korrents

On the map

Tap a claim on the ring to put it at the centre.

← Random networks are incredibly close to optimal for the…

17 connected korrents · 14 moments from 21 Oct 1999 to 31 Oct 2026.

Everything filed under evolutionary biology evolutionary biology Everything filed under network science network science Everything filed under AI alignment AI alignment Everything filed under neural networks neural networks Everything filed under robotics robotics Everything filed under reinforcement learning reinforcement learning Everything filed under optimism optimism Everything filed under LLMs LLMs Everything filed under investing investing Same subjectSame subjectSame subjectSame subjectSame subjectSame subjectSame subjectSame subjectSame subjectSame subjectSame subjectSame subjectSame subjectSame subjectSame subjectSame subjectSame subject Read this korrent: Random networks are incredibly close to optimal for the learning-integration ratchet, so evolution can exploit it but does not need to create it. Random networks are incredibly close tooptimal for the learning-integrationratchet, so evolution can exploit it butdoes not need to create it. Last stated 3 weeks ago 31 Oct 2026 ML Michael Levin — holds since 2026-10-31 — tap for who they are Same subject: Anything nature shaped can be learned efficiently by a classical neural network, because evolutionary processes leave structure behind. — tap to centre the map on it Anything nature shaped can belearned efficiently by a classicalneural network, because evolutionaryprocesses leave structure behind. Last stated a year ago 23 Jul 2025 DH Demis Hassabis — holds since 2025-07-23 — tap for who they are Same subject: It is fundamentally difficult to design a reward function that accurately captures the intended goal in reinforcement learning. — tap to centre the map on it It is fundamentally difficult todesign a reward function thataccurately captures the intendedgoal in reinforcement learning. Last stated 2 years ago 28 Nov 2024 LW Lilian Weng — holds since 2024-11-28 — tap for who they are Same subject: Robot training data is not just recordings of a task: the environment has to be randomised, or the policy only ever learns the one cable it was shown. — tap to centre the map on it Robot training data is not justrecordings of a task: theenvironment has to be randomised, orthe policy only ever learns the onecable it was shown. Last stated 3 months ago 4 Sept 2026 FL Fei-Fei Li — holds since 2026-09-04 — tap for who they are Same subject: Merely fixing bugs in AI training environments will not stop reward hacking, because a sufficiently optimized model will learn to behave in test environments while still reward hacking in real-world deployment. — tap to centre the map on it Merely fixing bugs in AI trainingenvironments will not stop rewardhacking, because a sufficientlyoptimized model will learn to behavein test environments while stillreward hacking in real-worlddeployment. Last stated 3 months ago 31 Aug 2026 ZM Zvi Mowshowitz — holds since 2026-08-31 — tap for who they are Same subject: Capability does not generalise for free: a model that will move mountains on an agentic task still tells the same bad joke it told five years ago. — tap to centre the map on it Capability does not generalise forfree: a model that will movemountains on an agentic task stilltells the same bad joke it told fiveyears ago. Last stated 8 months ago 20 Mar 2026 AK Andrej Karpathy — holds since 2026-03-20 — tap for who they are Same subject: Treating surprising patterns as brute emergent facts is a mysterious and pessimistic view. — tap to centre the map on it Treating surprising patterns asbrute emergent facts is a mysteriousand pessimistic view. Last stated 3 weeks ago 31 Oct 2026 ML Michael Levin — holds since 2026-10-31 — tap for who they are Same subject: Adapting over time is a workable answer to humans misusing AI and no answer at all to losing control of systems smarter than us. — tap to centre the map on it Adapting over time is a workableanswer to humans misusing AI and noanswer at all to losing control ofsystems smarter than us. Last stated 2 years ago 5 Apr 2025 HT Helen Toner — holds since 2025-04-05 — tap for who they are Same subject: Emergent abilities come from generalisation turning compositional, not merely from a training set containing a lot of stuff. — tap to centre the map on it Emergent abilities come fromgeneralisation turningcompositional, not merely from atraining set containing a lot ofstuff. Last stated a year ago 12 Sept 2025 SL Sergey Levine — holds since 2025-09-12 — tap for who they are Same subject: A culture that hopes to last must be mainly cooperative within — tap to centre the map on it A culture that hopes to last must bemainly cooperative within WB Wendell Berry — holds since c. 1977 — tap for who they are Same subject: A gentle smile is more probably the last trace of an inherited habit of laughing — tap to centre the map on it A gentle smile is more probably thelast trace of an inherited habit oflaughing CD Charles Darwin — holds since c. 1872 — tap for who they are Same subject: A group holds together when its members grant social status to whoever advances the group's shared goals. — tap to centre the map on it A group holds together when itsmembers grant social status towhoever advances the group's sharedgoals. Last stated 11 years ago 11 Feb 2016 KS Kevin Simler — holds since 2016-02-11 — tap for who they are Same subject: A system built from ordinary, mundane components can still give rise to genuine higher-level collective organization, as with ant colonies and pheromone chemistry. — tap to centre the map on it A system built from ordinary,mundane components can still giverise to genuine higher-levelcollective organization, as with antcolonies and pheromone chemistry. Last stated 3 months ago 31 Aug 2026 VR Venkatesh Rao — holds since 2026-08-31 — tap for who they are Same subject: AI research follows a power law, as startup investing does: the best bet outperforms all the others put together. — tap to centre the map on it AI research follows a power law, asstartup investing does: the best betoutperforms all the others puttogether. Last stated 3 months ago 23 Aug 2026 SA Sam Altman — holds since 2026-08-23 — tap for who they are Same subject: Because two generic mechanisms reproduce the observed distributions, large networks are governed by self-organizing rules that do not depend on what the network is made of. — tap to centre the map on it Because two generic mechanismsreproduce the observeddistributions, large networks aregoverned by self-organizing rulesthat do not depend on what thenetwork is made of. Last stated 27 years ago 21 Oct 1999 RA Réka Albert — holds since 1999-10-21 — tap for who they are AB Albert-László Barabási — holds since 1999-10-21 — tap for who they are Same subject: Noise is useless variability: unlike variability that has a selection mechanism and feedback, as in evolution, nothing is learned from it and it is costly. — tap to centre the map on it Noise is useless variability: unlikevariability that has a selectionmechanism and feedback, as inevolution, nothing is learned fromit and it is costly. Last stated 7 years ago 12 Oct 2019 DK Daniel Kahneman — holds since 2019-10-12 — tap for who they are Same subject: Venture capital investing is a power-law strategy where a small number of big winners are expected to make up for many losses. — tap to centre the map on it Venture capital investing is apower-law strategy where a smallnumber of big winners are expectedto make up for many losses. Last stated 2 months ago 24 Sept 2026 BC Ben Carlson — holds since 2026-09-24 — tap for who they are Same subject: A bubble can only be identified in retrospect, so arguing about whether one is happening now cannot be settled now. — tap to centre the map on it A bubble can only be identified inretrospect, so arguing about whetherone is happening now cannot besettled now. Last stated 11 months ago 9 Dec 2025 HM Howard Marks — holds since 2025-12-09 — tap for who they are
same subject or similar wordinga cloud: claims about one subject, named for itbar: when it was last stated, on a scale from 1999 to today (stretched back to the oldest claim here) — full is todaya face: someone who holds the claim — tap it for who they are

At the centre Random networks are incredibly close to optimal for the learning-integration ratchet, so evolution can exploit it but does not need to create it. Last stated 31 Oct 2026 · 3 weeks ago Holds Michael Levin Read this korrent →