A Visual Guide to How Diffusion Models Work

Stay Ahead, Stay ONMINE

A Visual Guide to How Diffusion Models Work

This article is aimed at those who want to understand exactly how Diffusion Models work, with no prior knowledge expected. I’ve tried to use illustrations wherever possible to provide visual intuitions on each part of these models. I’ve kept mathematical notation and equations to a minimum, and where they are necessary I’ve tried to define […]

Intro

I’ve framed this article around three main questions:

What exactly is it that diffusion models learn?
How and why do diffusion models work?
Once you’ve trained a model, how do you get useful stuff out of it?

The examples will be based on the glyffuser, a minimal text-to-image diffusion model that I previously implemented and wrote about. The architecture of this model is a standard text-to-image denoising diffusion model without any bells or whistles. It was trained to generate pictures of new “Chinese” glyphs from English definitions. Have a look at the picture below — even if you’re not familiar with Chinese writing, I hope you’ll agree that the generated glyphs look pretty similar to the real ones!

Random examples of glyffuser training data (left) and generated data (right).

What exactly is it that diffusion models learn?

Generative Ai models are often said to take a big pile of data and “learn” it. For text-to-image diffusion models, the data takes the form of pairs of images and descriptive text. But what exactly is it that we want the model to learn? First, let’s forget about the text for a moment and concentrate on what we are trying to generate: the images.

Probability distributions

Broadly, we can say that we want a generative AI model to learn the underlying probability distribution of the data. What does this mean? Consider the one-dimensional normal (Gaussian) distribution below, commonly written 𝒩(μ,σ²) and parameterized with mean μ = 0 and variance σ² = 1. The black curve below shows the probability density function. We can sample from it: drawing values such that over a large number of samples, the set of values reflects the underlying distribution. These days, we can simply write something like x = random.gauss(0, 1) in Python to sample from the standard normal distribution, although the computational sampling process itself is non-trivial!

Values sampled from an underlying distribution (here, the standard normal 𝒩(0,1)) can then be used to estimate the parameters of that distribution.

We could think of a set of numbers sampled from the above normal distribution as a simple dataset, like that shown as the orange histogram above. In this particular case, we can calculate the parameters of the underlying distribution using maximum likelihood estimation, i.e. by working out the mean and variance. The normal distribution estimated from the samples is shown by the dotted line above. To take some liberties with terminology, you might consider this as a simple example of “learning” an underlying probability distribution. We can also say that here we explicitly learnt the distribution, in contrast with the implicit methods that diffusion models use.

Conceptually, this is all that generative AI is doing — learning a distribution, then sampling from that distribution!

Data representations

What, then, does the underlying probability distribution of a more complex dataset look like, such as that of the image dataset we want to use to train our diffusion model?

First, we need to know what the representation of the data is. Generally, a machine learning (ML) model requires data inputs with a consistent representation, i.e. format. For the example above, it was simply numbers (scalars). For images, this representation is commonly a fixed-length vector.

The image dataset used for the glyffuser model is ~21,000 pictures of Chinese glyphs. The images are all the same size, 128 × 128 = 16384 pixels, and greyscale (single-channel color). Thus an obvious choice for the representation is a vector x of length 16384, where each element corresponds to the color of one pixel: x = (x₁,x₂,…,x₁₆₃₈₄). We can call the domain of all possible images for our dataset “pixel space”.

An example glyph with pixel values labelled (downsampled to 32 × 32 pixels for readability).

Dataset visualization

We make the assumption that our individual data samples, x, are actually sampled from an underlying probability distribution, q(x), in pixel space, much as the samples from our first example were sampled from an underlying normal distribution in 1-dimensional space. Note: the notation x ∼ q(x) is commonly used to mean: “the random variable x sampled from the probability distribution q(x).”

This distribution is clearly much more complex than a Gaussian and cannot be easily parameterized — we need to learn it with a ML model, which we’ll discuss later. First, let’s try to visualize the distribution to gain a better intution.

As humans find it difficult to see in more than 3 dimensions, we need to reduce the dimensionality of our data. A small digression on why this works: the manifold hypothesis posits that natural datasets lie on lower dimensional manifolds embedded in a higher dimensional space — think of a line embedded in a 2-D plane, or a plane embedded in 3-D space. We can use a dimensionality reduction technique such as UMAP to project our dataset from 16384 to 2 dimensions. The 2-D projection retains a lot of structure, consistent with the idea that our data lie on a lower dimensional manifold embedded in pixel space. In our UMAP, we see two large clusters corresponding to characters in which the components are arranged either horizontally (e.g. 明) or vertically (e.g. 草). An interactive version of the plot below with popups on each datapoint is linked here.

Click here for an interactive version of this plot.

Let’s now use this low-dimensional UMAP dataset as a visual shorthand for our high-dimensional dataset. Remember, we assume that these individual points have been sampled from a continuous underlying probability distribution q(x). To get a sense of what this distribution might look like, we can apply a KDE (kernel density estimation) over the UMAP dataset. (Note: this is just an approximation for visualization purposes.)

This gives a sense of what q(x) should look like: clusters of glyphs correspond to high-probability regions of the distribution. The true q(x) lies in 16384 dimensions — this is the distribution we want to learn with our diffusion model.

