations. [Operator] can take action on your behalf in ways that are new to AI systems — but at the moment it requires a lot of hand-holding, and may cause you to throw up your hands in frustration.  My most frustrating experience with Operator was my first one: trying to order groceries. “Help me buy groceries on Instacart,” I said, expecting it to ask me some basic questions. Where do I live? What store do I usually buy groceries from? What kinds of groceries do I want?  It didn’t ask me any of that. Instead, Operator opened Instacart in the browser tab and begin searching for milk in grocery stores located in Des Moines, Iowa. The prompt “Help me buy groceries on Instacart,” viewed declaratively, describes groceries being purchased using Instacart. It doesn’t have a lot of the information someone would need to buy groceries, like what exactly to buy, when it would be delivered, and to where. It’s worth repeating: LLMs are not good at knowing when to ask additional questions unless explicitly programmed to do so in the use case. Newton gave a vague request and expected follow-up questions. Instead, the LLM filled in all the missing details with the “average”. The average item was milk. The average location was Des Moines, Iowa. Newton doesn’t mention when it was scheduled to be delivered, but if the “average” delivery time is tomorrow, then that was likely the default. If we engineered this application specifically for ordering groceries, keeping in mind the declarative nature of AI and the information it “knows”, then we could make thoughtful design choices that improve functionality. We would need to prompt the user to specify when and where they want groceries up front (non-public information). With that information, we could find an appropriate grocery store near them. We would need access to that grocery store’s inventory (more non-public information). If we have access to the user’s previous orders, we could also pre-populate a cart with items typical to their order. If not, we may add a few suggested items and guide them to add more. By limiting the use case, we only have to deal with two sources of non-public information. This is a more tractable problem than Operator’s “agent that does it all” approach. Newton also mentions that this process took eight minutes to complete, and “complete” means that Operator did everything up to placing the order. This is a long time with very little human-in-the-loop iteration. Like we said before, an iteration loop is very important for human-AI interaction. A better-designed application would generate smaller steps along the way and provide more frequent interaction. We could prompt the user to describe what to add to their shopping list. The user might say, “Add barbeque sauce to the list,” and see the list update. If they see a vinegar-based barbecue sauce, they can refine that by saying, “Replace that with a barbeque sauce that goes well with chicken,” and might be happier when it’s replaced by a honey barbecue sauce. These frequent iterations make the LLM a creative tool rather than a does-it-all agent. The does-it-all agent looks automagical in marketing, but a more guided approach provides more utility with a less frustrating and more delightful experience. Elsewhere in the article, Newton gives an example of a prompt that Operator performed well: “Put together a lesson plan on the Great Gatsby for high school students, breaking it into readable chunks and then creating assignments and connections tied to the Common Core learning standard.” This prompt describes an output using much more specificity. It also solely relies on general information—the Great Gatsby, the Common Core standard, and a general sense of what assignments are. The general-information use case lends itself better to AI generation, and the prompt is explicit and detailed in its request. In this case, very little guidance was given to create the prompt, so it worked better. (In fact, this prompt comes from Ethan Mollick who has used it to evaluate AI chatbots.) This is the risk of general-purpose AI applications like Operator. The quality of the result relies heavily on the use case and specificity provided by the user. An application with a more specific use case allows for more design guidance and can produce better output more reliably. The limitations of deep research Newton also reviewed deep research, which, according to OpenAI’s website, is an “agent that uses reasoning to synthesize large amounts of online information and complete multi-step research tasks for you.” Deep research came out after Newton’s review of Operator. Newton chose an intentionally tricky prompt that prods at some of the tool’s limitations regarding fresh information and non-general information: “I wanted to see how OpenAI’s agent would perform given that it was researching a story that was less than a day old, and for which much of the coverage was behind paywalls that the agent would not be able to access. And indeed, the bot struggled more than I expected.” Near the end of the article, Newton elaborates on some of the shortcomings he noticed with deep research. OpenAI’s deep research suffers from the same design problem that almost all AI products have: its superpowers are completely invisible and must be harnessed through a frustrating process of trial and error. Generally speaking, the more you already know about something, the more useful I think deep research is. This may be somewhat counterintuitive; perhaps you expected that an AI agent would be well suited to getting you up to speed on an important topic that just landed on your lap at work, for example.  In my early tests, the reverse felt true. Deep research excels for drilling deep into subjects you already have some expertise in, letting you probe for specific pieces of information, types of analysis, or ideas that are new to you. The “frustrating trial and error” shows a mismatch between Newton’s expectations and a necessary aspect of many generative AI applications. A good response requires more information than the user will probably give in the first attempt. The challenge is to design the application and set the user’s expectations so that this interaction is not frustrating but exciting. Newton’s more poignant criticism is that the application requires already knowing something about the topic for it to work well. From the perspective of our framework, this makes sense. The more you know about a topic, the more detail you can provide. And as you iterate, having knowledge about a topic helps you observe and evaluate the output. Without the ability to describe it well or evaluate the results, the user is less likely to use the tool to generate good output. A version of deep research designed for lawyers to perform legal research could be powerful. Lawyers have an extensive and common vocabulary for describing legal matters, and they’re more likely to see a result and know if it makes sense. Generative AI tools are fallible, though. So, the tool should focus on a generation-evaluation loop rather than writing a final draft of a legal document. The article also highlights many improvements compared to Operator. Most notably, the bot asked clarifying questions. This is the most impressive aspect of the tool. Undoubtedly, it helps that deep search has a focused use-case of retrieving and summarizing general information instead of a does-it-all approach. Having a focused use case narrows the set of likely interactions, letting you design better guidance into the prompt flow. Good application design with generative AI Designing effective generative AI applications requires thoughtful consideration of how users interact with the technology, the types of information they need, and the limitations of the underlying models. Here are some key principles to guide the design of generative AI tools: 1. Constrain the input and focus on providing details Applications are inputs and outputs. We want the outputs to be useful and pleasant. By giving a user a conversational chatbot interface, we allow for a vast surface area of potential inputs, making it a challenge to guarantee useful outputs. One strategy is to limit or guide the input to a more manageable subset. For example, FigJam, a collaborative whiteboarding tool, uses pre-set template prompts for timelines, Gantt charts, and other common whiteboard artifacts. This provides some structure and predictability to the inputs. Users still have the freedom to describe further details like color or the content for each timeline event. This approach ensures that the AI has enough specificity to generate meaningful outputs while giving users creative control. 2. Design frequent iteration and evaluation into the tool Iterating in a tight generation-evaluation loop is essential for refining outputs and ensuring they meet user expectations. OpenAI’s Dall-E is great at this. Users quickly iterate on image prompts and refine their descriptions to add additional detail. If you type “a picture of a cheeseburger on a plate”, you may then add more detail by specifying “with pepperjack cheese”. AI code generating tools work well because users can run a generated code snippet immediately to see if it works, enabling rapid iteration and validation. This quick evaluation loop produces better results and a better coder experience.  Designers of generative AI applications should pull the user in the loop early, often, in a way that is engaging rather than frustrating. Designers should also consider the user’s knowledge level. Users with domain expertise can iterate more effectively. Referring back to the FigJam example, the prompts and icons in the app quickly communicate “this is what we call a mind map” or “this is what we call a gantt chart” for users who want to generate these artifacts but don’t know the terms for them. Giving the user some basic vocabulary can help them better generate desired results quickly with less frustration. 3. Be mindful of the types of information needed LLMs excel at tasks involving general knowledge already in the base training set. For example, writing class assignments involves absorbing general information, synthesizing it, and producing a written output, so LLMs are very well-suited for that task. Use cases that require non-general information are more complex. Some questions the designer and engineer should ask include: Does this application require fresh information? Maybe this is knowledge of current events or a user’s current bank account balance. If so, that information needs to be retrieved and incorporated into the model. How much non-general information does the LLM need to know? If it’s a lot of information—like a corpus of company documentation and communication—then the model may need to be fine tuned in batch ahead of time. If the information is relatively small, a retrieval augmented generation (RAG) approach at query time may suffice.  How many sources of non-general information—small and finite or potentially infinite? General purpose agents like Operator face the challenge of potentially infinite non-general information sources. Depending on what the user requires, it could need to access their contacts, restaurant reservation lists, financial data, or even other people’s calendars. A single-purpose restaurant reservation chatbot may only need access to Yelp, OpenTable, and the user’s calendar. It’s much easier to reconcile access and authentication for a handful of known data sources. Is there context-specific information that can only come from the user? Consider our restaurant reservation chatbot. Is the user making reservations for just themselves? Probably not. “How many people and who” is a detail that only the user can provide, an example of non-public information that only the user knows. We shouldn’t expect the user to provide this information upfront and unguided. Instead, we can use prompt suggestions so they include the information. We may even be able to design the LLM to ask these questions when the detail is not provided. 4. Focus on specific use cases Broad, all-purpose chatbots often struggle to deliver consistent results due to the complexity and variability of user needs. Instead, focus on specific use cases where the AI’s shortcomings can be mitigated through thoughtful design. Narrowing the scope helps us address many of the issues above. We can identify common requests for the use case and incorporate those into prompt suggestions. We can design an iteration loop that works well with the type of thing we’re generating. We can identify sources of non-general information and devise solutions to incorporate it into the model or prompt. 5. Translation or summary tasks work well A common task for ChatGPT is to rewrite something in a different style, explain what some computer code is doing, or summarize a long document. These tasks involve converting a set of information from one form to another. We have the same concerns about non-general information and context. For instance, a Chatbot asked to explain a code script doesn’t know the system that script is part of unless that information is provided. But in general, the task of transforming or summarizing information is less prone to missing details. By definition, you have provided the details it needs. The result should have the same information in a different or more condensed form. The exception to the rules There is a case when it doesn’t matter if you break any or all of these rules—when you’re just having fun. LLMs are creative tools by nature. They can be an easel to paint on, a sandbox to build in, a blank sheet to scribe. Iteration is still important; the user wants to see the thing they’re creating as they create it. But unexpected results due to lack of information or omitted details may add to the experience. If you ask for a cheeseburger recipe, you might get some funny or interesting ingredients. If the stakes are low and the process is its own reward, don’t worry about the rules.

