lukeplato

I think an unrestricted model with sufficient inference time compute could create unprecedented catastrophes.

Not sure why you would be opposed to whistleblower protections

hopefully nuclear power plants

Michael Edward Johnson has an interesting theory called vasocomputation with the core hypothesis:

vasomuscular tension stabilizes local neural patterns. A sustained thought is a pattern of vascular clenching that reduces dynamic range in nearby neurons. The thought (congealed pattern) persists until the muscle relaxes.

https://opentheory.net/2023/07/principles-of-vasocomputation...

https://x.com/johnsonmxe/status/1863603206649208983

I don't see why a mixture of experts couldn't be distilled into a single model and unified latent space

reading this thread was a good reminder that being intelligent but closed off to alternative hypotheses is the same thing as being ignorant

I remember reading that for k-d trees to be able to split on k dimensions the dataset needs to be > 2^k, which becomes unwieldy pretty quickly

This could make for an interesting UI for exploring clusters in data. I only wish K-d trees could handle higher dimensions

interatomic potential is a promising scale and there's been some interesting recent developments using equivariant graph neural networks: https://www.nature.com/articles/s41467-022-29939-5 https://arxiv.org/pdf/2206.07697

When is a good time to buy AAPL on the dip?

you should use a throwaway account without your name so as to not impact your company

The story distinctly says the sea monster uses a pouch for small kids, so they wouldn't have to fear it as they get older if that part is explained

https://twitter.com/pfau/status/1674766269113937920

models trained on gpt output might be more distilled and specialized but it wouldn't be improving generalization

My impression of Ilya is that it would be far more likely to be a safety related issue than a business/profits related issue

Would be cool if this shared the user found specs to create a database of API specs for the web

Not sure if it's clickbait, but there is good reason to believe in a deep connection between 4d gauge field theories (i.e. of the standard model) and number theory through representation theory. This involves other work than mentioned in this article though it's still inscrutable for most at this point (motivic cohomology/Hilbert–Pólya in RH, supersymmetry in YM mass gap, cosmic galois in renormalization).

From Peter Woit's blog:

It’s worth noting that while there are many connections to the ideas originating with Langlands, this new work shows that the “Langlands program” has expanded into a striking vision relating different areas of mathematics, with a strong connection to deep ideas about quantization and quantum field theory. The way in which these ideas bring together number theory and quantum field theory provide new evidence for the deep unity of fundamental ideas about mathematics and physics.

https://www.math.columbia.edu/~woit/wordpress/?p=13578

it's not rolled out yet

I recommend watching the Hinton talk on dark knowledge on YouTube

I imagine this is for training a multimodal transformer to retrieve previously viewed content and personalize generative conversational bots

optimal programs are very inhuman

sounds revolutionary

check out metarationality: https://metarationality.com/stem-fluidity-bridge

Is there any potential improvements over transformers for interpretablity or alignment?

It's important to avoid objectification in our relations with the world, especially with knowledge.

Touches on being vs having existential modes, transjectivity and also I & Thou.

They should really consider changing their company name at this point

https://en.wikipedia.org/wiki/With_Folded_Hands

This take is a little outdated. Many people actively working on this technology no longer view it this simplistically. ex: https://twitter.com/bio_bootloader/status/164051244495839641...

Is it even possible for our generation, let alone past or future generations, to choose ignorance? Can ignorance only be chosen before a discovery occurs?

AGI goes hand-in-hand with stewardship and utopic visions but I certainly don't see our generation being the ideal candidate.

Plato's allegories are very similar to Buddhist metaphysics. I recommend Vervaeke's meaning crisis series which draws connections between Buddhism, ancient Greek philosophy, Gnosticism, neoplatonism, etc.

https://www.youtube.com/playlist?list=PLND1JCRq8Vuh3f0P5qjrS...

wish there was a follow up discussion about those ideas instead of a context change. And yes, Altman has tweeted something similar before:

haven’t seen this as a twitter thread, so: what true thing do you believe that few people agree with you on?

absolute equivalence of brahman and atman

https://twitter.com/sama/status/1607482407770554373

they are specifically pointing out that the process of RLHF, which is intended to add guard rails on the chat bots trajectory through an all encompassing latent space of internet data, has an unintentional side-effect of creating a highly characterized alter-ego that can more easily be summoned.

The theory is well-thought-out and necessarily rich. The psychological approach of analysis from the alignment crowd is much overdue.