Tech leaders respond to the rapid rise of DeepSeek

Stay Ahead, Stay ONMINE

Tech leaders respond to the rapid rise of DeepSeek

Join our daily and weekly newsletters for the latest updates and exclusive content on industry-leading AI coverage. Learn More If you hadn’t heard, there’s a new AI star in town: DeepSeek, the subsidiary of Hong Kong-based quantitative analysis (quant) firm High-Flyer Capital Management, has sent shockwaves throughout Silicon Valley and the wider world with its […]

Join our daily and weekly newsletters for the latest updates and exclusive content on industry-leading AI coverage. Learn More

If you hadn’t heard, there’s a new AI star in town: DeepSeek, the subsidiary of Hong Kong-based quantitative analysis (quant) firm High-Flyer Capital Management, has sent shockwaves throughout Silicon Valley and the wider world with its release earlier this week of a new open source large reasoning model, DeepSeek R1, that matches OpenAI’s most powerful available model o1 — and at a fraction of the cost to users and to the company itself (when training it).

While the advent of DeepSeek R1 has already reshuffled a consistently topsy turvy, fast moving, intensely competitive market for new AI models — previous months saw OpenAI jockeying with Anthropic and Google for the most powerful proprietary models available, while Meta Platforms often came in with “close enough” open source rivals — the difference this time is the company behind the hot model is based in China, the geopolitical “frenemy” of the U.S., and whose tech sector was widely viewed, until this moment, as inferior to that of Silicon Valley.

As such, it’s caused no shortage of hand-wringing and existentialism from U.S. and Western bloc techies, who are suddenly doubting OpenAI and the general big tech strategy of throwing more money and more compute (graphics processing units, GPUs, the powerful gaming chips typically used to train AI models) toward the problem of inventing ever more powerful models.

Yet some Western tech leaders have had a largely positive public response to DeepSeek’s rapid ascent.

Marc Andreessen, a co-inventor of the pioneering Mosaic web browser, co-founder of the Netscape browser company and current general partner at the famed Andreessen Horowitz (a16z) venture capital firm, posted on X today: “Deepseek R1 is one of the most amazing and impressive breakthroughs I’ve ever seen — and as open source, a profound gift to the world [robot emoji, salute emoji].”

Yann LeCun, the Chief AI Scientist for Meta’s Fundamental AI Research (FAIR) division, posted on his LinkedIn account:

“To people who see the performance of DeepSeek and think:
‘China is surpassing the US in AI.’
You are reading this wrong.
The correct reading is:
‘Open source models are surpassing proprietary ones.’

DeepSeek has profited from open research and open source (e.g. PyTorch and Llama from Meta)
They came up with new ideas and built them on top of other people’s work.
Because their work is published and open source, everyone can profit from it.
That is the power of open research and open source.”

And even Mark “Zuck” Zuckerberg, Meta AI’s founder and CEO, seemed to seek to counter the rise of DeepSeek with his own post on Facebook promising that a new version of Facebook’s open source AI model family Llama would be “the leading state of the art model” when it is released sometime this year. As he put it:

“This will be a defining year for AI. In 2025, I expect Meta AI will be the leading assistant serving more than 1 billion people, Llama 4 will become the leading state of the art model, and we’ll build an AI engineer that will start contributing increasing amounts of code to our R&D efforts. To power this, Meta is building a 2GW+ datacenter that is so large it would cover a significant part of Manhattan. We’ll bring online ~1GW of compute in ’25 and we’ll end the year with more than 1.3 million GPUs. We’re planning to invest $60-65B in capex this year while also growing our AI teams significantly, and we have the capital to continue investing in the years ahead. This is a massive effort, and over the coming years it will drive our core products and business, unlock historic innovation, and extend American technology leadership. Let’s go build!“

He even shared a graphic showing the 2 gigawatt datacenter mentioned in his post overlaid on Manhattan:

Clearly, even as he espouses a commitment to open source AI, Zuck is not convinced that DeepSeek’s approach of optimizing for efficiency while leveraging far fewer GPUs than major labs is the right one for Meta, or for the future of AI.

But with U.S. companies raising and/or spending record sums on new AI infrastructure that many experts have noted depreciate rapidly (due to hardware/chip and software advancements), the question remains which vision of the future will win out in the end to become the dominant AI provider for the world. Or maybe it will always be a multiplicity of models each with a smaller market share? Stay tuned, because this competition is getting closer and fiercer than ever.

