iiinova — where your digital self keeps evolving

Reposted by Joel Lehman

Kenneth Stanley @kennethstanley.bsky.social · 20/05/2025

Could a major opportunity to improve representation in deep learning be hiding in plain sight? Check out our new position paper: Questioning Representational Optimism in Deep Learning: The Fractured Entangled Representation Hypothesis. Paper: arxiv.org/abs/2505.11581

14710

Reposted by Joel Lehman

Tarin Ziyaee @tarinz.bsky.social · 24/01/2025

"Corner cases". "Long tails". You can play whack-a-mole with them. Or, you can confront their root causes directly. Our new paper "Evolution and The Knightian Blindspot of Machine Learning" argues we should do the latter.

2178

Joel Lehman @joelbot3000.bsky.social · 24/01/2025

3) ML and RL have a rich history of imaginative new formalisms, like @dhadfieldmenell's CIRL, @marcgbellemare's distributional RL, etc. Highlighting this potential blindspot may unleash the field's substantial creativity, either in refuting it, or usefully encompassing it.

140

Joel Lehman @joelbot3000.bsky.social · 24/01/2025

2) Open-endedness: Field that rhymes most w/ unknown unknowns -- it explicitly aims to endlessly generate them. We believe OE algos can simultaneously aim towards robustness to them Related to @jeffclune's AI-GAs, @_rockt, @kenneth0stanley, @err_more, @MichaelD1729, @pyoudeyer

130

Joel Lehman @joelbot3000.bsky.social · 24/01/2025

1) Artificial Life: Relative to its grand aspirations to recreate life's tapestry digitally, ALife is underappreciated. scaling + creativity may uncover novel robust neural architectures See work done by @risi1979 @drmichaellevin @hardmaru @BertChakovsky @sina_lana + many others

130

Joel Lehman @joelbot3000.bsky.social · 24/01/2025

So what to do? The message could seem negative, but we're optimistic there are many possible avenues to dealing w/ unknown unknowns. Some include fields currently more peripheral to ML, like Artificial Life or Open-endedness; others involve imagining new ML formalisms & algos

130

Joel Lehman @joelbot3000.bsky.social · 24/01/2025

Paradigms like meta-learning ("learning how to learn") are exciting and seem like potential solutions. But they still assume a (meta-)frozen world, and need not incentivize to learn how to deal w/ the unknown (paper has more on other paradigms).

130

Joel Lehman @joelbot3000.bsky.social · 24/01/2025

E.g. given 1 additional edge-case example, sometimes more effective to 1) filter many divergent models through it, b/c more reflective of: "face a novel problem 0-shot" then 2) just train on it, which will help generalize to similar situations but not further unknown unknowns

120

Joel Lehman @joelbot3000.bsky.social · 24/01/2025

Rather than rely only on IID-aimed generalization, evolution takes bitter lesson to logical extreme: learns specialized architectures / learning algos that help organisms generalize to unforeseen situations, tested over time by shocks in a constantly-changing world.

130

Joel Lehman @joelbot3000.bsky.social · 24/01/2025

This isn't a dig at LLMs, which are amazing but still interestingly fragile at times. Generalization of big NNs is great, but underlying assumption is train world = test world = static. The paper argues NN generalization does not directly target robustness to open unknown future.

110

Joel Lehman @joelbot3000.bsky.social · 24/01/2025

Contrasting evolution with machine learning helps highlight the blind spot: a "dumb" algo w/ no gradients or formalisms can yet create much more open-world robustness. In hindsight it makes sense: If algo implicitly denies a problem's existence, why would they best solve it?

120

Joel Lehman @joelbot3000.bsky.social · 24/01/2025

Evolution, like science or VC, can be seen as making many diverse bets, that future experiments may invalidate (diversify-and-filter). Organisms able to persist through many unexpected shocks are lindy, i.e. likely to persist through more. D&F can be integrated into ML methods.

120

Joel Lehman @joelbot3000.bsky.social · 24/01/2025

Interestingly, evolution's products = remarkably robust. Invasive species evolve in one habitat, dominate another. Humans zero-shot generalize from US driving to the UK (i.e. w/o any UK data) -- still a big challenge for AI. How does evolution do it, w/o gradients or foresight?

130

Joel Lehman @joelbot3000.bsky.social · 24/01/2025

Most open-world AI (like LLMs) rely on "anticipate-and-train": Collect as much diverse data as possible, in anticipation of everything the model might later encounter. This often works! But training assumes a static, frozen world. This leads to fragility under new situations.

130

Joel Lehman @joelbot3000.bsky.social · 24/01/2025

In short, we 1) highlight a blindspot in ML to unknown unknowns, through contrast with evolution, 2) abstract principles underlying evolution's robustness to UUs, 3) examine RL's formalisms to see what causes the blindspot, and 4) propose research directions to help alleviate it.

150

Joel Lehman @joelbot3000.bsky.social · 24/01/2025

new paper: "Evolution and the Knightian Blindspot of Machine Learning" Our ever-changing world bubbles with surprise and complexity. General AI must include handling unforeseen situations with grace. Yet this issue largely lies outside AI's formalisms: a blind spot. (1/n)

2284

Reposted by Joel Lehman

Eugene Vinitsky 🍒 @eugenevinitsky.bsky.social · 02/01/2025

It’s the new year. Delete slack from your phone. Open your email after lunch. Disconnect the WiFi on Saturdays. Go to the woods on the weekend. Purchase a one way ticket to Alaska. Join a community of bears. Film a documentary. Post it on LinkedIn.

0381

Joel Lehman @joelbot3000.bsky.social · 09/12/2024

Economics papers were a bit different in the 80s? From "Let's Take the Con out of Econometrics" by Edward Leamer, >3k citations

0110

Joel Lehman @joelbot3000.bsky.social · 25/11/2024

"Move then with new desires. / For where we used to build and love / Is no-man's land, and only ghosts can live / Between two fires." -C. Day-Lewis

050

Reposted by Joel Lehman

Max @maxbittker.bsky.social · 10/08/2023

“Friendship Feed" is a feed with the goal of surfacing the people you care about, even if they don’t post a lot! Mixology: Find your mutuals, shuffle them, and show one post per person from their latest few. LMKWYT :) bsky.app/profile/did:plc:wmhp7mubpg…

59940