J.C. Baillie
@jcbaillie
Followers
3K
Following
1K
Media
109
Statuses
2K
Founder of Developmental Labs & founder of @Novaquark, maker of @dualuniverse. AI & Robotics researcher. Former head of Softbank Robotics Europe AI Lab.
Around the world
Joined December 2009
A beautiful tribute to Dual Universe here: https://t.co/g8nsttgBIq
3
2
3
"Dual Universe, the game that will live forever". I like that! https://t.co/Ebj40oowKT
1
1
11
The Era of Experience & The Age of Design Great talk by Richard Sutton: Humans have the ability to take design to the next level, and actually design things that can design (“make AI that can make AI”), taking us from Replicators into “The Age of Design”. https://t.co/MLL4kDeH3T
8
38
272
Just reread Permutation City by @gregeganSF. I'm convinced that a deeper understanding of consciousness sits at the crossroads of dust theory, digital physics, many-worlds & computational irreducibility. Might turn these thoughts into a blog post. Anyone else exploring this?
1
0
4
Exactly my workflow with AI coding at the moment (mostly for python code, for my C++ projects I barely touch AI). Don't fall for the "vibe coding" hype. Using AI requires work, and possibly improves code quality if you take the time to understand what's suggested (it shows you
Noticing myself adopting a certain rhythm in AI-assisted coding (i.e. code I actually and professionally care about, contrast to vibe code). 1. Stuff everything relevant into context (this can take a while in big projects. If the project is small enough just stuff everything
0
0
1
Today, we're releasing ARC-AGI-2. It's an AI benchmark designed to measure general fluid intelligence, not memorized skills – a set of never-seen-before tasks that humans find easy, but current AI struggles with. It keeps the same format as ARC-AGI-1, while significantly
56
326
2K
Information is not an intrinsic property of any given system A. Rather, it should be defined in relation to another system B, as a measure of how much causal influence A can have on B. For example, a book does not contain information from the perspective of a dog, but from the
0
0
1
As you query LLMs on more and more complex problems that you could handle yourself but decide to delegate, there is an interesting trade-off that I've not seen discussed: the more complex the problem, the longer it will take you to check the response of the LLM (typically, on
2
1
3
Complete hardware + software setup for running Deepseek-R1 locally. The actual model, no distillations, and Q8 quantization for full quality. Total cost, $6,000. All download and part links below:
722
4K
28K
Everybody in AI should read this: a very interesting article from @katbeuls and @paulvaneecke, which presents an alternative and exciting approach to language learning compared to LLMs:
direct.mit.edu
Abstract. Humans acquire their native languages by taking part in communicative interactions with their caregivers. These interactions are meaningful, intentional, and situated in their everyday...
1
3
11
Just finished reading "There Is No Antimemetics Division" by @qntm, really great sf work. The only problem is, as soon as you finish reading it, you cannot remember it!😶Good that there is a free copy here:
qntm.org
An important announcement There Is No Antimemetics Division has been picked up for publication by Penguin Random House! Find out more here. Everything below is for the original self-published version...
0
0
2
What is DeWeb ? 🔍 The internet's freedoms are under threat, central hosting providers can control and restrict your data instantly. 💾❌ Governments are increasingly censoring online content.📵 Onion services have been the go-to for hosting censorship-resistant applications,
8
56
130
I think we've been deprived of a lot more than flying cars and Mars colonies. I think civilization is currently burdened by a debilitating pessimism. Not just prophecies of doom, because they've always existed. There's something deeper. The term technological fix has become as
14
84
394
I'm often described (due to my work on computable physics) as being an advocate for the Simulation Hypothesis. But I'm not. In fact, I think the Hypothesis is nonsensical, and that the most famous argument in its favor actually proves the opposite of what it purports. (1/10)
109
250
2K
Searching the computational universe with a better-than-random bias. This is the way!
Introducing ASAL: Automating the Search for Artificial Life with Foundation Models https://t.co/uUq63UNrjv Artificial Life (ALife) research holds key insights that can transform and accelerate progress in AI. By speeding up ALife discovery with AI, we accelerate our
0
0
1
Major result here. I must admit I'm surprised, but I agree with François that this is not yet AGI. I have this impression that, in fact, any well-defined test can be beaten eventually. But I still believe that "true" AGI would require embodiement, developmental learning, world
Today OpenAI announced o3, its next-gen reasoning model. We've worked with OpenAI to test it on ARC-AGI, and we believe it represents a significant breakthrough in getting AI to adapt to novel tasks. It scores 75.7% on the semi-private eval in low-compute mode (for $20 per task
0
0
1
I had the pleasure of sitting down with @DavidDeutschOxf in his lovely garden in Oxford a few months ago. Here’s our conversation. Links to watch on YouTube or listen on popular podcast platforms below. Timestamps: 1:01 - A tragic view of human history 7:04 - Why did it take so
48
183
931
Bootstrapping Developmental AIs From Simple Competences to Intelligent, Human-Compatible AIs by @MarkStefik and Robert Price https://t.co/Ffv0YVDsl9
0
1
10