Daily insights on business use cases with VB Daily

If you want to impress your boss, VB Daily has you covered. We give you the inside scoop on what companies are doing with generative AI, from regulatory shifts to practical deployments, so you can share insights for maximum ROI.

Read our Privacy Policy

Thanks for subscribing. Check out more VB newsletters here.

An error occured.

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.

FortiGate admins advised to run a compromise assessment

While the data was apparently collected just over two years ago, it is unknown why it’s being released now. In a post last week analyzing the dump, researchers at Censys noted that the Belsen Group is new. It’s possible that this threat actor recently bought or assembled the data now

AMD’s unpatched chip microcode glitch may require extreme measures by CISOs

“If a system is compromised to this level, the ability to deploy malicious microcode to the CPU could make for a very insidious attack vector that would be very hard to identify and address,” Villanustre said. “Creating these types of sophisticated attacks would require significant resources, but it could be

Palo Alto Networks firewalls have UEFI flaws, Secure Boot bypasses

Palo Alto’s firewall device operating system, PAN-OS, is based on Red Hat Linux, which uses Grand Unified Bootloader version 2 (GRUB2). The company signs its GRUB2 bootloader and other components with its own certificates, which are stored in the UEFI certificate store to establish the chain of trust. However, in

3 hidden benefits of Dedicated Internet Access for enterprises

Reliable Internet connectivity is crucial for businesses to function efficiently and remain competitive in today’s digital world. While standard broadband connections may work for basic tasks, they often struggle to meet the demands of modern businesses. Activities like video conferencing, using cloud applications, or transferring large amounts of data require

Putin Says He’s Ready to Discuss Oil, Energy Issues With Trump

Russian President Vladimir Putin said he’s ready to discuss crude oil prices and other energy issues at a face-to-face meeting with US leader Donald Trump. Both Russia and the US are top global oil producers and consumers, Putin said in an interview with state-owned Rossiya 24 TV. Price movements in the commodity impacts both economies, he said. “We have a lot to discuss here, and there are other energy issues that might be mutually interesting,” Putin said. “I doubt that Mr. Trump — even if we are hearing about the possibility of additional sanction on Russia — would make decisions that could hurt the US economy.” Putin’s comments follow Trump’s Thursday address to world leaders in Davos, when he urged Saudi Arabia and other members of the Organization of Petroleum Exporting Countries to “bring down the cost of oil.” “If the price came down, the Russia-Ukraine war would end immediately,” Trump added. Earlier this week, Trump also stepped up pressure on Russia to make a peace deal with Ukraine “soon,” or else he would “have no other choice” but to impose additional taxes, tariffs and sanctions on Russian imports to the US, along with “other participating countries.” Earlier this month, the US imposed its harshest sanctions so far on Russia’s energy industry, the key source of the Kremlin’s revenue that funds the war in Ukraine. The restrictions could disrupt Russia’s exports of petroleum and reduce the global supply surplus. WHAT DO YOU THINK? Generated by readers, the comments included herein do not reflect the views and opinions of Rigzone. All comments are subject to editorial review. Off-topic, inappropriate or insulting comments will be removed. MORE FROM THIS AUTHOR Bloomberg

Trump plans to use emergency powers to fast-track generation co-located with AI

President Donald Trump plans to speed power plant development for co-located artificial intelligence data centers using his energy emergency declaration, he said Thursday. “I can get the approvals done myself without having to go through years of waiting,” Trump told the World Economic Forum in Davos, Switzerland. “I’m going to give emergency declarations so that they can start building them almost immediately.” Earlier this week, Trump hosted several of the founders of the Stargate Project at the White House. The Stargate Project is a joint venture between SoftBank, OpenAI, Oracle and MGX. The project envisions “immediately” spending $100 billion on co-located data centers with an initial focus on a site in Abilene, Texas. “We need to double the energy we currently have in the United States … for AI to really be as big as we want to have it” to compete with China and other countries, Trump said. Co-locating generation and data centers “was largely my idea,” Trump said. “Nobody thought this was possible … I told them that what I want you to do is build your electric generating plant right next to your plant as a separate building connected.” Coal-fired generation could be used as a backup for the data centers, according to Trump. “Nothing can destroy coal, not the weather, not a bomb, nothing,” Trump said. “It might make it a little smaller, might make it a little different shape, but coal is very strong as a backup.” Trump’s co-location comments appear to align with newly appointed Federal Energy Regulatory Commission Chairman Mark Christie, according to ClearView Energy Partners. “Chairman Christie has supported co-location on the condition that developers site data centers adjacent to new power plants,” the research firm said in a client note on Thursday. “He has raised concerns about data centers or other energy-intensive facilities