ChatGPT launched in 2022 and kicked off the Generative Ai boom. In the two years since, academics, technologists, and armchair experts have written libraries worth of articles on the technical underpinnings of generative AI and about the potential capabilities of both current and future generative AI models.

Surprisingly little has been written about how we interact with these tools—the human-AI interface. The point where we interact with AI models is at least as important as the algorithms and data that create them. “There is no success where there is no possibility of failure, no art without the resistance of the medium” (Raymond Chandler). In that vein, it’s useful to examine human-AI interaction and the strengths and weaknesses inherent in that interaction. If we understand the “resistance in the medium” then product managers can make smarter decisions about how to incorporate generative AI into their products. Executives can make smarter decisions about what capabilities to invest in. Engineers and designers can build around the tools’ limitations and showcase their strengths. Everyday people can know when to use generative AI and when not to.

Imagine walking into a restaurant and ordering a cheeseburger. You don’t tell the chef how to grind the beef, how hot to set the grill, or how long to toast the bun. Instead, you simply describe what you want: “I’d like a cheeseburger, medium rare, with lettuce and tomato.” The chef interprets your request, handles the implementation, and delivers the desired outcome. This is the essence of declarative interaction—focusing on the what rather than the how.

Now, imagine interacting with a Large Language Model (LLM) like ChatGPT. You don’t have to provide step-by-step instructions for how to generate a response. Instead, you describe the result you’re looking for: “A user story that lets us implement A/B testing for the Buy button on our website.” The LLM interprets your prompt, fills in the missing details, and delivers a response. Just like ordering a cheeseburger, this is a declarative mode of interaction.

Explaining the steps to make a cheeseburger is an imperative interaction. Our LLM prompts sometimes feel imperative. We might phrase our prompts like a question: ”What is the tallest mountain on earth?” This is equivalent to describing “the answer to the question ‘What is the tallest mountain on earth?’” We might phrase our prompt as a series of instructions: ”Write a summary of the attached report, then read it as if you are a product manager, then type up some feedback on the report.” But, again, we’re describing the result of a process with some context for what that process is. In this case, it is a sequence of descriptive results—the report then the feedback.

This is a more useful way to think about LLMs and generative AI. In some ways it is more accurate; the neural network model behind the curtain doesn’t explain why or how it produced one output instead of another. More importantly though, the limitations and strengths of generative AI make more sense and become more predictable when we think of these models as declarative.

LLMs as a declarative mode of interaction

Computer scientists use the term “declarative” to describe coding languages. SQL is one of the most common. The code describes the output table and the procedures in the database figure out how to retrieve and combine the data to produce the result. LLMs share many of the benefits of declarative languages like SQL or declarative interactions like ordering a cheeseburger.

  1. Focus on desired outcome: Just as you describe the cheeseburger you want, you describe the output you want from the LLM. For example, “Summarize this article in three bullet points” focuses on the result, not the process.
  2. Abstraction of implementation: When you order a cheeseburger, you don’t need to know how the chef prepares it. When submitting SQL code to a server, the server figures out where the data lives, how to fetch it, and how to aggregate it based on your description. You as the user don’t need to know how. With LLMs, you don’t need to know how the model generates the response. The underlying mechanisms are abstracted away.
  3. Filling in missing details: If you don’t specify onions on your cheeseburger, the chef won’t include them. If you don’t specify a field in your SQL code, it won’t show up in the output table. This is where LLMs differ slightly from declarative coding languages like SQL. If you ask ChatGPT to create an image of “a cheeseburger with lettuce and tomato” it may also show the burger on a sesame seed bun or include pickles, even if that wasn’t in your description. The details you omit are inferred by the LLM using the “average” or “most likely” detail depending on the context, with a bit of randomness thrown in. Ask for the cheeseburger image six times; it may show you three burgers with cheddar cheese, two with Swiss, and one with pepper jack.

Like other forms of declarative interaction, LLMs share one key limitation. If your description is vague, ambiguous, or lacks enough detail, then the result may not be what you hoped to see. It is up to the user to describe the result with sufficient detail.

This explains why we often iterate to get what we’re looking for when using LLMs and generative AI. Going back to our cheeseburger analogy, the process to generate a cheeseburger from an LLM may look like this.

  • “Make me a cheeseburger, medium rare, with lettuce and tomatoes.” The result also has pickles and uses cheddar cheese. The bun is toasted. There’s mayo on the top bun.
  • “Make the same thing but this time no pickles, use pepper jack cheese, and a sriracha mayo instead of plain mayo.” The result now has pepper jack, no pickles. The sriracha mayo is applied to the bottom bun and the bun is no longer toasted.
  • “Make the same thing again, but this time, put the sriracha mayo on the top bun. The buns should be toasted.” Finally, you have the cheeseburger you’re looking for.

