Five Things AI: Distrust, Data Centers, Coupled Oscillators, Knowledge Agents, World Model Trained Agents
There is so much AI out there! Find out what it does!
Heya and welcome back to Five Things AI!
This week the AI story splits in two directions at once. On one side, the money finally works: revenue is clearing the cost of the hardware, generative AI is scaling three times faster than the cloud ever did, and the buildout that looked like a leap of faith is starting to pencil out. On the other side, the mood is souring and the engineering is getting smarter. People are asking why they should love a technology that concentrates wealth and burns through power, while researchers quietly demonstrate that you don't need a trillion-parameter Nobel laureate to do most of the work. Coupled oscillators that compute with physics at a fraction of the energy, knowledge agents that beat frontier models with better structure, world models that make agents sharper without ever training them as agents. The pattern this week: the economics validate the giants, but the cleverness is migrating to whoever builds lean. Here are five things worth your attention.
Why Does Everyone Hate AI?
Finally, AI is tightly linked in the public mind with the tech oligarchs who are pushing it. There is widespread awareness of the growing concentration of wealth and power at the top and how this is distorting our politics and harming our society. Aside from the MAGA faithful, Americans overwhelmingly favor government policies to reduce wealth inequality:
And AI is widely perceived, for good reason, as a technology that will increase the concentration of wealth at the top. Indeed, as I said, the AI companies themselves have already told us that the technology will have extremely negative effects on workers.
I am certain that AI will bring plenty of positive effects and will revolutionize the way we work and live. Founders like Altman and Amodei need to calm down the fears and focus more on the amazing potential AI has in medicine and in science in general. And while data centers are consuming plenty of energy, I believe that AI will lead to numerous breakthroughs that will reduce global warming.
AI Demand Begins to Justify Massive Cost of Data-Center Buildout
Much of the AI boom has been measured from the supply side, through disclosures from public semiconductor companies like Nvidia Corp. and hyperscalers like Alphabet. Demand has been harder to quantify because many of the most important AI labs, including OpenAI and Anthropic, remain private.
Generative AI revenue, excluding China, reached $110 billion over the past 12 months and is scaling three times faster than any previous information technology wave including the internet, mobile applications and the cloud, according to the report.
When we look at the Gartner hype cycle, we can assume that AI will quickly reach the plateau of productivity, which we can also see by looking at the revenue numbers.
Introducing Un-0: Generating Images with Coupled Oscillators
At Unconventional AI, we’re building a new kind of computer, one that harnesses the laws of physics to do the computing. Our goal is to run modern AI on a fraction of the energy today’s machines need, around 1,000x less. As a first step, we ask: can we train a physical dynamical system to generate images at scale?
The best AI models today are conventional deep networks with transformer backbones. However, there is also a long history of alternatives that seek energy efficiency by leveraging the dynamics of a physical system, such as the noisy, time-varying behavior of analog circuits that compute with analog voltage and current instead of conventional digitized numbers.
Wow. I’m intrigued. The power consumption of traditional AI models is extreme and even if they don’t get 1000s less, it would be huge.
Knowledge Agents: Beat Frontier Models with Better Structure
First, a significant portion of frontier models’ huge footprint is “knowledge.” I might call it pseudo-knowledge, since it’s probabilistic and there’s no guarantee it’ll give you the right answer... but the biggest models have been trained on an enormously broad set of data. This is captured in numerical weights as “parametric” knowledge. While that’s very useful if you’re casually asking Claude Opus or GPT-5.5 about some random topic, it’s entirely irrelevant if you either already have the data you want to reference or the data isn’t publicly available anyway—so it could never train on it. The latter is quite common in fields that are specialist (areas of medical research), secretive (high finance), or proprietary (frontier, company-specific materials science). If I don’t need it... well, a lot of the massive size of the model to cover every random subject matter is a huge waste.
I liked this article a lot, basically because somebody finally pointed out that just using the best LLM out there is a bit lazy - and expensive.
Alibaba’s model never trained as an agent — and improved agent performance across seven benchmarks
The research team trained agents inside the resulting simulator and found performance gains that exceeded what training against real environments alone produced. In a separate test, using world model training as a warm-up before agentic fine-tuning improved performance across seven benchmarks, including three the model had never seen during training.
The paper accompanying the release identified a gap in prior agent research. “We argue that world modeling is a crucial missing piece in the path to general agents.”
This means that we will see another boost in Agentic AI soonish. We live in interesting times.