Trump Order Offers a Chance to Revive Keystone XL Pipeline

President Donald Trump appears to have opened the door for construction of the Keystone XL pipeline, the controversial oil conduit that even its former developer doesn’t want to build. A Biden administration executive order that revoked Trump’s March 2019 permit for the pipeline was among the directives rescinded by the newly elected president Monday. The decision appears to have put Keystone XL back in play — even if it may now be little more than symbolic. It’s unlikely that the multibillion-dollar 1,200-mile project to build the pipeline from Canada to Nebraska would proceed anytime soon — if ever. South Bow Corp., the oil pipeline business spun off from TC Energy Corp., hasn’t indicated interest in a revival. “We’ve moved on from Keystone XL,” South Bow’s chief executive officer, Bevin Wirzba, told Bloomberg last June. Former President Joe Biden revoked Trump’s permit allowing Keystone to cross the US-Canada border hours after taking office in January 2021. Since then, parts of the system — which runs through Alberta, Montana, South Dakota and Nebraska — have been dismantled. ‘Virtually all of the permits along the way have expired,” Anthony Swift, a senior director at the Natural Resources Defense Council, said in an interview. “So it would be starting from ground one to resuscitate the project.” A White House spokesman didn’t comment on the matter. Trump and other Republicans attacked Biden’s decision to kill the pipeline, blaming the move for increasing gas prices and using it to paint Democrats as anti-oil. The pipeline’s shift in fortunes unfolded despite Trump’s insistence that the US doesn’t need Canadian crude oil. “We don’t need their oil and gas,” Trump said Thursday in a remote presentation to the World Economic Forum in Davos, Switzerland. “We have more than anybody.” The case of the Keystone pipeline underscores his zeal to revoke many Biden

Oil Falls Recording Weekly Decline Amid Trade War Threats

Oil posted its first weekly decline this year after President Donald Trump threatened trade wars and demanded OPEC+ lower crude prices. West Texas Intermediate settled little changed above $74 a barrel after fluctuating between gains and losses for most of the session. Earlier on Friday, crude extended losses after Russian President Vladimir Putin reiterated he’s open to discussing Ukraine and oil prices with Trump. This week, Trump had threatened more penalties on Moscow if Putin doesn’t “make a deal” to end the prolonged war in Ukraine. Recently imposed US sanctions on Russia’s oil have been tightening the global market, and loosening them may increase supplies available to customers in Asia that had been scrambling to seek alternatives. The sanctions were introduced before Trump took office to boost Ukraine’s leverage in possible peace negotiations. Oil markets have been reacting to Trump’s statements since he took office on Monday. The first week of his new term began with tariff threats aimed at Canada, Mexico and China, followed by a pledge that he would ask the OPEC+ producer group to “bring down the cost of oil.” “He just wants the price lower,” Nadia Martin Wiggen, a director at Svelland Capital, said in a Bloomberg Television interview. “He wants gasoline prices lower for the consumer, and he wants oil prices just to be lower, at least than what the Biden administration had. On the flip side, he wants producers to keep producing and especially American ones, and that’s where we see him in a tough situation.” Futures in New York fell 4.1% this week, but remain higher for the year amid cold weather in the Northern Hemisphere and the Russia sanctions. Some of the strength in crude and freight markets that followed the sanctions has since cooled. One of Trump’s executive orders this week

Storm alert! How wild weather can wreak havoc on energy infrastructure

There is no doubt that with climate change comes more extreme weather. In the UK, the Met Office has been naming storms now for ten years. These have been coming with greater regularity, often bringing devastation in their wake. The 2023/24 storm season saw 12 named storms, the greatest number of named storms since the first season in 2015. Storms often cause devastation on the infrastructure that delivers our energy and for the workers responsible for keeping the lights on. Storms can cause billions of pounds of damage and can even put lives at risk. As the UK awaits the impact of Storm Eowyn, which is bringing gusts of up to 100mph across Ireland and Scotland, we have brought together some of significant impacts recent storms have caused. Wind turbine bursts into flames A wind turbine burst into flames as hurricane-force winds battered parts of Scotland in 2011. Firefighters were called as the huge 330ft turbine in North Ayrshire caught fire in the afternoon. However, the blaze died out within minutes of their arrival. North Sea rig floats away Dozens of workers were evacuated after Storm Babet in 2023 caused a North Sea oil rig east of Aberdeen lose anchors amid turbulent waves. Forty-five non-essential staff working aboard the Stena Spey were airlifted to other sites after four out of eight anchors became detached due to the severe weather. WATCH: Wind turbine sheds blades In December 2023, an Ayeshire wind turbine lost its blades and disintegrated as it was battered by Storm Gerrit. An onlooker captured the astonishing moment the 35m turbine at Auchencloigh farm span out of control. Irish Sea oil rig evacuated Last year 30 workers were evacuated from this oil rig in the Irish Sea but it was a preventative measure to ensure their safety. Petrofac (LON:PFC)