This example demonstrates one of the main points of friction with human-AI interaction. Human beings are really bad at describing what they want with sufficient detail on the first attempt.

When we asked for a cheeseburger, we had to refine our description to be more specific (the type of cheese). In the second generation, some of the inferred details (whether the bun was toasted) changed from one iteration to the next, so then we had to add that specificity to our description as well. Iteration is an important part of AI-human generation.

Insight: When using generative AI, we need to design an iterative human-AI interaction loop that enables people to discover the details of what they want and refine their descriptions accordingly.

To iterate, we need to evaluate the results. Evaluation is extremely important with generative AI. Say you’re using an LLM to write code. You can evaluate the code quality if you know enough to understand it or if you can execute it and inspect the results. On the other hand, hypothetical questions can’t be tested. Say you ask ChatGPT, “What if we raise our product prices by 5 percent?” A seasoned expert could read the output and know from experience if a recommendation doesn’t take into account important details. If your product is property insurance, then increasing premiums by 5 percent may mean pushback from regulators, something an experienced veteran of the industry would know. For non-experts in a topic, there’s no way to tell if the “average” details inferred by the model make sense for your specific use case. You can’t test and iterate.

Insight: LLMs work best when the user can evaluate the result quickly, whether through execution or through prior knowledge.

The examples so far involve general knowledge. We all know what a cheeseburger is. When you start asking about non-general information—like when you can make dinner reservations next week—you delve into new points of friction.

In the next section we’ll think about different types of information, what we can expect the AI to “know”, and how this impacts human-AI interaction.

What did the AI know, and when did it know it?

Above, I explained how generative AI is a declarative mode of interaction and how that helps understand its strengths and weaknesses. Here, I’ll identify how different types of information create better or worse human-AI interactions.

Understanding the information available

When we describe what we want to an LLM, and when it infers missing details from our description, it draws from different sources of information. Understanding these sources of information is important. Here’s a useful taxonomy for information types:

  • General information used to train the base model.
  • Non-general information that the base model is not aware of.
    • Fresh information that is new or changes rapidly, like stock prices or current events.
    • Non-public information, like facts about you and where you live or about your company, its employees, its processes, or its codebase.

General information vs. non-general information

LLMs are built on a massive corpus of written word data. A large part of GPT-3 was trained on a combination of books, journals, Wikipedia, Reddit, and CommonCrawl (an open-source repository of web crawl data). You can think of the models as a highly compressed version of that data, organized in a gestalt manner—all the like things are close together. When we submit a prompt, the model takes the words we use (and any words added to the prompt behind the scenes) and finds the closest set of related words based on how those things appear in the data corpus. So when we say “cheeseburger” it knows that word is related to “bun” and “tomato” and “lettuce” and “pickles” because they all occur in the same context throughout many data sources. Even when we don’t specify pickles, it uses this gestalt approach to fill in the blanks.

This training information is general information, and a good rule of thumb is this: if it was in Wikipedia a year ago then the LLM “knows” about it. There could be new articles on Wikipedia, but that didn’t exist when the model was trained. The LLM doesn’t know about that unless told.

Now, say you’re a company using an LLM to write a product requirements document for a new web app feature. Your company, like most companies, is full of its own lingo. It has its own lore and history scattered across thousands of Slack messages, emails, documents, and some tenured employees who remember that one meeting in Q1 last year. The LLM doesn’t know any of that. It will infer any missing details from general information. You need to supply everything else. If it wasn’t in Wikipedia a year ago, the LLM doesn’t know about it. The resulting product requirements document may be full of general facts about your industry and product but could lack important details specific to your firm.

This is non-general information. This includes personal info, anything kept behind a log-in or paywall, and non-digital information. This non-general information permeates our lives, and incorporating it is another source of friction when working with generative AI.

Non-general information can be incorporated into a generative AI application in three ways:

  • Through model fine-tuning (supplying a large corpus to the base model to expand its reference data).
  • Retrieved and fed it to the model at query time (e.g., the retrieval augmented generation or “RAG” technique).
  • Supplied by the user in the prompt.

Insight: When designing any human-AI interactions, you should think about what non-general information is required, where you will get it, and how you will expose it to the AI.

Fresh information

Any information that changes in real-time or is new can be called fresh information. This includes new facts like current events but also frequently changing facts like your bank account balance. If the fresh information is available in a database or some searchable source, then it needs to be retrieved and incorporated into the application. To retrieve the information from a database, the LLM must create a query, which may require specific details that the user didn’t include.

Here’s an example. I have a chatbot that gives information on the stock market. You, the user, type the following: “What is the current price of Apple? Has it been increasing or decreasing recently?”

  • The LLM doesn’t have the current price of Apple in its training data. This is fresh, non-general information. So, we need to retrieve it from a database.
  • The LLM can read “Apple”, know that you’re talking about the computer company, and that the ticker symbol is AAPL. This is all general information.
  • What about the “increasing or decreasing” part of the prompt? You did not specify over what period—increasing in the past day, month, year? In order to construct a database query, we need more detail. LLMs are bad at knowing when to ask for detail and when to fill it in. The application could easily pull the wrong data and provide an unexpected or inaccurate answer. Only you know what these details should be, depending on your intent. You must be more specific in your prompt.

A designer of this LLM application can improve the user experience by specifying required parameters for expected queries. We can ask the user to explicitly input the time range or design the chatbot to ask for more specific details if not provided. In either case, we need to have a specific type of query in mind and explicitly design how to handle it. The LLM will not know how to do this unassisted.

Insight: If a user is expecting a more specific type of output, you need to explicitly ask for enough detail. Too little detail could produce a poor quality output.

Non-public information

Incorporating non-public information into an LLM prompt can be done if that information can be accessed in a database. This introduces privacy issues (should the LLM be able to access my medical records?) and complexity when incorporating multiple non-public sources of information.

Let’s say I have a chatbot that helps you make dinner reservations. You, the user, type the following: “Help me make dinner reservations somewhere with good Neapolitan pizza.”

  • The LLM knows what a Neapolitan pizza is and can infer that “dinner” means this is for an evening meal.
  • To do this task well, it needs information about your location, the restaurants near you and their booking status, or even personal details like dietary restrictions. Assuming all that non-public information is available in databases, bringing them all together into the prompt takes a lot of engineering work.
  • Even if the LLM could find the “best” restaurant for you and book the reservation, can you be confident it has done that correctly? You never specified how many people you need a reservation for. Since only you know this information, the application needs to ask for it upfront.

If you’re designing this LLM-based application, you can make some thoughtful choices to help with these problems. We could ask about a user’s dietary restrictions when they sign up for the app. Other information, like the user’s schedule that evening, can be given in a prompting tip or by showing the default prompt option “show me reservations for two for tomorrow at 7PM”. Promoting tips may not feel as automagical as a bot that does it all, but they are a straightforward way to collect and integrate the non-public information.

Some non-public information is large and can’t be quickly collected and processed when the prompt is given. These need to be fine-tuned in batch or retrieved at prompt time and incorporated. A chatbot that answers information about a company’s HR policies can obtain this information from a corpus of non-public HR documents. You can fine-tune the model ahead of time by feeding it the corpus. Or you can implement a retrieval augmented generation technique, searching a corpus for relevant documents and summarizing the results. Either way, the response will only be as accurate and up-to-date as the corpus itself.

