I did in fact go back and listen to that part, but I interpreted that clarifying expansion as referring to the latter part of your quoted segment only, and the former part of your quoted segment to be separate—using cryptocurrency as a bridging topic to get to cryptography afterwards. Anyway, your interpretation is entirely reasonable as well, and you probably have a much better Eliezer-predictor than I do; it just seemed oddly unconservative to interpolate that much into a transcript proper as part of what was otherwise described as an error correction pass.

We have people in crypto[graphy] who are good at breaking things, and they're the reason why anything is not on fire. Some of them might go into breaking AI systems instead, because that's where you learn anything.

Was there out-of-band clarification that Eliezer meant “cryptography” here (at 01:28:41)? He verbalized “crypto”, and I interpreted it as “cryptocurrency” myself, partly to tie things in with both the overall context of the podcast and the hosts' earlier preemptively-retracted question which was more clearly about cryptocurrency. Certainly I would guess that the first statement there is informally true either way, and there's a lot of overlap. (I don't interpret the “cryptosystem” reference a few sentences later to bias it much, to be clear, due to that overlap.)

If you weren't already aware, look up “circuit bending” if you want to see a grungier DIY/hacker-style subcultural activity around producing interesting sounds via ad-hoc manipulations of electronics.

Tiberium at HN seems to think not. Copied and lightly reformatted, with the 4chan URLs linkified:

It seems that the leak originated from 4chan [1]. Two people in the same thread had access to the weights and verified that their hashes match [2][3] to make sure that the model isn't watermarked. However, the leaker made a mistake of adding the original download script which had his unique download URL to the torrent [4], so Meta can easily find them if they want to.

I haven't looked at the linked content myself yet.

Long before we get to the “LLMs are showing a number of abilities that we don't really understand the origins of” part (which I think is the most likely here), a number of basic patterns in chess show up in the transcript semi-directly depending on the tokenization. The full set of available board coordinates is also countable and on the small side. Enough games and it would be possible to observe that “. N?3” and “. N?5” can come in sequence but the second one has some prerequisites (I'm using the dot here to point out that there's adjacent text cues showing which moves are from which side), that if there's a “0-0” there isn't going to be a second one in the same position later, that the pawn moves “. ?2” and “. ?1” never show up… and so on. You could get a lot of the way toward inferring piece positions by recognizing the alternating move structure and then just taking the last seen coordinates for a piece type, and a layer of approximate-rule-based discrimination would get you a lot further than that.

I wonder whether, when approving applications for the full models for research, they watermark the provided data somehow to be able to detect leaks. Would that be doable by using the low-order bits of the weights or something, for instance?

I mean in the paragraph “If anyone wants to check it out, the app can be found at: https://cyborg-psychologiy.com”. I suppose I should have said “the main pseudo-link”. 🙂 What you gave in this last comment points to the blog post, not (directly) to the app.

The main link might be broken due to an extraneous ‘i’.

Without knowing what implications this might have, I notice that the first two points against “People might neglect real romance” are analogous to arguments against “People won't bother with work if they have a basic income” based on a “scarcity decompensation threshold” model: avoiding getting trapped in a really bad relationship/job by putting a floor on alternatives, and avoiding having so little confidence/money that you can't put in the activation energy to engage with the pool/market to begin with.

I wonder if anyone's tried targeting the avian furry market with these, since those would seem to be the most obvious class of “people who might not mind looking like a duck”. I can't seem to get a good grip on that intersection via search engines, mostly because I get a lot more results related to the non-protective masks more visibly associated with that group.