We showed that for a simple distribution such as the 1-D Gaussian, we could calculate the parameters (mean and variance) from our data. However, for complex distributions such as images, we need to call on ML methods. Moreover, what we will find is that for diffusion models in practice, rather than parameterizing the distribution directly, they learn it implicitly through the process of learning how to transform noise into data over many steps.

Takeaway

The aim of generative AI such as diffusion models is to learn the complex probability distributions underlying their training data and then sample from these distributions.

How and why do diffusion models work?

Diffusion models have recently come into the spotlight as a particularly effective method for learning these probability distributions. They generate convincing images by starting from pure noise and gradually refining it. To whet your interest, have a look at the animation below that shows the denoising process generating 16 samples.

In this section we’ll only talk about the mechanics of how these models work but if you’re interested in how they arose from the broader context of generative models, have a look at the further reading section below.

What is “noise”?

Let’s first precisely define noise, since the term is thrown around a lot in the context of diffusion. In particular, we are talking about Gaussian noise: consider the samples we talked about in the section about probability distributions. You could think of each sample as an image of a single pixel of noise. An image that is “pure Gaussian noise”, then, is one in which each pixel value is sampled from an independent standard Gaussian distribution, 𝒩(0,1). For a pure noise image in the domain of our glyph dataset, this would be noise drawn from 16384 separate Gaussian distributions. You can see this in the previous animation. One thing to keep in mind is that we can choose the means of these noise distributions, i.e. center them, on specific values — the pixel values of an image, for instance.

For convenience, you’ll often find the noise distributions for image datasets written as a single multivariate distribution 𝒩(0,I) where I is the identity matrix, a covariance matrix with all diagonal entries equal to 1 and zeroes elsewhere. This is simply a compact notation for a set of multiple independent Gaussians — i.e. there are no correlations between the noise on different pixels. In the basic implementations of diffusion models, only uncorrelated (a.k.a. “isotropic”) noise is used. This article contains an excellent interactive introduction on multivariate Gaussians.

Diffusion process overview

Below is an adaptation of the somewhat-famous diagram from Ho et al.’s seminal paper “Denoising Diffusion Probabilistic Models” which gives an overview of the whole diffusion process:

Diagram of the diffusion process adapted from Ho *et al*. 2020. The glyph 锂, meaning “lithium”, is used as a representative sample from the dataset.

I found that there was a lot to unpack in this diagram and simply understanding what each component meant was very helpful, so let’s go through it and define everything step by step.

We previously used x ∼ q(x) to refer to our data. Here, we’ve added a subscript, xₜ, to denote timestep t indicating how many steps of “noising” have taken place. We refer to the samples noised a given timestep as x ∼ q(xₜ). x₀ is clean data and xₜ (t = T) ∼ 𝒩(0,1) is pure noise.

We define a forward diffusion process whereby we corrupt samples with noise. This process is described by the distribution q(xₜ|xₜ₋₁). If we could access the hypothetical reverse process q(xₜ₋₁|xₜ), we could generate samples from noise. As we cannot access it directly because we would need to know x₀, we use ML to learn the parameters, θ, of a model of this process, 𝑝θ(𝑥ₜ₋₁∣𝑥ₜ). (That should be p subscript θ but medium cannot render it.)

In the following sections we go into detail on how the forward and reverse diffusion processes work.

Forward diffusion, or “noising”

Used as a verb, “noising” an image refers to applying a transformation that moves it towards pure noise by scaling down its pixel values toward 0 while adding proportional Gaussian noise. Mathematically, this transformation is a multivariate Gaussian distribution centered on the pixel values of the preceding image.

In the forward diffusion process, this noising distribution is written as q(xₜ|xₜ₋₁) where the vertical bar symbol “|” is read as “given” or “conditional on”, to indicate the pixel means are passed forward from q(xₜ₋₁) At t = T where T is a large number (commonly 1000) we aim to end up with images of pure noise (which, somewhat confusingly, is also a Gaussian distribution, as discussed previously).

The marginal distributions q(xₜ) represent the distributions that have accumulated the effects of all the previous noising steps (marginalization refers to integration over all possible conditions, which recovers the unconditioned distribution).

Since the conditional distributions are Gaussian, what about their variances? They are determined by a variance schedule that maps timesteps to variance values. Initially, an empirically determined schedule of linearly increasing values from 0.0001 to 0.02 over 1000 steps was presented in Ho et al. Later research by Nichol & Dhariwal suggested an improved cosine schedule. They state that a schedule is most effective when the rate of information destruction through noising is relatively even per step throughout the whole noising process.

Forward diffusion intuition

As we encounter Gaussian distributions both as pure noise q(xₜ, t = T) and as the noising distribution q(xₜ|xₜ₋₁), I’ll try to draw the distinction by giving a visual intuition of the distribution for a single noising step, q(x₁∣x₀), for some arbitrary, structured 2-dimensional data:

Each noising step q(xₜ|xₜ₋₁) is a Gaussian distribution conditioned on the previous step.

The distribution q(x₁∣x₀) is Gaussian, centered around each point in x₀, shown in blue. Several example points x₀⁽ⁱ⁾ are picked to illustrate this, with q(x₁∣x₀ = x₀⁽ⁱ⁾) shown in orange.