Insight: When designing an AI application, you need to be aware of non-public information and how to retrieve it. Some of that information can be pulled from databases. Some needs to come from the user, which may require prompt suggestions or explicitly asking.

If you understand the types of information and treat human-AI interaction as declarative, you can more easily predict which AI applications will work and which ones won’t. In the next section we’ll look at OpenAI’s Operator and deep research products. Using this framework, we can see where these applications fall short, where they work well, and why.

Critiquing OpenAI’s Operator and deep research through a declarative lens

I have now explained how thinking of generative AI as declarative helps us understand its strengths and weaknesses. I also identified how different types of information create better or worse human-AI interactions.

Now I’ll apply these ideas by critiquing two recent products from OpenAI—Operator and deep research. It’s important to be honest about the shortcomings of AI applications. Bigger models trained on more data or using new techniques might one day solve some issues with generative AI. But other issues arise from the human-AI interaction itself and can only be addressed by making appropriate design and product choices.

These critiques demonstrate how the framework can help identify where the limitations are and how to address them.

The limitations of Operator

Journalist Casey Newton of Platformer reviewed Operator in an article that was largely positive. Newton has covered AI extensively and optimistically. Still, Newton couldn’t help but point out some of Operator’s frustrating limitations.

[Operator] can take action on your behalf in ways that are new to AI systems — but at the moment it requires a lot of hand-holding, and may cause you to throw up your hands in frustration. 

My most frustrating experience with Operator was my first one: trying to order groceries. “Help me buy groceries on Instacart,” I said, expecting it to ask me some basic questions. Where do I live? What store do I usually buy groceries from? What kinds of groceries do I want? 

It didn’t ask me any of that. Instead, Operator opened Instacart in the browser tab and begin searching for milk in grocery stores located in Des Moines, Iowa.

The prompt “Help me buy groceries on Instacart,” viewed declaratively, describes groceries being purchased using Instacart. It doesn’t have a lot of the information someone would need to buy groceries, like what exactly to buy, when it would be delivered, and to where.

It’s worth repeating: LLMs are not good at knowing when to ask additional questions unless explicitly programmed to do so in the use case. Newton gave a vague request and expected follow-up questions. Instead, the LLM filled in all the missing details with the “average”. The average item was milk. The average location was Des Moines, Iowa. Newton doesn’t mention when it was scheduled to be delivered, but if the “average” delivery time is tomorrow, then that was likely the default.

If we engineered this application specifically for ordering groceries, keeping in mind the declarative nature of AI and the information it “knows”, then we could make thoughtful design choices that improve functionality. We would need to prompt the user to specify when and where they want groceries up front (non-public information). With that information, we could find an appropriate grocery store near them. We would need access to that grocery store’s inventory (more non-public information). If we have access to the user’s previous orders, we could also pre-populate a cart with items typical to their order. If not, we may add a few suggested items and guide them to add more. By limiting the use case, we only have to deal with two sources of non-public information. This is a more tractable problem than Operator’s “agent that does it all” approach.

Newton also mentions that this process took eight minutes to complete, and “complete” means that Operator did everything up to placing the order. This is a long time with very little human-in-the-loop iteration. Like we said before, an iteration loop is very important for human-AI interaction. A better-designed application would generate smaller steps along the way and provide more frequent interaction. We could prompt the user to describe what to add to their shopping list. The user might say, “Add barbeque sauce to the list,” and see the list update. If they see a vinegar-based barbecue sauce, they can refine that by saying, “Replace that with a barbeque sauce that goes well with chicken,” and might be happier when it’s replaced by a honey barbecue sauce. These frequent iterations make the LLM a creative tool rather than a does-it-all agent. The does-it-all agent looks automagical in marketing, but a more guided approach provides more utility with a less frustrating and more delightful experience.

Elsewhere in the article, Newton gives an example of a prompt that Operator performed well: “Put together a lesson plan on the Great Gatsby for high school students, breaking it into readable chunks and then creating assignments and connections tied to the Common Core learning standard.” This prompt describes an output using much more specificity. It also solely relies on general information—the Great Gatsby, the Common Core standard, and a general sense of what assignments are. The general-information use case lends itself better to AI generation, and the prompt is explicit and detailed in its request. In this case, very little guidance was given to create the prompt, so it worked better. (In fact, this prompt comes from Ethan Mollick who has used it to evaluate AI chatbots.)

This is the risk of general-purpose AI applications like Operator. The quality of the result relies heavily on the use case and specificity provided by the user. An application with a more specific use case allows for more design guidance and can produce better output more reliably.

The limitations of deep research

Newton also reviewed deep research, which, according to OpenAI’s website, is an “agent that uses reasoning to synthesize large amounts of online information and complete multi-step research tasks for you.”

Deep research came out after Newton’s review of Operator. Newton chose an intentionally tricky prompt that prods at some of the tool’s limitations regarding fresh information and non-general information: “I wanted to see how OpenAI’s agent would perform given that it was researching a story that was less than a day old, and for which much of the coverage was behind paywalls that the agent would not be able to access. And indeed, the bot struggled more than I expected.”

Near the end of the article, Newton elaborates on some of the shortcomings he noticed with deep research.

OpenAI’s deep research suffers from the same design problem that almost all AI products have: its superpowers are completely invisible and must be harnessed through a frustrating process of trial and error.

Generally speaking, the more you already know about something, the more useful I think deep research is. This may be somewhat counterintuitive; perhaps you expected that an AI agent would be well suited to getting you up to speed on an important topic that just landed on your lap at work, for example. 

In my early tests, the reverse felt true. Deep research excels for drilling deep into subjects you already have some expertise in, letting you probe for specific pieces of information, types of analysis, or ideas that are new to you.

The “frustrating trial and error” shows a mismatch between Newton’s expectations and a necessary aspect of many generative AI applications. A good response requires more information than the user will probably give in the first attempt. The challenge is to design the application and set the user’s expectations so that this interaction is not frustrating but exciting.

Newton’s more poignant criticism is that the application requires already knowing something about the topic for it to work well. From the perspective of our framework, this makes sense. The more you know about a topic, the more detail you can provide. And as you iterate, having knowledge about a topic helps you observe and evaluate the output. Without the ability to describe it well or evaluate the results, the user is less likely to use the tool to generate good output.

A version of deep research designed for lawyers to perform legal research could be powerful. Lawyers have an extensive and common vocabulary for describing legal matters, and they’re more likely to see a result and know if it makes sense. Generative AI tools are fallible, though. So, the tool should focus on a generation-evaluation loop rather than writing a final draft of a legal document.

The article also highlights many improvements compared to Operator. Most notably, the bot asked clarifying questions. This is the most impressive aspect of the tool. Undoubtedly, it helps that deep search has a focused use-case of retrieving and summarizing general information instead of a does-it-all approach. Having a focused use case narrows the set of likely interactions, letting you design better guidance into the prompt flow.

Good application design with generative AI

Designing effective generative AI applications requires thoughtful consideration of how users interact with the technology, the types of information they need, and the limitations of the underlying models. Here are some key principles to guide the design of generative AI tools:

1. Constrain the input and focus on providing details

Applications are inputs and outputs. We want the outputs to be useful and pleasant. By giving a user a conversational chatbot interface, we allow for a vast surface area of potential inputs, making it a challenge to guarantee useful outputs. One strategy is to limit or guide the input to a more manageable subset.