USA Crude Oil Inventories Drop 1 Million Barrels WoW

U.S. commercial crude oil inventories, excluding those in the Strategic Petroleum Reserve (SPR), decreased by 1.0 million barrels from the week ending January 10 to the week ending January 17, the U.S. Energy Information Administration (EIA) highlighted in its latest weekly petroleum status report. Crude oil stocks, excluding the SPR, stood at 411.7 million barrels on January 17, 412.7 million barrels on January 10, and 420.7 million barrels on January 19, 2024, the report revealed. Crude oil in the SPR came in at 394.6 million barrels on January 17, 394.3 million barrels on January 10, and 356.5 million barrels on January 19, 2024, the report showed. Total petroleum stocks – including crude oil, total motor gasoline, fuel ethanol, kerosene type jet fuel, distillate fuel oil, residual fuel oil, propane/propylene, and other oils – stood at 1.621 billion barrels on January 17, the report revealed. This figure was down 3.9 million barrels week on week and up 24.2 million barrels year on year, the report outlined. “At 411.7 million barrels, U.S. crude oil inventories are about six percent below the five year average for this time of year,” the EIA noted in its latest weekly petroleum status report. “Total motor gasoline inventories increased by 2.3 million barrels from last week and are one percent below the five year average for this time of year,” it added. “Finished gasoline inventories and blending components inventories increased last week. Distillate fuel inventories decreased by 3.1 million barrels last week and are about six percent below the five year average for this time of year,” it continued. “Propane/propylene inventories decreased by 3.7 million barrels from last week and are eight percent above the five year average for this time of year,” the EIA went on to state. In its report, the EIA said U.S. crude

Meta wants everyone to know that it, too, is investing a lot in AI

In his Facebook post this week, Zuckerberg wrote that Meta will also “significantly” grow out its AI teams and will build an “AI engineer” AI agent that will contribute code to Meta’s R&D efforts. Planned investments for this year already represent a 50% increase over the company’s 2024 spending, and Zuckerberg noted, “we have the capital to continue investing in the years ahead.” His announcement comes just four days after OpenAI dropped its “Stargate Project” bombshell, and four days before Meta’s planned fourth-quarter financial reporting on Jan. 29. It also follows on the heels of US President Donald Trump’s controversial executive order on AI that specifically revokes previous administration policies that he said “act as barriers to American AI innovation.” “This will be a defining year for AI,” Zuckerberg wrote in his Facebook post, also saying he expects that Meta AI will be the world’s “leading assistant,” serving more than 1 billion people, and that Llama 4 will become the “leading state-of-the-art model.” “This is a massive effort, and over the coming years it will drive our core products and business, unlock historic innovation, and extend American technology leadership,” he wrote. “Let’s go build!” Big tech racing to build the data centers of the future Undoubtedly one of the biggest stories out of the tech world this week was the US President Donald Trump-endorsed Stargate Project, an ambitious, $500 billion initiative to build out AI infrastructure in the US for OpenAI over the next four years. It is an industry collaboration, with participation from OpenAI, Oracle, SoftBank, MGX, Arm, Nvidia and Microsoft, and will deploy $100 billion “immediately,” OpenAI said in a blog post earlier this week. Other industry players are investing as well. Microsoft has announced its intent to spend $80 billion in fiscal year 2025, more than half

Data center growth puts unprecedented pressure on power grids

The spike in electricity needs is unprecedented, McKinsey wrote, and will be difficult to meet. In another report released this week, Goldman Sachs also predicts a surge in data center power demand, and expects an increase of 160% by 2030, compared to 2023 levels. Data center builds typically have timelines of 18 to 24 months, gas and renewable power plans have timelines of three to five years, and new transmission development can take seven to ten years. In fact, it’s the power transmission grid that’s often the biggest constraint, the research firm says, more even than the power generation capacity. According to a report by Grid Strategies, construction of new high-voltage transmission lines has actually slowed recently, dropping steadily from a peak of 4,000 miles of new high-voltage lines in 2012 to just 55 in 2023. To meet demands, the DOE estimates that 115,000 miles will be needed by 2040, doubling current grid capacity. Companies are already planning ahead for shortfalls.