In practice, the main usage of these distributions is to generate specific instances of noised samples for training (discussed further below). We can calculate the parameters of the noising distributions at any timestep t directly from the variance schedule, as the chain of Gaussians is itself also Gaussian. This is very convenient, as we don’t need to perform noising sequentially—for any given starting data x₀⁽ⁱ⁾, we can calculate the noised sample xₜ⁽ⁱ⁾ by sampling from q(xₜ∣x₀ = x₀⁽ⁱ⁾) directly.

Forward diffusion visualization

Let’s now return to our glyph dataset (once again using the UMAP visualization as a visual shorthand). The top row of the figure below shows our dataset sampled from distributions noised to various timesteps: xₜ ∼ q(xₜ). As we increase the number of noising steps, you can see that the dataset begins to resemble pure Gaussian noise. The bottom row visualizes the underlying probability distribution q(xₜ).

The dataset xₜ (above) sampled from its probability distribution q(xₜ) (below) at different noising timesteps.

Reverse diffusion overview

It follows that if we knew the reverse distributions q(xₜ₋₁∣xₜ), we could repeatedly subtract a small amount of noise, starting from a pure noise sample xₜ at t = T to arrive at a data sample x₀ ∼ q(x₀). In practice, however, we cannot access these distributions without knowing x₀ beforehand. Intuitively, it’s easy to make a known image much noisier, but given a very noisy image, it’s much harder to guess what the original image was.

So what are we to do? Since we have a large amount of data, we can train an ML model to accurately guess the original image that any given noisy image came from. Specifically, we learn the parameters θ of an ML model that approximates the reverse noising distributions, pθ(xₜ₋₁ ∣ xₜ) for t = 0, …, T. In practice, this is embodied in a single noise prediction model trained over many different samples and timesteps. This allows it to denoise any given input, as shown in the figure below.

The ML model predicts added noise at any given timestep t.

Next, let’s go over how this noise prediction model is implemented and trained in practice.

How the model is implemented

First, we define the ML model — generally a deep neural network of some sort — that will act as our noise prediction model. This is what does the heavy lifting! In practice, any ML model that inputs and outputs data of the correct size can be used; the U-net, an architecture particularly suited to learning images, is what we use here and frequently chosen in practice. More recent models also use vision transformers.

We use the U-net architecture (Ronneberger *et al*. 2015) for our ML noise prediction model. We train the model by minimizing the difference between predicted and actual noise.

Then we run the training loop depicted in the figure above:

We take a random image from our dataset and noise it to a random timestep tt. (In practice, we speed things up by doing many examples in parallel!)
We feed the noised image into the ML model and train it to predict the (known to us) noise in the image. We also perform timestep conditioning by feeding the model a timestep embedding, a high-dimensional unique representation of the timestep, so that the model can distinguish between timesteps. This can be a vector the same size as our image directly added to the input (see here for a discussion of how this is implemented).
The model “learns” by minimizing the value of a loss function, some measure of the difference between the predicted and actual noise. The mean square error (the mean of the squares of the pixel-wise difference between the predicted and actual noise) is used in our case.
Repeat until the model is well trained.

Note: A neural network is essentially a function with a huge number of parameters (on the order of 10⁶ for the glyffuser). Neural network ML models are trained by iteratively updating their parameters using backpropagation to minimize a given loss function over many training data examples. This is an excellent introduction. These parameters effectively store the network’s “knowledge”.

A noise prediction model trained in this way eventually sees many different combinations of timesteps and data examples. The glyffuser, for example, was trained over 100 epochs (runs through the whole data set), so it saw around 2 million data samples. Through this process, the model implicity learns the reverse diffusion distributions over the entire dataset at all different timesteps. This allows the model to sample the underlying distribution q(x₀) by stepwise denoising starting from pure noise. Put another way, given an image noised to any given level, the model can predict how to reduce the noise based on its guess of what the original image. By doing this repeatedly, updating its guess of the original image each time, the model can transform any noise to a sample that lies in a high-probability region of the underlying data distribution.

Reverse diffusion in practice

We can now revisit this video of the glyffuser denoising process. Recall a large number of steps from sample to noise e.g. T = 1000 is used during training to make the noise-to-sample trajectory very easy for the model to learn, as changes between steps will be small. Does that mean we need to run 1000 denoising steps every time we want to generate a sample?

Luckily, this is not the case. Essentially, we can run the single-step noise prediction but then rescale it to any given step, although it might not be very good if the gap is too large! This allows us to approximate the full sampling trajectory with fewer steps. The video above uses 120 steps, for instance (most implementations will allow the user to set the number of sampling steps).

Recall that predicting the noise at a given step is equivalent to predicting the original image x₀, and that we can access the equation for any noised image deterministically using only the variance schedule and x₀. Thus, we can calculate xₜ₋ₖ based on any denoising step. The closer the steps are, the better the approximation will be.

Too few steps, however, and the results become worse as the steps become too large for the model to effectively approximate the denoising trajectory. If we only use 5 sampling steps, for example, the sampled characters don’t look very convincing at all:

There is then a whole literature on more advanced sampling methods beyond what we’ve discussed so far, allowing effective sampling with much fewer steps. These often reframe the sampling as a differential equation to be solved deterministically, giving an eerie quality to the sampling videos — I’ve included one at the end if you’re interested. In production-level models, these are usually preferred over the simple method discussed here, but the basic principle of deducing the noise-to-sample trajectory is the same. A full discussion is beyond the scope of this article but see e.g. this paper and its corresponding implementation in the Hugging Face diffusers library for more information.