For example, FigJam, a collaborative whiteboarding tool, uses pre-set template prompts for timelines, Gantt charts, and other common whiteboard artifacts. This provides some structure and predictability to the inputs. Users still have the freedom to describe further details like color or the content for each timeline event. This approach ensures that the AI has enough specificity to generate meaningful outputs while giving users creative control.

2. Design frequent iteration and evaluation into the tool

Iterating in a tight generation-evaluation loop is essential for refining outputs and ensuring they meet user expectations. OpenAI’s Dall-E is great at this. Users quickly iterate on image prompts and refine their descriptions to add additional detail. If you type “a picture of a cheeseburger on a plate”, you may then add more detail by specifying “with pepperjack cheese”.

AI code generating tools work well because users can run a generated code snippet immediately to see if it works, enabling rapid iteration and validation. This quick evaluation loop produces better results and a better coder experience. 

Designers of generative AI applications should pull the user in the loop early, often, in a way that is engaging rather than frustrating. Designers should also consider the user’s knowledge level. Users with domain expertise can iterate more effectively.

Referring back to the FigJam example, the prompts and icons in the app quickly communicate “this is what we call a mind map” or “this is what we call a gantt chart” for users who want to generate these artifacts but don’t know the terms for them. Giving the user some basic vocabulary can help them better generate desired results quickly with less frustration.

3. Be mindful of the types of information needed

LLMs excel at tasks involving general knowledge already in the base training set. For example, writing class assignments involves absorbing general information, synthesizing it, and producing a written output, so LLMs are very well-suited for that task.

Use cases that require non-general information are more complex. Some questions the designer and engineer should ask include:

  • Does this application require fresh information? Maybe this is knowledge of current events or a user’s current bank account balance. If so, that information needs to be retrieved and incorporated into the model.
  • How much non-general information does the LLM need to know? If it’s a lot of information—like a corpus of company documentation and communication—then the model may need to be fine tuned in batch ahead of time. If the information is relatively small, a retrieval augmented generation (RAG) approach at query time may suffice. 
  • How many sources of non-general information—small and finite or potentially infinite? General purpose agents like Operator face the challenge of potentially infinite non-general information sources. Depending on what the user requires, it could need to access their contacts, restaurant reservation lists, financial data, or even other people’s calendars. A single-purpose restaurant reservation chatbot may only need access to Yelp, OpenTable, and the user’s calendar. It’s much easier to reconcile access and authentication for a handful of known data sources.
  • Is there context-specific information that can only come from the user? Consider our restaurant reservation chatbot. Is the user making reservations for just themselves? Probably not. “How many people and who” is a detail that only the user can provide, an example of non-public information that only the user knows. We shouldn’t expect the user to provide this information upfront and unguided. Instead, we can use prompt suggestions so they include the information. We may even be able to design the LLM to ask these questions when the detail is not provided.

4. Focus on specific use cases

Broad, all-purpose chatbots often struggle to deliver consistent results due to the complexity and variability of user needs. Instead, focus on specific use cases where the AI’s shortcomings can be mitigated through thoughtful design.

Narrowing the scope helps us address many of the issues above.

  • We can identify common requests for the use case and incorporate those into prompt suggestions.
  • We can design an iteration loop that works well with the type of thing we’re generating.
  • We can identify sources of non-general information and devise solutions to incorporate it into the model or prompt.

5. Translation or summary tasks work well

A common task for ChatGPT is to rewrite something in a different style, explain what some computer code is doing, or summarize a long document. These tasks involve converting a set of information from one form to another.

We have the same concerns about non-general information and context. For instance, a Chatbot asked to explain a code script doesn’t know the system that script is part of unless that information is provided.

But in general, the task of transforming or summarizing information is less prone to missing details. By definition, you have provided the details it needs. The result should have the same information in a different or more condensed form.

The exception to the rules

There is a case when it doesn’t matter if you break any or all of these rules—when you’re just having fun. LLMs are creative tools by nature. They can be an easel to paint on, a sandbox to build in, a blank sheet to scribe. Iteration is still important; the user wants to see the thing they’re creating as they create it. But unexpected results due to lack of information or omitted details may add to the experience. If you ask for a cheeseburger recipe, you might get some funny or interesting ingredients. If the stakes are low and the process is its own reward, don’t worry about the rules.

Shape
Shape
Stay Ahead

Explore More Insights

Stay ahead with more perspectives on cutting-edge power, infrastructure, energy,  bitcoin and AI solutions. Explore these articles to uncover strategies and insights shaping the future of industries.

Shape

Mobile demands spur enterprise Wi-Fi upgrades

Performance requirements (reduced latency, jitter, packet drops): 67.1% Increased bandwidth consumption: 59.9% User mobility (roaming, broader coverage): 53.3% Connectivity for operational technology (industrial systems, medical imaging, video surveillance): 50.0% Return-to-office policies driving up office occupancy: 48.7% Density requirements (users congregating): 44.7% End-of-life network equipment: 40.8% Need for location-based services: 40.1%

Read More »

Anthropic signs billion-dollar deal with Google Cloud

US-based AI company Anthropic has signed a major deal with Google Cloud that is said to be worth tens of billions of dollars. As part of the deal, Anthropic will have access to up to one million of Google’s purpose-built Tensor Processing Unit (TPU) AI accelerators. “Anthropic and Google have

Read More »

Crude Pauses After Sanction-Fueled Rally

Oil swung between gains and losses on the back of the biggest weekly increase since June as attention turned to the wider outlook for supply as the US and China made progress on trade. West Texas Intermediate slipped 0.3% to settle just above $61, extending its decline for a second day. Top Chinese and US negotiators said they came to terms on a range of points, setting the table for President Donald Trump and counterpart Xi Jinping to finalize a deal to ease trade tensions between the world’s two biggest economies and crude importers. Still, oil was little changed Monday after adding nearly 7% last week when the US sanctioned Russian oil giants Rosneft and Lukoil to squeeze Russia over its ongoing war in Ukraine. The move added output risks to a market that’s showing signs of entering a surplus. Lukoil announced in a statement that it intends to sell its international assets due to the latest sanctions. “The market is taking a breather here,” said Dennis Kissler, senior vice president for trading at BOK Financial. “While US-China negotiations continue, no real outcome has been agreed as of yet…and the sanctions on Russia may halt some shipments though it’s more likely most of that oil will still find a home.” The Trump administration is seeking to make Russia’s trade harder, costlier and riskier, but without forcing a sudden supply shock that might spike global oil prices, officials familiar with the matter said over the weekend. The measures helped oil rebound from a five-month low last week, but part of the move was likely driven by extreme market positioning. Traders had amassed record bearish wagers on the global Brent benchmark in anticipation of oversupply in the next few months. In the meantime, commodity trading advisers, or CTAs, are set to accelerate

Read More »

