Student in fundamental and applied mathematics, interested in theoretical computer science and AI alignment
Technoprogressive, biocosmist, rationalist, defensive accelerationist, longtermist
Matrice Jacobine
Karma: 1,541
I don’t think it shows that. It arguably suggests that abundant pretraining data doesn’t matter as much as one could think. As opposed to good pretraining data.
I think it straightforwardly show the reverse (good pretraining data doesn’t matter as much as one could think, but abundant pretraining data does)? Olmedo himself notes: “What holds the 1930 model back is that it is severely undertrained (only 260B tokens), rather than its pre-training data.”
I presume that the codebases + agentic coding transcripts that they SFT’d on were high quality, right?
SFT is posttraining??
Matrice Jacobine’s Shortform
This is interesting because, it is not discussed enough that AlphaGo worked the same way LLMs work: it was pretrained on a large dataset of human moves, then posttrained through reinforcement learning. The shift to AlphaZero removed the pretraining, showing that pretraining data didn’t matter at all for the model’s capabilities (quite the contrary, AlphaZero is superior to AlphaGo). While this coding model is still pretrained, it also shows that good pretraining data doesn’t matter nearly as much as one could think.
Substack is designed for news, not for posts people will be reading in five years. Reading a Substack’s archive is a pain, and most Substacks are poorly indexed by Google. Therefore, if you take the intermittent-posting approach, I recommend posting to both a Substack and a separate stand-alone website (for inspiration, consider Andy Masley or dynomight’s pages). Your Substack allows people who are interested in your thoughts to subscribe and be automatically notified when you have something new to say, while the stand-alone website serves as a canonical reference.
(x)
I believe, from memory, that Zajko identified herself as a woman in court, but I cannot access to the relevant Maryland local media rn bc it blocks European countries and the Internet Archive is down rn. For second-hand evidence e.g. here refers to Zajko as “a cis woman involved in the affair”.
This is interesting in that the most serious violence against you came from Zajko. What was Zajko’s motivation here? Did he wanted to protect CFAR/MIRI? How did he came up living with Ziz, when he was that antagonistic against you?
Zajko is a cis woman FTR.
Were you a source for the original Rolling Stone article?
(Wait, considering Vaniver has commented here too, what comment do you mean, and in what direction do you think the difference is?)
Yeah I see. Cretarei say they don’t want to talk about Ziz again after that, but I see how you could have a broader definition of what “writings about the Irish” mean in the analogy (e.g. discussion of psychology).
I assume prosecutors are waiting for less circumstantial evidence to emerge?
A murder trial could take years and years to fully resolve IIUC.
Ziz and Zajko aren’t charged with murder.
I do not think it is true that the Ziz crew “are still committing crimes” or that “their violence and retaliation is not contained”. I am also not quite sure what you are suggesting with the mention of Snyder. I would point to @AprilSR’s comment here (and subsequent subthread). Slimepriestess isn’t banned, Chase isn’t banned.
@Vaniver do you still hold that “Less Wrong the website should accept letters from murderers in asylums”?
Maybe? Sorry for assuming people weren’t disingenuous trolls trying to derail a very serious topic involving several actual completed murders and suicides for their personal political vendettas.
I’m aware. I was just describing their philosophical worldview to contrast with Ziz’s ‘vegan anarcho-transhumanism’. But the whole point of Holly leaving rationalism and EA to start PauseAI was because she wanted AI x-risk concerns to reach people with all kind of worldviews (beyond the ~left-libertarian utilitarian transhumanism of rationalism and EA), and this has been a successful strategy hence everyone from Bernie Sanders to Ben Bernanke to Steve Bannon being into AI x-risk now.
IDK. The concerns in the footnote of attracting crazy people (even against
hertheir wishes), you are aware that I am concerned about as well due to events last year. But I definitely do not want anyone to feel inhibited from writing and socializing if they are well-meaning and it is good for them. Maybe lettingherthem post personal blogposts but not frontpage posts would be a good compromise? But maybe as mentioned find another adjacent space.
I’m going to research the last two attempts to someone’s life that were related to the topic of ai risk. If it turns out the aspiring assassins were mainly influenced by the writings of Ziz, as opposed some other prominent rationalist, then I’ll get back to you.
Seems extremely implausible to me FTR. Looking at incidents of domestic terrorism against AGI companies, Sam Kirchner (if we’re counting him) was a StopAI guy, so left-wing human-centric anti-transhumanist and close to Torres (admittedly Zajko did praise Torres and quoted such anti-transhumanist stuff too though). Daniel Moreno-Gama was a right-wing (NRxish?) human-centric anti-transhumanist (and racist eugenicist). Amanda Tom and Muhamad Tarik Hussein appear to be generic progressives presumably motivated by near-term ethics concerns, not existential risk (unlike the previous two).
Ironically (with a little amount of bad faith) the one ~criminal activity that I can trace to someone influenced by Ziz is that one of the DOGE guys had an account here where he talked about veganism being necessary prefigurative praxis for a sentience-friendly singularity and praised Ziz when pointed to her in the comments.
To clarify: are you just banning
herthem from making classical LW posts, or also banningherthem from having a personal blog (i.e. treatingherthem similarly as the automatically sorted out LLM slop)?[Edit: btw Gwen I’m she/her-ing out of habit but please feel free to clarify your preferred pronouns!]
Like most macrostrategy-brained cult leaders (e.g. Leverage, Ziz, arguably FTX, etc.) in this space, the answer appear to be some variation of “amass as much power and money as possible and take over the world using AI” e.g. this mail where he pitch a super-secret Epstein-and-Thiel-funded AGI project under Arecebo Observatory.
So… what’s the general take on the hantavirus outbreak?