Alternative intuition from score function

To me, it was still not 100% clear why training the model on noise prediction generalises so well. I found that an alternative interpretation of diffusion models known as “score-based modeling” filled some of the gaps in intuition (for more information, refer to Yang Song’s definitive article on the topic.)

The dataset xₜ sampled from its probability distribution q(xₜ) at different noising timesteps; below, we add the score function ∇ₓ log q(xₜ).

I try to give a visual intuition in the bottom row of the figure above: essentially, learning the noise in our diffusion model is equivalent (to a constant factor) to learning the score function, which is the gradient of the log of the probability distribution: ∇ₓ log q(x). As a gradient, the score function represents a vector field with vectors pointing towards the regions of highest probability density. Subtracting the noise at each step is then equivalent to moving following the directions in this vector field towards regions of high probability density.

As long as there is some signal, the score function effectively guides sampling, but in regions of low probability it tends towards zero as there is little to no gradient to follow. Using many steps to cover different noise levels allows us to avoid this, as we smear out the gradient field at high noise levels, allowing sampling to converge even if we start from low probability density regions of the distribution. The figure shows that as the noise level is increased, more of the domain is covered by the score function vector field.

Summary

The aim of diffusion models is learn the underlying probability distribution of a dataset and then be able to sample from it. This requires forward and reverse diffusion (noising) processes.
The forward noising process takes samples from our dataset and gradually adds Gaussian noise (pushes them off the data manifold). This forward process is computationally efficient because any level of noise can be added in closed form a single step.
The reverse noising process is challenging because we need to predict how to remove the noise at each step without knowing the original data point in advance. We train a ML model to do this by giving it many examples of data noised at different timesteps.
Using very small steps in the forward noising process makes it easier for the model to learn to reverse these steps, as the changes are small.
By applying the reverse noising process iteratively, the model refines noisy samples step by step, eventually producing a realistic data point (one that lies on the data manifold).

Takeaway

Diffusion models are a powerful framework for learning complex data distributions. The distributions are learnt implicitly by modelling a sequential denoising process. This process can then be used to generate samples similar to those in the training distribution.

Once you’ve trained a model, how do you get useful stuff out of it?

Earlier uses of generative AI such as “This Person Does Not Exist” (ca. 2019) made waves simply because it was the first time most people had seen AI-generated photorealistic human faces. A generative adversarial network or “GAN” was used in that case, but the principle remains the same: the model implicitly learnt a underlying data distribution — in that case, human faces — then sampled from it. So far, our glyffuser model does a similar thing: it samples randomly from the distribution of Chinese glyphs.

The question then arises: can we do something more useful than just sample randomly? You’ve likely already encountered text-to-image models such as Dall-E. They are able to incorporate extra meaning from text prompts into the diffusion process — this in known as conditioning. Likewise, diffusion models for scientific scientific applications like protein (e.g. Chroma, RFdiffusion, AlphaFold3) or inorganic crystal structure generation (e.g. MatterGen) become much more useful if can be conditioned to generate samples with desirable properties such as a specific symmetry, bulk modulus, or band gap.

Conditional distributions

We can consider conditioning as a way to guide the diffusion sampling process towards particular regions of our probability distribution. We mentioned conditional distributions in the context of forward diffusion. Below we show how conditioning can be thought of as reshaping a base distribution.

A simple example of a joint probability distribution p(x, y), shown as a contour map, along with its two marginal 1-D probability distributions, p(x) and p(y). The highest points of p(x, y) are at (x₁, y₁) and (x₂, y₂). The conditional distributions p(x∣y = y₁) and p(x∣y = y₂) are shown overlaid on the main plot.

Consider the figure above. Think of p(x) as a distribution we want to sample from (i.e., the images) and p(y) as conditioning information (i.e., the text dataset). These are the marginal distributions of a joint distribution p(x, y). Integrating p(x, y) over y recovers p(x), and vice versa.

Sampling from p(x), we are equally likely to get x₁ or x₂. However, we can condition on p(y = y₁) to obtain p(x∣y = y₁). You can think of this as taking a slice through p(x, y) at a given value of y. In this conditioned distribution, we are much more likely to sample at x₁ than x₂.

In practice, in order to condition on a text dataset, we need to convert the text into a numerical form. We can do this using large language model (LLM) embeddings that can be injected into the noise prediction model during training.

Embedding text with an LLM

In the glyffuser, our conditioning information is in the form of English text definitions. We have two requirements: 1) ML models prefer fixed-length vectors as input. 2) The numerical representation of our text must understand context — if we have the words “lithium” and “element” nearby, the meaning of “element” should be understood as “chemical element” rather than “heating element”. Both of these requirements can be met by using a pre-trained LLM.

The diagram below shows how an LLM converts text into fixed-length vectors. The text is first tokenized (LLMs break text into tokens, small chunks of characters, as their basic unit of interaction). Each token is converted into a base embedding, which is a fixed-length vector of the size of the LLM input. These vectors are then passed through the pre-trained LLM (here we use the encoder portion of Google’s T5 model), where they are imbued with additional contextual meaning. We end up with a array of n vectors of the same length d, i.e. a (n, d) sized tensor.

We can convert text to a numerical embedding imbued with contextual meaning using a pre-trained LLM.

Note: in some models, notably Dall-E, additional image-text alignment is performed using contrastive pretraining. Imagen seems to show that we can get away without doing this.