Petrofac’s 7,300 Jobs at Risk

UK energy services provider Petrofac Ltd has applied to enter administration after the company’s latest plans to restructure its balance sheet unexpectedly fell through, putting thousands of jobs at risk. The company has applied to the High Court of England and Wales to appoint administrators, according to a statement on Monday. It comes after European grid operator TenneT canceled Petrofac’s work at a large offshore energy project in the North Sea, rendering the financial restructuring unviable.  The firm employs around 7,300 people globally and has been trying to strike a deal with creditors for over a year, while safeguarding key business deals that could keep the firm afloat. Its collapse would raise the pressure on the UK government to protect British jobs, with the Labour administration already coming under fire for blocking new North Sea oil licenses. The filing from the company, which employs about 2,000 people in the UK, adds to a string of challenges facing the government, including a bruising by-election defeat, pressure to raise taxes in the autumn budget, and ongoing public outrage over troubled utility Thames Water, Britain’s biggest and most indebted water supplier.  A representative for the Department for Energy Security and Net Zero said that Petrofac’s administration is a “product of longstanding issues in their global business,” adding that the UK arm is continuing to operate as normal. “The government will continue to work with the UK company as it focuses on its long-term future,” the spokesperson said.  Meanwhile, alternative restructuring and M&A solutions for the group are still being explored with creditors including its bondholder group, Petrofac said. Key Contract Petrofac’s business with TenneT was particularly crucial given it represents over 80% of revenue in the group’s engineering and construction division, according to court documents filed earlier this year. But since Petrofac was not able to meet

Read More »

Energy Department Announces New Public-Private Partnership Model, Two Supercomputers, to Accelerate American Dominance in Science and Technology

WASHINGTON— The U.S. Department of Energy (DOE) today announced two new AMD-accelerated artificial intelligence (AI) supercomputers at Oak Ridge National Laboratory (ORNL), one of which will be built at record speeds thanks to a new public-private partnership model. The supercomputers will help expand America’s leadership in scientific computing, strengthen national security, and drive the next generation of Gold Standard Science and innovation. With the new public-private partnership model, the Lux AI cluster, powered by AMD Instinct™ MI355X GPUs, AMD EPYC™ CPUs, and AMD Pensando™ advanced networking, will be deployed in early 2026 to expand DOE’s near-term AI capacity and accelerate work on critical national priorities, including fusion, fission, materials discovery, quantum, advanced manufacturing, and grid modernization. Lux will provide a secure, open, and efficient AI software stack to strengthen America’s innovation base and enhance U.S. competitiveness. “Winning the AI race requires new and creative partnerships that will bring together the brightest minds and industries American technology and science has to offer,” said U.S. Secretary of Energy Chris Wright. “That’s why the Trump administration is announcing the first example of a new commonsense approach to computing partnerships with Lux. We are also announcing, as part of a competitive procurement process, Discovery. Working with AMD and HPE, we’re bringing new capacity online faster than ever before, turning shared innovation into national strength, and proving that America leads when private-public partners build together.” “We are proud and honored to partner with the U.S. Department of Energy and Secretary Wright to accelerate America’s AI compute infrastructure,” said AMD chair and CEO Dr. Lisa Su. “This partnership exemplifies public-private collaboration at its best. With Discovery and Lux, we are delivering leadership compute systems that combine performance and energy efficiency to advance America’s research priorities and strengthen U.S. leadership in AI, energy, and national security.” DOE

Read More »

US electric utilities entering investment ‘super cycle,’ says Morningstar DBRS

U.S. electric utilities are entering a five-year capital expenditure “super-cycle” as they build out transmission and generation networks to meet new demand from data centers, Morningstar DBRS said in a Monday research note. “Investment in electricity infrastructure is projected to be $1.4 trillion from 2025 to 2030, double the amount invested in the prior 10 years,” the firm said. And based on data from the North American Electric Reliability Corp., Morningstar said many regions can expect load growth to increase from previously estimated 6.1% to around 11.6% over the next decade. “The challenges posed by the rapid buildout of data centers are overlaid on existing concerns for most utilities, including decarbonization and guaranteeing the reliability of grid infrastructure while increasing the contribution of renewable power,” Morningstar said. “The projected surge in demand provides opportunities for utilities that must be counterbalanced with structural changes and regulatory support for utilities and rate payers.” “We anticipate that regulated utilities with supportive regulatory commissions, solid credit ratings, and access to capital markets will deploy the needed capex to take advantage of the data center boom,” Bukola Folashakin, Morningstar assistant vice president of corporate ratings, said in a statement. “We expect this capex investment in turn to make such utility locations attractive for more data center construction, potentially creating a cycle of increased revenue for as long as data centers remain economically viable.” Some states — Morningstar named California, Texas, and Louisiana — will see elevated risks of resource inadequacy next year that “in extreme conditions, could trigger electricity shortfalls,” according to the note. Morningstar’s analysis mirror’s recent capital expenditure estimates from the Edison Electric Institute, which expects U.S electricity generation to grow “for the foreseeable future.” The group, which represents investor-owned utilities, said generation rose 3% in 2024 and generation investments as a share of

Read More »

Residential electricity prices up more than 6% in August: EIA

Dive Brief: The retail price of electricity for residential customers was up 6.1% in August over the same month last year, reaching an average of 17.62 cents/kWh, according to the U.S. Energy Information Administration’s monthly update on electricity released Friday.  Total net generation and retail sales were down slightly, by 0.8% and 0.5% respectively, which experts attributed to a mild summer in much of the eastern part of the country. Nationally, there were 9.1% fewer cooling days this August compared to August 2024, though parts of the West saw between 5% and 50% more.  Natural gas prices were up 47.4% to just over $3/MMBtu, per Henry Hub, and natural gas consumption fell 5.6%.  Dive Insight: Aaron Denman, who heads Bain & Company’s Americas utilities and renewables practice, said the reasons behind the rise in prices are complicated and nuanced, and vary depending on region. “I think if you look at aggregate headline numbers, it’s hard to blame data centers for much of the increase, and I think part of that is just because it’s still such a small percentage of the overall load,” Denman said. “But I think when you get into specific regions, they’re so large, they’re so meaningful, that you can start to see some of that impact.” He said Bain’s projections put data centers at about 40% to 60% of new load growth in the coming years, with manufacturing and electrification making up most of the remainder. He said the longterm fundamentals remain strong for load growth, despite some recent negative economic data. U.S. factory activity shrank in September for a seventh consecutive month, according to Bloomberg. “This [much] load growth, this fast, is going to put upward pressure on residential prices,” Denman said. “We’re going to need to see creative solutions to minimize the impact on particularly residential

Read More »

EIA Shows Production Has Outweighed Demand All Year

In its latest short term energy outlook (STEO), which was released on October 7, the U.S. Energy Information Administration (EIA) showed that world petroleum and other liquid fuels production outweighed consumption in the first, second, and third quarters of this year. According to the STEO, in the third quarter, production averaged 107.43 million barrels per day and consumption averaged 104.83 million barrels per day. In the second quarter, production came in at 105.06 million barrels per day and consumption was 104.05 million barrels per day, and in the first quarter, production averaged 103.62 million barrels per day and consumption was 102.33 million barrels per day, the STEO highlighted. The last time consumption came in higher than demand was in the third quarter of 2024, according to the STEO, which showed that, in that quarter, production averaged 103.09 million barrels per day and consumption averaged 103.45 million barrels per day. In the first quarter of last year, production was 102.60 million barrels per day and consumption was 101.79 million barrels per day, in the second quarter of 2024 production was 103.23 million barrels per day and consumption was 102.93 million barrels per day, and in the fourth quarter of last year production was 103.83 million barrels per day and consumption was 103.46 million barrels per day, the STEO showed. Looking ahead, the STEO projected that production will average 107.31 million barrels per day in the fourth quarter of this year. The EIA forecast in the STEO that consumption will come in at 104.72 million barrels per day in the fourth quarter. In its latest STEO, the EIA projected that production will average 106.39 million barrels per day in the first quarter of next year, 106.97 million barrels per day in the second quarter, 107.55 million barrels per day in the third

