speaker: Mark Hamalainen
talk title: Longevity Bio Fellowship
event: Foresight Institute AI x Bio Workshop 2025
video: https://www.youtube.com/watch?v=rQalRF1MSsM
LLM summary: The speaker outlines the stagnation in aging research despite two decades of effort, attributing it to biology's outdated non-quantitative paradigms, the pharmaceutical industry's misalignment with complex, non-reductionist challenges like aging—which exceeds baseline biology in complexity as unoptimized byproducts—and severe funding shortages due to long ROI timelines. Critiquing failed reductionist theories (e.g., telomeres, antioxidants) and non-translatable model organism lifespan extensions, the Longevity Biotech Fellowship (LBF) prioritizes biostasis, organ replacement, and advanced bioengineering over pharmaceuticals, leveraging AI for acceleration in cryoprotectant optimization, connectomics, and data modeling. Key bottlenecks include massive data needs (far beyond the Protein Data Bank's multi-billion-dollar cost), automated perturbation experiments for genetic CAD/digital twins, and gene delivery; recent initiatives like the Thalion Initiative's $710M comparative study offer hope, but LBF aims to catalyze dozens more high-impact entities through expert cohorts and networking.
Introduction and LBF Purpose
So thanks Alison for the intro and it's awesome working with Foresight Institute and Vitalia (Vitalism?). My co-founder Nathan is here as well at the LBF who also is a co-founder of Vitalism.
What is the point of the LBF? The LBF is very concerned about the rate of progress on solving ?aging and we really don't like this idea of mortality. I've been in the industry for over 20 years now since I started doing research. We still have zero approved treatments for aging. People don't realize how bad the situation is and we think it's an emergency and it doesn't matter how much sauna, lifting, supplements, Zone 2 training you do, you're going to only shift this a couple pixels. Let's be honest with ourselves, come on guys.
Core Problems in Aging Research
So what's the problem? We have an outdated paradigm I think and the way that we do biology. Biology is one of the last sciences to become quantitative, to become data-driven, to become built using models. We have this industry called the pharmaceutical industry and it has a lot of weight and inertia and institutional power but it's not designed to solve aging. It was never designed to solve a problem like aging.
Aging is just fundamentally very, very difficult. It's more difficult to understand aging than it is to understand biology, actually, because you can think of biology as a generative algorithm that produces your soma that's actually only about 18 megabytes if you compress it. Aging is everything, all the byproducts, all the stuff that it wasn't coded for, that wasn't selected for. It's actually potentially more complicated than biology.
And funders just aren't interested in problems that are that difficult to be honest for the most part. There are a few exceptions out there and we appreciate you greatly but for the most part people want something with a faster ROI. Yeah, and so because of that the amount of funding is trivial.
Failures of Reductionist Models
So another problem is that people keep trying to come up with reductionist models of aging. There's been, you know, telomeres, antioxidants, the information theory of aging, there's a long list of them and they always fail. I've been watching them fail one at a time and there's a really good paper that anybody who believes in reductionist theories of aging should read which is called "No Privileged Level of Causality." Everything in your body is affecting everything else, every molecule in your cell is affecting every other molecule, it's a mess. Yeah, and also a lot of wishful thinking.
So we actually have people like to say that we've extended lifespan in model organisms. In vertebrates perhaps that's true but that's mostly due to effects that wouldn't have any use in mammals like diapause, things like that. And we only actually have one thing that extends the lifespan in mice so far if you use the ITP 900-day rule where you look at different strains. So yeah, a lot of bottlenecks. We have a paper on this. You can just find it on the Longevity Biotech Fellowship website. And we have a roadmap.
LBF Focus Areas
So we focus not on pharmaceuticals. We focus on biostasis, replacement, and advanced bioengineering. I don't have time to get into those right now. But if you want to go super deep into them, come to an LBF cohort, and you'll spend... I think the next one is going to be a five-day long one. And you're going to spend five days with people who are experts in this. You're going to be brainstorming, problem-solving. It's not passive information consumption. We're going to put you to work.
Role of AI in Acceleration
What about AI? I don't put AI on the roadmap because it's just a tool to make everything go faster. Both cryo, replacement, and bioengineering we can make all those go faster if we use AI in our workflows. And I can see someone in particular that likes this slide. Retro is one of my favorite companies in the space.
Yeah, biostasis, everybody here probably knows about that. You can make improvements to biostasis through modeling. For instance, if you're going to try to predict better cryoprotectants or mixes of cryoprotectants or do actual dry lab experiments where you're trying to lower the search space.
In replacement, we're going to need connectomics. If you're going to do gradual brain replacement, you need to be able to study while you're developing that process whether your tissue grafts are forming the correct connections. And so connectomics requires AI to analyze the data. That's just one example of many.