US GPU export limits could bring cold war to AI, data center markets

Eighteen countries, including the UK, Canada, Sweden, France, Germany, Japan, and South Korea, are exempted from the AI export caps. The Biden administration had previously banned the export of some powerful AI chips to China, Russia, and other adversaries in rules from 2022 and 2023. But other countries friendly to the US, including Mexico, Israel, India, and Saudi Arabia, would be subject to the quotas. The export limits would take effect 120 days from the Jan. 13 order, and it’s unclear whether the incoming Trump administration will amend or rewrite the rule, although Trump has targeted China as a primary economic competitor of the US. The cost of AI In addition to cutting off most of the world from large AI chip purchases, the rule will force countries such as China and Russia to pump up their own AI capabilities, ultimately reducing US AI leadership, claims Aible’s Sengupta.

Sustainability, grid demands, AI workloads will challenge data center growth in 2025

Cloud training for AI models Uptime believes that most AI models will be trained in the cloud rather than on dedicated enterprise infrastructure, as cloud services provide a more cost-effective way to fine-tune foundation models for specific use cases. The incremental training required to fine-tune a foundation model can be done cost-effectively on cloud platforms without the need for a large, expensive on-premises cluster. Enterprises can leverage on-demand cloud resources to customize the foundation model as needed, without investing the capital and operational costs of dedicated hardware. “Because fine-tuning requires only a relatively small amount of training, for many it just wouldn’t make sense to buy a huge, expensive dedicated AI cluster for this purpose. The foundation model, which has already been trained by someone else, has taken the burden of most of the training away from us,” said Dr. Owen Rogers, research director for cloud computing at Uptime. “Instead, we could just use on-demand cloud services to tweak the foundation model for our needs, only paying for the resources we need for as long as we need them.” Data center collaboration with utilities Uptime expects new and expanded data center developers will be asked to provide or store power to support grids. That means data centers will need to actively collaborate with utilities to manage grid demand and stability, potentially shedding load or using local power sources during peak times. Uptime forecasts that data center operators “running non-latency-sensitive workloads, such as specific AI training tasks, could be financially incentivized or mandated to reduce power use when required.” “The context for all of this is that the [power] grid, even if there were no data centers, would have a problem meeting demand over time. They’re having to invest at a rate that is historically off the charts. It’s not just

UK Government’s Bold AI Plan: A Game-Changer for Data Centers and Economic Growth?

The UK government has presently announced its comprehensive “AI Opportunities Action Plan,” positioning artificial intelligence as a cornerstone for economic growth and public service transformation over the next decade. The bold initiative, spearheaded by Prime Minister Keir Starmer, aims to make Britain a global leader in AI development and adoption, with significant implications for the data center industry. Britain’s ambitious AI roadmap taps into the growing synergy between artificial intelligence and data infrastructure. With dedicated AI Growth Zones and a focus on sustainable energy, the UK is setting the stage for an AI-driven economy that aligns with the next generation of data center demands. The data center industry should watch these developments closely, as they signal opportunities for long-term growth in a rapidly evolving market. AI Infrastructure Prioritization Meets Major Private Sector Investments The UK government plan introduces “AI Growth Zones,” areas designed to streamline planning approvals for data centers and enhance access to energy infrastructure. These zones will focus on de-industrialized regions, providing a dual benefit of revitalizing local economies and accelerating the rollout of AI infrastructure. The first such zone will be established in Culham, Oxfordshire, leveraging local expertise in sustainable energy research, including fusion technologies. Leading tech firms, including Vantage Data Centers, Nscale, and Kyndryl, have committed £14 billion to AI infrastructure development under the plan, creating 13,250 jobs across the UK, according to a press release. Vantage Data Centers alone plans to invest over £12 billion to establish one of Europe’s largest campuses in Wales and additional facilities nationwide, generating 11,500 jobs. Plan Harnesses AI for Both Public, Private Sectors A significant component of the plan is a proposed 20x increase in public compute capacity by 2030, starting with the development of a new supercomputer to support AI innovation. Alongside this supercharging of

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.

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

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

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

Stay Ahead, Stay ONMINE