Read More »

Qualcomm goes all-in on inferencing with purpose-built cards and racks

From a strategy perspective, there is a longer term enterprise play here, noted Moor’s Kimball; Humain is Qualcomm’s first customer, and a cloud service provider (CSP) or hyperscaler will likely be customer number two. However, at some point, these rack-scale systems will find their way into the enterprise. “If I were the AI200 product marketing lead, I would be thinking about how I demonstrate this as a viable platform for those enterprise workloads that will be getting ‘agentified’ over the next several years,” said Kimball. It seems a natural step, as Qualcomm saw success with its AI100 accelerator, a strong inference chip, he noted. Right now, Nvidia and AMD dominate the training market, with CUDA and ROCm enjoying a “stickiness” with customers. “If I am a semiconductor giant like Qualcomm that is so good at understanding the performance-power balance, this inference market makes perfect sense to really lean in on,” said Kimball. He also pointed to the company’s plans to re-enter the datacenter CPU space with its Oryon CPU, which is featured in Snapdragon and loosely based on technology it acquired with its $1.4 billion Nuvia acquisition. Ultimately, Qualcomm’s move demonstrates how wide open the inference market is, said Kimball. The company, he noted, has been very good at choosing target markets and has seen success when entering those markets. “That the company would decide to go more ‘in’ on the inference market makes sense,” said Kimball. He added that, from an ROI perspective, inferencing will “dwarf” training in terms of volume and dollars.

Read More »

AI data center building boom risks fueling future debt bust, bank warns

However, that’s only one part of the problem. Meeting the power demands of AI data centers will require the energy sector to make large investments. Then there’s data center demand for microprocessors, rare earth elements, and other valuable metals such as copper, which could, in a bust, make data centers the most expensively-assembled unwanted assets in history. “Financial stability consequences of an AI-related asset price fall could arise through multiple channels. If forecasted debt-financed AI infrastructure growth materializes, the potential financial stability consequences of such an event are likely to grow,” warned the BoE blog post. “For companies who depend on the continued demand for massive computational capacity to train and run inference on AI models, an algorithmic breakthrough or other event which challenges that paradigm could cause a significant re-evaluation of asset prices,” it continued. According to Matt Hasan, CEO of AI consultancy aiRESULTS, the underlying problem is the speed with which AI has emerged. “What we’re witnessing isn’t just an incremental expansion, it’s a rush to construct power-hungry, mega-scale data centers,” he told Network World. The dot.com reversal might be the wrong comparison; it dented the NASDAQ and hurt tech investment, but the damage to organizations investing in e-commerce was relatively limited. AI, by contrast, might have wider effects for large enterprises because so many have pinned their business prospects on its potential. “Your reliance on these large providers means you are indirectly exposed to the stability of their debt. If a correction occurs, the fallout can impact the services you rely on,” said Hasan.

Read More »

Intel sees supply shortage, will prioritize data center technology

“Capacity constraints, especially on Intel 10 and Intel 7 [Intel’s semiconductor manufacturing process], limited our ability to fully meet demand in Q3 for both data center and client products,” said Zinsner, adding that Intel isn’t about to add capacity to Intel 10 and 7 when it has moved beyond those nodes. “Given the current tight capacity environment, which we expect to persist into 2026, we are working closely with customers to maximize our available output, including adjusting pricing and mix to shift demand towards products where we have supply and they have demand,” said Zinsner. For that reason, Zinzner projects that the fourth quarter will be roughly flat versus the third quarter in terms of revenue. “We expect Intel products up modestly sequentially but below customer demand as we continue to navigate supply environment,” said Zinsner. “We expect CCG to be down modestly and PC AI to be up strongly sequentially as we prioritize wafer capacity for server shipments over entry-level client parts.”

Read More »

How to set up an AI data center in 90 days

“Personally, I think that a brownfield is very creative way to deal with what I think is the biggest problem that we’ve got right now, which is time and speed to market,” he said. “On a brownfield, I can go into a building that’s already got power coming into the building. Sometimes they’ve already got chiller plants, like what we’ve got with the building I’m in right now.” Patmos certainly made the most of the liquid facilities in the old printing press building. The facility is built to handle anywhere from 50 to over 140 kilowatts per cabinet, a leap far beyond the 1–2 kW densities typical of legacy data centers. The chips used in the servers are Nvidia’s Grace Blackwell processors, which run extraordinarily hot. To manage this heat load, Patmos employs a multi-loop liquid cooling system. The design separates water sources into distinct, closed loops, each serving a specific function and ensuring that municipal water never directly contacts sensitive IT equipment. “We have five different, completely separated water loops in this building,” said Morgan. “The cooling tower uses city water for evaporation, but that water never mixes with the closed loops serving the data hall. Everything is designed to maximize efficiency and protect the hardware.” The building taps into Kansas City’s district chilled water supply, which is sourced from a nearby utility plant. This provides the primary cooling resource for the facility. Inside the data center, a dedicated loop circulates a specialized glycol-based fluid, filtered to extremely low micron levels and formulated to be electronically safe. Heat exchangers transfer heat from the data hall fluid to the district chilled water, keeping the two fluids separate and preventing corrosion or contamination. Liquid-to-chip and rear-door heat exchangers are used for immediate heat removal.

Read More »

INNIO and VoltaGrid: Landmark 2.3 GW Modular Power Deal Signals New Phase for AI Data Centers

Why This Project Marks a Landmark Shift The deployment of 2.3 GW of modular generation represents utility-scale capacity, but what makes it distinct is the delivery model. Instead of a centralized plant, the project uses modular gas-reciprocating “power packs” that can be phased in step with data-hall readiness. This approach allows staged energization and limits the bottlenecks that often stall AI campuses as they outgrow grid timelines or wait in interconnection queues. AI training loads fluctuate sharply, placing exceptional stress on grid stability and voltage quality. The INNIO/VoltaGrid platform was engineered specifically for these GPU-driven dynamics, emphasizing high transient performance (rapid load acceptance) and grid-grade power quality, all without dependence on batteries. Each power pack is also designed for maximum permitting efficiency and sustainability. Compared with diesel generation, modern gas-reciprocating systems materially reduce both criteria pollutants and CO₂ emissions. VoltaGrid markets the configuration as near-zero criteria air emissions and hydrogen-ready, extending allowable runtimes under air permits and making “prime-as-a-service” viable even in constrained or non-attainment markets. 2025: Momentum for Modular Prime Power INNIO has spent 2025 positioning its Jenbacher platform as a next-generation power solution for data centers: combining fast start, high transient performance, and lower emissions compared with diesel. While the 3 MW J620 fast-start lineage dates back to 2019, this year the company sharpened its data center narrative and booked grid stability and peaking projects in markets where rapid data center growth is stressing local grids. This momentum was exemplified by an 80 MW deployment in Indonesia announced earlier in October. The same year saw surging AI-driven demand and INNIO’s growing push into North American data-center markets. Specifications for the 2.3 GW VoltaGrid package highlight the platform’s heat tolerance, efficiency, and transient response, all key attributes for powering modern AI campuses. VoltaGrid’s 2025 Milestones VoltaGrid’s announcements across 2025 reflect

Read More »

Inside Google’s multi-architecture revolution: Axion Arm joins x86 in production clusters