Training the diffusion model with text conditioning

The exact method that this embedding vector is injected into the model can vary. In Google’s Imagen model, for example, the embedding tensor is pooled (combined into a single vector in the embedding dimension) and added into the data as it passes through the noise prediction model; it is also included in a different way using cross-attention (a method of learning contextual information between sequences of tokens, most famously used in the transformer models that form the basis of LLMs like ChatGPT).

Conditioning information can be added *via* multiple different methods but the training loss remains the same.

In the glyffuser, we only use cross-attention to introduce this conditioning information. While a significant architectural change is required to introduce this additional information into the model, the loss function for our noise prediction model remains exactly the same.

Testing the conditioned diffusion model

Let’s do a simple test of the fully trained conditioned diffusion model. In the figure below, we try to denoise in a single step with the text prompt “Gold”. As touched upon in our interactive UMAP, Chinese characters often contain components known as radicals which can convey sound (phonetic radicals) or meaning (semantic radicals). A common semantic radical is derived from the character meaning “gold”, “金”, and is used in characters that are in some broad sense associated with gold or metals.

Even with a single sampling step, conditioning guides denoising towards the relevant regions of the probability distribution.

The figure shows that even though a single step is insufficient to approximate the denoising trajectory very well, we have moved into a region of our probability distribution with the “金” radical. This indicates that the text prompt is effectively guiding our sampling towards a region of the glyph probability distribution related to the meaning of the prompt. The animation below shows a 120 step denoising sequence for the same prompt, “Gold”. You can see that every generated glyph has either the 釒 or 钅 radical (the same radical in traditional and simplified Chinese, respectively).

Takeaway

Conditioning enables us to sample meaningful outputs from diffusion models.

Further remarks

I found that with the help of tutorials and existing libraries, it was possible to implement a working diffusion model despite not having a full understanding of what was going on under the hood. I think this is a good way to start learning and highly recommend Hugging Face’s tutorial on training a simple diffusion model using their diffusers Python library (which now includes my small bugfix!).

I’ve omitted some topics that are crucial to how production-grade diffusion models function, but are unnecessary for core understanding. One is the question of how to generate high resolution images. In our example, we did everything in pixel space, but this becomes very computationally expensive for large images. The general approach is to perform diffusion in a smaller space, then upscale it in a separate step. Methods include latent diffusion (used in Stable Diffusion) and cascaded super-resolution models (used in Imagen). Another topic is classifier-free guidance, a very elegant method for boosting the conditioning effect to give much better prompt adherence. I show the implementation in my previous post on the glyffuser and highly recommend this article if you want to learn more.

Fun extras

Diffusion sampling using the DPMSolverSDEScheduler developed by Katherine Crowson and implemented in Hugging Face diffusers—note the smooth transition from noise to data.

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.

Broadcom grows revenues by 20% following VMware purchase, as customers fume about subscription costs

The company is “more than halfway” through the renewals, Tan said, adding, “we probably have at least another year plus, maybe a year and a half to go” in transitioning major accounts to VCF. Most of Broadcom’s VMware contracts are for about three years, which is “very traditional,” he said.

Nvidia aims to bring AI to wireless

Key features of ARC-Compact include: Energy Efficiency: Utilizing the L4 GPU (72-watt power footprint) and an energy-efficient ARM CPU, ARC-Compact aims for a total system power comparable to custom baseband unit (BBU) solutions currently in use. 5G vRAN support: It fully supports 5G TDD, FDD, massive MIMO, and all O-RAN

Netgear’s enterprise ambitions grow with SASE acquisition

Addressing the SME security gap The acquisition directly addresses a portfolio gap that Netgear (Nasdaq:NTGR) has identified through customer feedback. According to Badjate, customers have been saying that they like the Netgear products, but they also really need more security capabilities. Netgear’s target market focuses on organizations with fewer than

IBM’s cloud crisis deepens: 54 services disrupted in latest outage

Rawat said IBM’s incident response appears slow and ineffective, hinting at procedural or resource limitations. The situation also raises concerns about IBM Cloud’s adherence to zero trust principles, its automation in threat response, and the overall enforcement of security controls. “The recent IBM Cloud outages are part of a broader

Canada’s Oil Sands Emissions Intensity Falls for Sixth Year

Canada’s oil sands industry reduced its emissions per barrel for the sixth straight year in 2023, even as one growing portion of the sector moved in the opposite direction, according to new Alberta government data released Thursday. The emissions intensity of all oil sands sites fell to the equivalent of 0.399 metric tons of carbon dioxide per cubic meter of bitumen produced, down from 0.404 in 2022, the data show. The gain reflects improvements at oil sands mines, where bitumen is dug from the ground. However, in situ oil sands, which use wells similar to traditional oil producers, saw emissions per barrel rise. Even with the efficiency improvement, total emissions rose to the equivalent 80.1 million metric tons of carbon dioxide, up from 78.8 million in 2022, the data show. That’s the highest in data back to 2011. The oil sands’ declining energy intensity — to the lowest in data stretching back to 2011 — is welcome news for an industry that has struggled with a reputation for being climate unfriendly, prompting some investors to shun it altogether. However, the rising emissions intensity at well sites presents a challenge for the sector, as the method’s lower costs make it increasingly popular among producers. While the average intensity of oil sands producers is higher than the average for the global oil industry overall, drillers ‘emissions profiles vary widely around the world, said Kevin Birn, chief analyst for Canadian oil markets for S&P Global. The oil sands “fits well within the range of carbon intensity of oil and gas we see in the world,” Birn said in an interview. All of the oil sands mines reduced their emissions intensity with Canadian Natural Resources Ltd.’s Horizon making the biggest gain for the year. In situ production facilities, which include the more than 250,000 barrel-a-day Suncor

