We're headed for a dual singularity, and recursive self-improving embodied cognition is going to be much more important than recursive self-improving LLMs.
The race for the most important results in science is a two-horse race between:
- LLMs backed by trillions of dollars, armies of geniuses, and the full weight of capitalism
and
- Michael Levin
My money is on Michael Levin.
5
194
Jonathan McCoy retweeted
🧵 Elon: “Longevity is an extremely solvable problem”
At Vitalist Bay (May 14 - May 17), we’re solving it!
1K+ pioneers, 100+ speakers, 50+ workshops, 40+ activities, 5 critical health tests
Join us to spark dozens of SpaceXs for longevity!
Jonathan McCoy retweeted
Being a scientist is still the best job in the world. The constant frustrations, all the crap that goes with trying to keep the lab alive and move things forward - it all resets and washes away in those occasional moments when we catch a glimpse of nature revealing something remarkable and not seen before. This has been a good week - I saw a few completely amazing, wild new things; it will be a total slog to get them polished and out, but it doesn't matter because they've been seen and a new piece of the roadmap is revealed. Just feeling immensely grateful for the opportunity, and mentally sending a message back in time to childhood me (and anyone else who dreams of this now).
Jonathan McCoy retweeted
Couldn't be happier to share the news of FRONT's launch, an initiative to repair brain tissue led by fellow Vitalists at ARPA-H.
Onwards!
NEW: No technology currently exists to repair damaged brain tissue and fully restore lost function. FRONT aims to enable millions with what is considered permanent brain damage to regain lost functions, including motor control, vision, and speech. arpa-h.gov/news-and-events/a…
Epistemic evals - towards interpretability of users...
@ttorres just seeing this - very helpful and I totally agree!
I've been experimenting with an evaluation approach in this direction. Roughly, the idea is to categorize opportunities by cohort intent. The hierarchy provides constraints for the prompt as a concept that could make it easier to collect failure mode annotations from the developer, SMEs or even users.
By abstracting by user intent, I was planning to perform evals on the semantic router. ie, I'll be training my router to identify the active user opportunity, and then routing them to the prompts, tools and retrievers I've evaluated in the context of delivering to specific value props.
I asked ChatGPT about the ways that synthetic biology and AI could create a risk to agriculture and food supply chains.
It did not mention mirror life.
What I asked about mirror life it shut down the thread.
Is this safety or intellectual dishonesty?
Jonathan McCoy retweeted
“Objectivity is the illusion that observations can be made without an observer.”
- Heinz von Foerster, an Austrian-American physicist and cybernetician
Self-models with belief systems give context to augment experience. Pragmatic philosophy of experience will get us there.
Google DeepMind's Logan Kilpatrick says AGI will be a product experience. Not a model.
His bet: whoever nails memory + context around decent model at a product level wins. Users will suddenly feel like they're talking to AGI.
Not from capability breakthrough, but experience breakthrough.
At some point, the AGI race is actually a UX problem.
1
85
I just completed the course AI Evals for Engineers and PMs by @HamelHusain and @sh_reya
My goal was to expand my understanding of AI eval best-practices and apply it to Epistemic Me's personalized recommendation approach.
Key takeaways and links below..
Here are some of my bookmarks shared in the discussion forums:
1. Why AI Eval Startups Fail - thomasliao.com/eval-startups
2. Interactive LLM-Powered Data Processing with DocWrangler - data-people-group.github.io/…
3. LLM Evaluation doesn't need to be complicated - philschmid.de/llm-evaluation
4. Evaluating the Effectiveness of LLM-Evaluators (aka LLM-as-Judge) - eugeneyan.com/writing/llm-ev…
5. Statistical Process Control: A Practitioner’s Guide - entropicthoughts.com/statist…
6. Who Validates the Validators? Aligning LLM-Assisted Evaluation of LLM Outputs with Human Preferences - arxiv.org/abs/2404.12272
1
4
92
Jonathan McCoy retweeted
People looking for a theory of everything that doesn’t include biology are just trying to one up Einstein (which is stupid) when really they should be extending Darwin. A theory of everything that doesn’t include the origin of the observer is a theory of nothing.
The mind leading AlphaFold is asking for innovative philosophy - repeatedly and urgently.
Demis Hassabis: "I think there is a need for some great philosophers. Where are they? The great next philosophers... I think we're going to need that to help navigate society to that next step. AGI and ASI is going to change humanity and the human condition."
2
150
"The formation of groups by single-cell organisms to improve survival is viewed as an important early step in the evolution of multicellularity."
nytimes.com/2025/03/31/scien…
78