Same with bioengineering, that's more obvious. Of course, we need to be able to collect a lot of data, model aging, comparative biology.
Data Bottlenecks and Recent Developments
I want to highlight just some cool news. There's this thing called the Thalion Initiative started by Todd White, which supposedly just raised $710 million to do a giant comparative biology study. I'd like to verify that, but if that's true, that's really cool. Because the bottleneck is data. How are we going to generate enough data to actually solve aging?
Who here knows how much the Protein Data Bank costs? Close your eyes. This is basically the training set for AlphaFold. Most people, when we interview people for the LBF, we ask them how much they think the training data set for AlphaFold cost, and they usually guess somewhere around five to 20 million. It was billions over like 40 years. So and that just for one super narrow problem within biology. The amount of data we're going to need to actually simulate cells and solve aging is so much more than that. Who's going to pay for that? How are we going to generate it? I think that's something we should all be thinking about today.
Yeah, because we need to collect a lot of data, do fully automated closed-loop genetic perturbation experiments in order to build genetic CAD tools, digital twins, a human intervention testing program. That's all the stuff that we need to do for bioengineering.
I'm just going to say that we also need to solve the delivery problem. If anyone doesn't know what the delivery problem is, you should really look it up. It's a major bottleneck. Everybody working in reprogramming is not going to be able to actually do anything in humans unless we solve the delivery problem. And so far everybody I've talked to, including Retro, says, I hope somebody else solves that problem.
LBF Mission and Call to Action
So why does the LBF exist? We want to help catalyze the creation of the entities that could solve these problems. The Thalion Initiative is a good example. Retro is a good example. The Wyss Institute is a good example. But we need dozens of these. We need an order of magnitude or two more of these if we're going to have a chance at solving this in our lifetime.
So if you want to join the LBF and dive into the deep, dark science, we'll teach you what are the different things that need to be built and we'll hopefully help you get to work on them or invest in them or find a co-founder. We are generally just like a network to make all the connections necessary to move forward fast, and we're focused on the technicals. We're not talking about the social stuff. That'll be for the next speaker.
So that's what I got. Oh yeah, and we have cool swag for sale over there. This is really hot.
Intuition
- Aging as Emergent Complexity: Biology is a compact (~18 MB compressed) generative algorithm for soma; aging is the vast, unselected byproducts, more complex than core biology, defying reductionism ("No Privileged Level of Causality" paper emphasizes holistic molecular interdependence).
- Paradigm Shift: Biology lags other sciences in quantification, data-driven modeling; pharma inertia unsuited for aging's multi-scale difficulty.
- Translational Bottlenecks: Model organism extensions (e.g., diapause) non-applicable to mammals; only one compound passes ITP's rigorous 900-day multi-strain mouse survival criteria.
- Non-Pharma Roadmap: Prioritize biostasis (cryoprotectant modeling), replacement (AI-analyzed connectomics for tissue graph integrity), bioengineering (automated genetic perturbations → CAD tools, digital twins).
- AI as Accelerator: Tool for dry-lab search space reduction, data analysis, not standalone; integrates across workflows.
- Data as Ultimate Bottleneck: Protein Data Bank (~$billions/40 years) is trivial vs. cell/aging simulation needs; requires closed-loop automation, massive comparative studies (e.g., Thalion).
- Delivery Problem: Critical unsolved hurdle for reprogramming/bioengineering in humans.
- Scaling Solution: Catalyze 10-100x more entities (Retro, Wyss, Thalion) via networks/cohorts for rapid technical progress.
Transcription errors?
- "zone tube training": Interpreted as "Zone 2 training" (common in longevity/fitness contexts for aerobic exercise); likely speech-to-text error for "Zone 2".
- "built done using models": Corrected to "built using models".
- "Acetus 900 day rule": Best guess "ITP 900-day rule" (Interventions Testing Program's multi-strain mouse median survival criterion of ~900 days for rigor); "Acetus" probable mishearing of "ITP" or similar acronym.
- "cryo": Expanded to "cryo" in context of biostasis (cryonics/cryopreservation).
- "Thalion Initiative": Retained as-is; possible "Thalian" or "Thalamus", but Todd White association suggests specific new entity—unverified.
- "Weiss Institute": Corrected to "Wyss Institute" (Harvard's bioengineering/longevity hub); common homophone error.
- "Longbio Fellowship": Expanded to "Longevity Biotech Fellowship" based on LBF context and website reference.
- "soma": Retained as biological term (body cells vs. germline).
- Minor phrasing (e.g., "But it's not designed" repeated "But" likely stutter; smoothed for readability while word-for-word).
- Overall: Dense technical speech with enthusiasm/pauses; no major ambiguities beyond proper nouns/acronyms; funding figures (Protein Data Bank "billion over 40 years") approximated as "billions" per public knowledge.