Cofounder & Chief Strategy Officer @ Stealth

Houston, TX
Joined September 2010
🕹️
Qualitative sample Pong gameplay of a transformer model, IRIS, which trained for 2 days compared to an agent, powered by Genius, which trained for just 2 hours both on 90% less data than the standard Atari 100k benchmark. 🔗 Full blog post: verses.ai/research-blog/achi…
1
8
2,683
I’m thrilled to share what I’ve been cooking up. My team is launching a new kind of social network – think less content, more commerce – and we’re offering early access to a select group of trailblazers and investors. Sign up now for consideration. list-lab.org/m/steven?follow…
🤖 Made with AI
1
6
104
Intelligence and understanding powered by curiosity and imagination. 🤔 #ActiveInference #FEP
Artificial Intelligence Made Simple expert @Machine01776819 has featured VERSES today in his in-depth analysis of reinforcement learning and its alternatives. He says "Karl Friston’s framework naturally seeks out information that resolves uncertainty about its world model. It’s curious by design, not by engineering patch. You don’t need curiosity hacks bolted onto reward maximization. Active Inference agents are also robust to changing goals. Your primary objective is understanding... You’re not locked into whatever reward function someone wrote on day one."
3
315
Steven Swanson retweeted
Either you crack general intelligence -- the ability to efficiently acquire arbitrary skills on your own -- or you don't have AGI. A big pile of task-specific skills memorized from handcrafted/generated environments isn't AGI, not matter how big.
New post: Thoughts on AI progress (Dec 2025) 1. What are we scaling?
101
110
23
1,173
118,977
Steven Swanson retweeted
Former New York Times Technology journalist, Craig Smith, featured Karl Friston in his Eye on AI interview this week. The wide ranging interview covered: 1. How the brain works by making predictions - and what this means for AI. 2. VERSES' AXIOM digital brain. 3. Real world applications of Genius - including our work with Analog. 4. How representing uncertainty is critical - because it allows systems to know what they don't know - and therefore judge an action by how much it reduces uncertainty.
1
5
17
1,209
Steven Swanson retweeted
Jensen Huang on being a CEO. Strategy means deciding what not to pursue. The work brings pain, determination, and endurance. It’s an honorable job, but built more on suffering than fame.
4
7
1
39
5,038
Alright, let's quantify it. AI bubble, yay or nay?
100%Definitely
0%Probably yes
0%Probably no
0%Definitely not
1 votes • Final results
116
Steven Swanson retweeted
A video made by pixl8studio recreates the impossible, maze-like world from M.C. Escher’s famous artwork “Relativity.”
34
405
24
1,788
100,777
Steven Swanson retweeted
To spot consciousness, we must provoke it. Agency is the system's fight to minimize surprise (Friston's Free Energy Principle). ​When an agent breaks its pattern, it causes a massive belief update (high KL Divergence). This is only meaningful if the resulting action has high, non-algorithmic complexity (Kolmogorov). The energy required to process this surprise and update the model is dictated by Landauer's Limit. ​In real life, this "Bayesian inference" isn't digital: biological systems may use analog resonance—where internal frequencies (priors) couple with data—to instantly estimate new beliefs. Consciousness is the energy-costly process of constantly managing this chaotic update cycle. 🤯 AI generated image [Gemini].
15
29
2
191
8,411
Steven Swanson retweeted
NEWS: VERSES Recognized in Gartner® 2025 Emerging Tech Impact Radar: Spatial AI “The Spatial Web, where digital + physical systems converge, is the infrastructure needed to enable the next frontier in AI.” — Gabriel René, CEO 🔗 hubs.li/Q03PymVQ0 #AI #Gartner
5
1
20
1,287
Steven Swanson retweeted
The reason it is so important for everyone to keep pretending that AGI is definitely right around the corner is that there is now over $1T of investment riding on this belief (either already expended, or committed) Current (and recent past) capex cannot be justified by current use cases and technology (currently spending $10-15 to make $1). To ever be in the black you'd need dramatically better tech/applications, and you'd need them fast -- before current datacenters depreciate, which is a 3-5 years timescale
148
284
65
2,161
338,446
Steven Swanson retweeted
nothing to worry about
16
36
3
315
28,821
Steven Swanson retweeted
Our CEO @greal1111 & Chief Scientist Karl Friston are featured in the Financial Times today discussing alternatives to LLMs. "One expert proposing a different path is Karl Friston... chief scienti... Verses, who thinks we still have much to learn from biological intelligence... But he argues they will never achieve AGI as they “have no agency under the hood.. That can only be achieved through 'active inference', which means having a theory of the world and the ability to predict and adapt." "As an example, generative AI models might be good at analysing historic patterns of traffic flows in a city. But what happens when Taylor Swift comes to town? A truly intelligent system has to have a constantly evolving model of how the world works and predict the likely effect of novel factors"
1
3
1
17
1,060
Richard Sutton, author of "The Bitter Lesson", now has a better lesson: "The majority of high-quality data sources...have either already been, or soon will be consumed." theaiinnovator.com/welcome-t…
1
222
I like to think of this research as giving autonomous vehicles #imagination. 🚘🤔✨
Dealing with objects that are hidden from view is one of the biggest hurdles to making self-driving vehicles safe on the road. Today we've published more details about our work on this challenge, which we have undertaken in collaboration with researchers from Volvo Cars Safety Centre. verses.ai/blog/trajectory-pr…
1
3
274
How many r's in Agentforce?
Marc Benioff said AI would be easy for Salesforce customers to adopt, but that hasn’t been the case. Now Salesforce is relaunching a program whose former members irked Benioff last fall for not staying on-message. More here in our latest Salesforce story theinformation.com/articles/…
234
Steven Swanson retweeted
oops. who could have seen this coming?!
52
132
5
882
50,695
For anyone who genuinely understands #robotics (or machine learning in general) this is NOT trivial. Huge props to the research team at @helloVERSES ! 🤖🦾🦿
VERSES® Unveils Robotics Architecture that Works Without Pre-Training. VERSES multi-agent robotics model outperforms current methods on Meta’s Habitat Benchmark without pre-training.
1
1
1
10
1,069
The more things change, the more they stay the @sama
630