ICYMI: Energy Secretary Wright, EPA Administrator Zeldin, and Interior Secretary Burgum Visit Alaska to Advance Unleashing American Energy

In Case You Missed It— U.S. Secretary of Energy Chris Wright traveled to Alaska earlier this week, where he joined U.S. Department of the Interior Secretary Doug Burgum and U.S. Environmental Protection Agency (EPA) Administrator Lee Zeldin to advance President Trump’s bold agenda to unleash Alaska’s extraordinary resource potential. In Anchorage and on the North Slope, Secretary Wright met with state, local and tribal leaders as well as Alaskan energy workers, held bilateral meetings with foreign energy ministers to advance American energy projects like Alaska LNG, and participated in a keynote fireside chat at the Alaska Sustainable Energy Conference. “I was proud to join my colleagues, Doug Burgum and Lee Zeldin, in Alaska to help fulfill President Trump’s mission of unleashing Alaska to create a safe, prosperous, and opportunity-filled future,” said Secretary Wright. “Alaska has always been a big bold place, a big bold idea, full of big bold people. Unfortunately, the last administration especially and many before them sought to shut down Alaska in the name of climate nonsense. Those days are over. The Trump Administration is fully committed to restoring the rights and liberties of Alaskans, because when Alaska is unleashed, the entire country and world benefits.” Click here to watch a highlight video of the trip and here to download and view photos from the visit. Roundtable in Anchorage On Sunday, Secretaries Wright and Burgum and Administrator Zeldin joined Alaska Governor Mike Dunleavy and Senator Dan Sullivan (R-Alaska) for a roundtable discussion in Anchorage with leaders from across Alaska’s resource development industry. WATCH: Alaska Resources Roundtable with Secretary Burgum, Secretary Wright and Administrator Zeldin Town hall in Utqiagvik (Barrow), Alaska The delegation then traveled to the capital of Alaska’s North Slope Borough, Utqiagvik, Alaska, for a town hall with the predominantly Alaska-Native community. The cabinet members heard

Kenya Says Tullow Asset Sale ‘Raises Concern’ for Oil Sector

Kenya lawmakers said a lack of detail over Tullow Oil Plc’s sale of oil assets to local company Gulf Energy “raises concerns” for the industry in the East African nation. Tullow ended a decade-long quest to develop discoveries in April, when it sold the inland fields for $120 million to Gulf Energy, a Nairobi-based oil and gas trader. Tullow has announced divestments in recent months toward reducing its debt below $1 billion this year. The deal for the assets has included few details, “which raises concerns about the continuity, governance, and strategic direction of the country’s nascent oil industry,” the national assembly energy committee said in a report tabled on Wednesday. “There is limited information regarding the terms of the exit, the implications of the transaction on Kenya’s commercial oil prospects, and its potential impact on the completion and timely approval of the Field Development Plan,” it said. Gulf Energy is an oil supplier, electricity generator and infrastructure developer, according to its website. It didn’t immediately respond to emails seeking comment. The sale of working interests in Kenya to Gulf Energy “continues to progress well,” a Tullow spokesperson said in a reply to questions. “We have a high level of confidence in it completing with the receipt of the first two payments totaling $80 million expected during 2025.” Passage of the development plan, which is still pending, became one of the obstacles Tullow faced with the project. Partners TotalEnergies SE and Africa Oil Corp. exited the joint venture in 2023 and the company failed to find a strategic partner to replace them. The payment profile of the Gulf Energy deal “isn’t too demanding,” though it’s unclear who is backing the company, said Ashley Kelty, an analyst at Panmure Liberum. WHAT DO YOU THINK? Generated by readers, the comments included herein do not reflect

Oil Rises as Solid USA Jobs Data Pushes Algos to Drop Short Bets

Oil rose as stronger-than-expected US jobs data eased concerns about an economic slowdown that would crimp demand, spurring algorithmic traders to reduce short positions. West Texas Intermediate climbed almost 2% to settle above $64 a barrel, notching the largest weekly gain since November. Crude followed equities higher after US job growth in May narrowly surpassed economist forecasts, allaying concerns of near-term demand deterioration. The figures also pushed economy-sensitive diesel futures to a two-week high. “Trading is relatively quiet today, with macroeconomic factors continuing to drive the narrative,” said Rebecca Babin, a senior energy trader at CIBC Private Wealth Group. “The unemployment data is easing concerns that demand will sharply decline due to tariff uncertainty.” The positive economic data spurred commodity trading advisers to ease off of their bearish tilt. The funds, which can accelerate price momentum, liquidated short positions to sit at negative 9% short in WTI on Friday, compared with 64% short on June 5, according to data from Bridgeton Research Group. The rally was supported by enduring risk-on sentiment from optimistic signs on trade talks between the US and China, the world’s largest importer of crude. President Donald Trump and his Chinese counterpart, Xi Jinping, agreed to further negotiations over tariffs and supplies of rare earth minerals. The positive signals come against the backdrop of an oil market that has been increasingly rangebound in recent weeks. Prices have traded in a $5 band since the middle of May, and a gauge of volatility for US crude futures is at the lowest since early April. Oil has been buffeted in Trump’s second term as trade tensions between the world’s two largest economies menace demand. At the same time, the OPEC+ alliance has been adding barrels back to the market at a faster-than-expected rate, further clouding an already weak outlook for the second half