Matt Kimball, VP and principal analyst with Moor Insights and Strategy, pointed out that AWS and Microsoft have already moved many workloads from x86 to internally designed Arm-based servers. He noted that, when Arm first hit the hyperscale datacenter market, the architecture was used to support more lightweight, cloud-native workloads with an interpretive layer where architectural affinity was “non-existent.” But now there’s much more focus on architecture, and compatibility issues “largely go away” as Arm servers support more and more workloads. “In parallel, we’ve seen CSPs expand their designs to support both scale out (cloud-native) and traditional scale up workloads effectively,” said Kimball. Simply put, CSPs are looking to monetize chip investments, and this migration signals that Google has found its performance-per-dollar (and likely performance-per-watt) better on Axion than x86. Google will likely continue to expand its Arm footprint as it evolves its Axion chip; as a reference point, Kimball pointed to AWS Graviton, which didn’t really support “scale up” performance until its v3 or v4 chip. Arm is coming to enterprise data centers too When looking at architectures, enterprise CIOs should ask themselves questions such as what instance do they use for cloud workloads, and what servers do they deploy in their data center, Kimball noted. “I think there is a lot less concern about putting my workloads on an Arm-based instance on Google Cloud, a little more hesitance to deploy those Arm servers in my datacenter,” he said. But ultimately, he said, “Arm is coming to the enterprise datacenter as a compute platform, and Nvidia will help usher this in.” Info-Tech’s Jain agreed that Nvidia is the “biggest cheerleader” for Arm-based architecture, and Arm is increasingly moving from niche and mobile use to general-purpose and AI workload execution.

Read More »

Microsoft will invest $80B in AI data centers in fiscal 2025

And Microsoft isn’t the only one that is ramping up its investments into AI-enabled data centers. Rival cloud service providers are all investing in either upgrading or opening new data centers to capture a larger chunk of business from developers and users of large language models (LLMs).  In a report published in October 2024, Bloomberg Intelligence estimated that demand for generative AI would push Microsoft, AWS, Google, Oracle, Meta, and Apple would between them devote $200 billion to capex in 2025, up from $110 billion in 2023. Microsoft is one of the biggest spenders, followed closely by Google and AWS, Bloomberg Intelligence said. Its estimate of Microsoft’s capital spending on AI, at $62.4 billion for calendar 2025, is lower than Smith’s claim that the company will invest $80 billion in the fiscal year to June 30, 2025. Both figures, though, are way higher than Microsoft’s 2020 capital expenditure of “just” $17.6 billion. The majority of the increased spending is tied to cloud services and the expansion of AI infrastructure needed to provide compute capacity for OpenAI workloads. Separately, last October Amazon CEO Andy Jassy said his company planned total capex spend of $75 billion in 2024 and even more in 2025, with much of it going to AWS, its cloud computing division.

Read More »

John Deere unveils more autonomous farm machines to address skill labor shortage

Join our daily and weekly newsletters for the latest updates and exclusive content on industry-leading AI coverage. Learn More Self-driving tractors might be the path to self-driving cars. John Deere has revealed a new line of autonomous machines and tech across agriculture, construction and commercial landscaping. The Moline, Illinois-based John Deere has been in business for 187 years, yet it’s been a regular as a non-tech company showing off technology at the big tech trade show in Las Vegas and is back at CES 2025 with more autonomous tractors and other vehicles. This is not something we usually cover, but John Deere has a lot of data that is interesting in the big picture of tech. The message from the company is that there aren’t enough skilled farm laborers to do the work that its customers need. It’s been a challenge for most of the last two decades, said Jahmy Hindman, CTO at John Deere, in a briefing. Much of the tech will come this fall and after that. He noted that the average farmer in the U.S. is over 58 and works 12 to 18 hours a day to grow food for us. And he said the American Farm Bureau Federation estimates there are roughly 2.4 million farm jobs that need to be filled annually; and the agricultural work force continues to shrink. (This is my hint to the anti-immigration crowd). John Deere’s autonomous 9RX Tractor. Farmers can oversee it using an app. While each of these industries experiences their own set of challenges, a commonality across all is skilled labor availability. In construction, about 80% percent of contractors struggle to find skilled labor. And in commercial landscaping, 86% of landscaping business owners can’t find labor to fill open positions, he said. “They have to figure out how to do

Read More »

2025 playbook for enterprise AI success, from agents to evals

Join our daily and weekly newsletters for the latest updates and exclusive content on industry-leading AI coverage. Learn More 2025 is poised to be a pivotal year for enterprise AI. The past year has seen rapid innovation, and this year will see the same. This has made it more critical than ever to revisit your AI strategy to stay competitive and create value for your customers. From scaling AI agents to optimizing costs, here are the five critical areas enterprises should prioritize for their AI strategy this year. 1. Agents: the next generation of automation AI agents are no longer theoretical. In 2025, they’re indispensable tools for enterprises looking to streamline operations and enhance customer interactions. Unlike traditional software, agents powered by large language models (LLMs) can make nuanced decisions, navigate complex multi-step tasks, and integrate seamlessly with tools and APIs. At the start of 2024, agents were not ready for prime time, making frustrating mistakes like hallucinating URLs. They started getting better as frontier large language models themselves improved. “Let me put it this way,” said Sam Witteveen, cofounder of Red Dragon, a company that develops agents for companies, and that recently reviewed the 48 agents it built last year. “Interestingly, the ones that we built at the start of the year, a lot of those worked way better at the end of the year just because the models got better.” Witteveen shared this in the video podcast we filmed to discuss these five big trends in detail. Models are getting better and hallucinating less, and they’re also being trained to do agentic tasks. Another feature that the model providers are researching is a way to use the LLM as a judge, and as models get cheaper (something we’ll cover below), companies can use three or more models to

Read More »

OpenAI’s red teaming innovations define new essentials for security leaders in the AI era

Join our daily and weekly newsletters for the latest updates and exclusive content on industry-leading AI coverage. Learn More OpenAI has taken a more aggressive approach to red teaming than its AI competitors, demonstrating its security teams’ advanced capabilities in two areas: multi-step reinforcement and external red teaming. OpenAI recently released two papers that set a new competitive standard for improving the quality, reliability and safety of AI models in these two techniques and more. The first paper, “OpenAI’s Approach to External Red Teaming for AI Models and Systems,” reports that specialized teams outside the company have proven effective in uncovering vulnerabilities that might otherwise have made it into a released model because in-house testing techniques may have missed them. In the second paper, “Diverse and Effective Red Teaming with Auto-Generated Rewards and Multi-Step Reinforcement Learning,” OpenAI introduces an automated framework that relies on iterative reinforcement learning to generate a broad spectrum of novel, wide-ranging attacks. Going all-in on red teaming pays practical, competitive dividends It’s encouraging to see competitive intensity in red teaming growing among AI companies. When Anthropic released its AI red team guidelines in June of last year, it joined AI providers including Google, Microsoft, Nvidia, OpenAI, and even the U.S.’s National Institute of Standards and Technology (NIST), which all had released red teaming frameworks. Investing heavily in red teaming yields tangible benefits for security leaders in any organization. OpenAI’s paper on external red teaming provides a detailed analysis of how the company strives to create specialized external teams that include cybersecurity and subject matter experts. The goal is to see if knowledgeable external teams can defeat models’ security perimeters and find gaps in their security, biases and controls that prompt-based testing couldn’t find. What makes OpenAI’s recent papers noteworthy is how well they define using human-in-the-middle

Read More »