Bartek Wlodarczyk (@bartek_wl) header
Bartek Wlodarczyk (@bartek_wl) profile picture

Co-founder @skyengineai. PhD, Medical Physics in Clinical Radiation Oncology and Imaging. AI @stanford. I like pixels and sounds and to train AI models.

London
Joined September 2025
Photo by Bartek Wlodarczyk (@bartek_wl)Photo by Bartek Wlodarczyk (@bartek_wl)Photo by Bartek Wlodarczyk (@bartek_wl)Photo by Bartek Wlodarczyk (@bartek_wl)Photo by Bartek Wlodarczyk (@bartek_wl)Photo by Bartek Wlodarczyk (@bartek_wl)Photo by Bartek Wlodarczyk (@bartek_wl)Photo by Bartek Wlodarczyk (@bartek_wl)Photo by Bartek Wlodarczyk (@bartek_wl)Photo by Bartek Wlodarczyk (@bartek_wl)
You can't ban recursive self-improvement (RSI) when it hasn't even been demonstrated. Statistical models recursively trained on their own outputs risk model collapse, not runaway intelligence. Self-improvement requires more than recycling statistical predictions.
1
1
150
Show me a statistical model that autonomously and indefinitely improves its own intelligence without external grounding, new information or human intervention. Until then, you're regulating a hypothetical, not a capability.
43
AI has intelligence-like computation but lacks continuously existing, recurrently integrated, self-modeling, intrinsically motivated stream of cognition + we don't yet know whether adding that would create consciousness or merely an extremely convincing functional imitation of it
1
1
180
Intelligence is not what you know, it is what you do when you don't know. And AI won't have this ability for years to come.
1
5
203
How did we get a potential AI apocalypse before we ever got a usable customer support chatbot?
1
3
193
This is what Dramodei and Altman fear: - Cost of fine tuned open source models trained on custom data was 95% less than frontier models - And they perform better than frontier models - And they could train them in less than 48 hours Big AI labs have no moat.
4
157
Bartek Wlodarczyk retweeted
This is what Dario and Sama fear - - the cost of fine tuned open source models trained on custom data was 95% less than the frontier models - And they performed better than the frontier models - And they could train them in less than 48 hours The big AI labs have no moat.. Even at the enterprise level.. Any sane company would prefer a custom trained open source model instead of paying API pricing to Anthropic and Open AI If open source keeps growing at the pace it is, frontier labs' business collapses.. only way out is oldest trick in manipulation - FUD -> Fear, Uncertainty, Death Step 1 - Seed the psyop - Test a swarm of agents trained to "hack" - The swarm does what its supposed to do - hack - Publish a report - AI is too dangerous Step 2 - weed out the doubters - Get an employee to quit your company - Go viral saying both companies are not doing enough to self regulate AI - AI is dangerous - will kill humanity by the end of the decade Step 3 - Prepare for fruition - write an "essay" to "pace the frontier" - Advocate regulation - Only a few holier than though companies get to build AI - Scare the world into agreeing with AI Doom - Effectively turn your competitors into criminals (genius business strategy) Endgame - - Become a cartel that controls AI (and the world) - Raise prices (coz business is unsustainable... ofc) - IPO (offload liability of said unsustainable business to the public) Laugh all the way to the bank 🤑 Win!
.@L3HarrisTech says Palantir helped them beat frontier AI models in less than 2 days, at 95% lower cost: "When we fine-tuned open source models trained on our own data, we were able to outperform the frontier models in less than 48 hours." "The cost of our fine-tuned open source model was 95% lower than the frontier models we were using." "AI is a commodity. It's all about the data. We view our data as a corporate asset. It's our unique hard-earned knowledge." "We believe American defense companies should not be a vassal for frontier AI labs, handing over our data and institutional knowledge, hoping to rent back the intelligence it creates." " We own the model, we own the compute, we own the advantage."
206
1,591
188
7,930
819,533
Fear that AI will destroy humanity is absurd. AI models don't understand physical world, nor are they capable of planning or reasoning. Major labs fuel narrative of existential threat for 2 reasons: to justify their valuations and to lobby for regulations to shut out open-source.
1
4
984
Astra is not AGI, let alone ASI. It is a highly capable AI system operating within a bounded architecture, trained objective and deployment stack. It's not general intelligence, autonomous scientific reasoning, open-ended self-improvement or superhuman competence across domains.
8
314
“AI will cure cancer” hype is mostly nonsense. Cancer isn’t one disease, biology isn’t a benchmark and predicting molecules is not the same as proving safety, efficacy, delivery, resistance or survival benefit in humans. AI can accelerate research. It does not abolish biology.
1
8
430
LLMs using external code harness are not neurosymbolic AI. NNs and symbolic representations shall be integrated into a unified system where symbolic knowledge and reasoning are core components of the model's inference or learning process.
2
2
343
We’re not in an AI singularity. Today’s models are powerful pattern engines, but they still depend on human-defined goals, infrastructure, data and oversight. Rapid progress is not the same as autonomous, self-sustaining intelligence improving beyond human control.
5
355
1/ There is no future version where LLMs retain full distributional flexibility and become impossible to jailbreak. The jailbreak isn't bug. It is the feature and the price of trying to sell interpolation as intelligence while pretending policy wrappers are actual guardrails.
1
3
685
10/ The reality is that there is no final fix waiting to be discovered. You can narrow the distribution. You can overfit refusals. You can increase false positives. You can make the model progressively more useless around anything near the boundary.
1
1
359
11/ But the more successful you are at doing that, the less of a still useful LLM you have left. That is the part the industry keeps refusing to admit or COMPREHEND ffs.
209