USA Targets Niche Gas That China Can’t Replace as Trade War Chip

The US is using its dominance of a niche petroleum gas as a bargaining chip in its trade war with China. America supplies China with almost all of its ethane, a product of the shale boom that’s used as a building block for making plastics. But the commerce department is now ordering shippers to apply for export licenses, and has told at least one, Enterprise Products Partners LP, that it intends to withhold permits for three China-bound cargoes. The trade war is throwing a spotlight on how the US and China rely on each other for certain commodities — dependencies that both nations are seeking to leverage as they negotiate terms to resolve their dispute. In this case, America is the world’s biggest producer of ethane, which is converted into ethylene for plastics factories, and China is its largest customer. The commerce department has cited risks that petroleum products like ethane could be diverted to the military, copying the playbook deployed by Beijing in justifying restrictions on what it calls dual-use items such as rare earths and other critical minerals. “Ethane is no longer just a byproduct of shale — it’s now a geopolitical weapon,” said Julian Renton, lead analyst covering natural gas liquids at East Daley Analytics. “China bet billions building infrastructure around US ethane, and Washington is now questioning whether that bet should continue to pay off.” America’s shale revolution and China’s rapid industrialization have coincided this century to create a market where cheap energy byproducts are parlayed into millions of tons of materials used as trash bags and shampoo bottles, car seats and computer keyboards. But companies that prospered from cooperation are now caught in the crossfire of an increasingly antagonistic trade relationship between Washington and Beijing. Chinese firms such as Satellite Chemical Co. operate giant petrochemical plants that

Alaska LNG Attracts Potential Partners in Asia, EU

The first round of partner selection for the planned Alaska LNG project attracted over 50 companies, lead developer Glenfarne Group LLC has said. The potential partners are from the United States, the European Union, Japan, Korea, Taiwan, Thailand and India. The companies “expressed interest for over $115 billion of contract value for various partnerships with the Project, including equipment and material supply, services, investment, and customer agreements”, Glenfarne said in an online statement. Alaska LNG, approved by the Federal Energy Regulatory Commission May 2020, will deliver natural gas from the state’s North Slope to both domestic and global markets. It is the only federally permitted liquefied natural gas (LNG) project on the United States Pacific Coast, according to co-developer Alaska Gasline Development Corp. (AGDC). “Alaska LNG’s economic fundamentals allow it to deliver LNG into Asia at prices that are lower than Henry Hub pricing from the U.S. Gulf Coast”, the statement said. Alaska LNG is planned to have an LNG export terminal with a capacity of 20 million metric tons per annum (MMtpa), an 807-mile 42-inch pipeline and a carbon capture plant with a storage capacity of seven MMtpa. Phase 1 aims to deliver gas about 765 miles from the North Slope to the Anchorage region. Phase 2 would install compression equipment and around 42 miles of pipeline under Cook Inlet to the Alaska LNG Export Facility in Nikiski, which would be constructed at the same time. “Glenfarne anticipates a final investment decision on the domestic portion of the Alaska LNG pipeline in late Q4 2025”, the statement said. Brendan Duval, chief executive and founder of New York City-based energy investor Glenfarne, commented, “The many expressions of interest received reinforce that the market recognizes Alaska LNG’s advantaged economics, fully permitted status, and powerful federal, state, and local support”. The project’s

LiquidStack launches cooling system for high density, high-powered data centers

The CDU is serviceable from the front of the unit, with no rear or end access required, allowing the system to be placed against the wall. The skid-mounted system can come with rail and overhead piping pre-installed or shipped as separate cabinets for on-site assembly. The single-phase system has high-efficiency dual pumps designed to protect critical components from leaks and a centralized design with separate pump and control modules reduce both the number of components and complexity. “AI will keep pushing thermal output to new extremes, and data centers need cooling systems that can be easily deployed, managed, and scaled to match heat rejection demands as they rise,” said Joe Capes, CEO of LiquidStack in a statement. “With up to 10MW of cooling capacity at N, N+1, or N+2, the GigaModular is a platform like no other—we designed it to be the only CDU our customers will ever need. It future-proofs design selections for direct-to-chip liquid cooling without traditional limits or boundaries.”

Enterprises face data center power design challenges

” Now, with AI, GPUs need data to do a lot of compute and send that back to another GPU. That connection needs to be close together, and that is what’s pushing the density, the chips are more powerful and so on, but the necessity of everything being close together is what’s driving this big revolution,” he said. That revolution in new architecture is new data center designs. Cordovil said that instead of putting the power shelves within the rack, system administrators are putting a sidecar next to those racks and loading the sidecar with the power system, which serves two to four racks. This allows for more compute per rack and lower latency since the data doesn’t have to travel as far. The problem is that 1 mW racks are uncharted territory and no one knows how to manage the power, which is considerable now. ”There’s no user manual that says, hey, just follow this and everything’s going to be all right. You really need to push the boundaries of understanding how to work. You need to start designing something somehow, so that is a challenge to data center designers,” he said. And this brings up another issue: many corporate data centers have power plugs that are like the ones that you have at home, more or less, so they didn’t need to have an advanced electrician certification. “We’re not playing with that power anymore. You need to be very aware of how to connect something. Some of the technicians are going to need to be certified electricians, which is a skills gap in the market that we see in most markets out there,” said Cordovil. A CompTIA A+ certification will teach you the basics of power, but not the advanced skills needed for these increasingly dense racks. Cordovil

HPE Nonstop servers target data center, high-throughput applications

HPE has bumped up the size and speed of its fault-tolerant Nonstop Compute servers. There are two new servers – the 8TB, Intel Xeon-based Nonstop Compute NS9 X5 and Nonstop Compute NS5 X5 – aimed at enterprise customers looking to upgrade their transaction processing network infrastructure or support larger application workloads. Like other HPE Nonstop systems, the two new boxes include compute, software, storage, networking and database resources as well as full-system clustering and HPE’s specialized Nonstop operating system. The flagship NS9 X5 features support for dual-fabric HDR200 InfiniBand interconnect, which effectively doubles the interconnect bandwidth between it and other servers compared to the current NS8 X4, according to an HPE blog detailing the new servers. It supports up to 270 networking ports per NS9 X system, can be clustered with up to 16 other NS9 X5s, and can support 25 GbE network connectivity for modern data center integration and high-throughput applications, according to HPE.

AI boom exposes infrastructure gaps: APAC’s data center demand to outstrip supply by 42%

“Investor confidence in data centres is expected to strengthen over the remainder of the decade,” the report said. “Strong demand and solid underlying fundamentals fuelled by AI and cloud services growth will provide a robust foundation for investors to build scale.” Enterprise strategies must evolve With supply constrained and prices rising, CBRE recommended that enterprises rethink data center procurement models. Waiting for optimal sites or price points is no longer viable in many markets. Instead, enterprises should pursue early partnerships with operators that have robust development pipelines and focus on securing power-ready land. Build-to-suit models are becoming more relevant, especially for larger capacity requirements. Smaller enterprise facilities — those under 5MW — may face sustainability challenges in the long term. The report suggested that these could become “less relevant” as companies increasingly turn to specialized colocation and hyperscale providers. Still, traditional workloads will continue to represent up to 50% of total demand through 2030, preserving value in existing facilities for non-AI use cases, the report added. The region’s projected 15 to 25 GW gap is more than a temporary shortage — it signals a structural shift, CBRE said. Enterprises that act early to secure infrastructure, invest in emerging markets, and align with power availability will be best positioned to meet digital transformation goals. “Those that wait may find themselves locked out of the digital infrastructure they need to compete,” the report added.

Cisco bolsters DNS security package

The software can block domains associated with phishing, malware, botnets, and other high-risk categories such as cryptomining or new domains that haven’t been reported previously. It can also create custom block and allow lists and offers the ability to pinpoint compromised systems using real-time security activity reports, Brunetto wrote. According to Cisco, many organizations leave DNS resolution to their ISP. “But the growth of direct enterprise internet connections and remote work make DNS optimization for threat defense, privacy, compliance, and performance ever more important,” Cisco stated. “Along with core security hygiene, like a patching program, strong DNS-layer security is the leading cost-effective way to improve security posture. It blocks threats before they even reach your firewall, dramatically reducing the alert pressure your security team manages.” “Unlike other Secure Service Edge (SSE) solutions that have added basic DNS security in a ‘checkbox’ attempt to meet market demand, Cisco Secure Access – DNS Defense embeds strong security into its global network of 50+ DNS data centers,” Brunetto wrote. “Among all SSE solutions, only Cisco’s features a recursive DNS architecture that ensures low-latency, fast DNS resolution, and seamless failover.”

HPE Aruba unveils raft of new switches for data center, campus modernization

And in large-scale enterprise environments embracing collapsed-core designs, the switch acts as a high-performance aggregation layer. It consolidates services, simplifies network architecture, and enforces security policies natively, reducing complexity and operational cost, Gray said. In addition, the switch offers the agility and security required at colocation facilities and edge sites. Its integrated Layer 4 stateful security and automation-ready platform enable rapid deployment while maintaining robust control and visibility over distributed infrastructure, Gray said. The CX 10040 significantly expands the capacity it can provide and the roles it can serve for enterprise customers, according to one industry analyst. “From the enterprise side, this expands on the feature set and capabilities of the original 10000, giving customers the ability to run additional services directly in the network,” said Alan Weckel, co-founder and analyst with The 650 Group. “It helps drive a lower TCO and provide a more secure network.” Aimed as a VMware alternative Gray noted that HPE Aruba is combining its recently announced Morpheus VM Essentials plug-in package, which offers a hypervisor-based package aimed at hybrid cloud virtualization environments, with the CX 10040 to deliver a meaningful alternative to Broadcom’s VMware package. “If customers want to get out of the business of having to buy VM cloud or Cloud Foundation stuff and all of that, they can replace the distributed firewall, microsegmentation and lots of the capabilities found in the old VMware NSX [networking software] and the CX 10k, and Morpheus can easily replace that functionality [such as VM orchestration, automation and policy management],” Gray said. The 650 Group’s Weckel weighed in on the idea of the CX 10040 as a VMware alternative:

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