ben_w

10 ms·

121.
by ben_w 1mo ago
Same way you get "more than 10%" prediction on "I don't know what moves Stockfish will make when I play against it, but I know I will lose". In fact, I will lose in part because I don't know what moves Stockf
122.
by ben_w 1mo ago
I mean much more generally than doom. It's like, Altman's reputation in general is not great, and that includes people who've worked there.
123.
by ben_w 1mo ago
Sure, sure. There's many others like this to help you combine whatever you do feel you can put a number to. For me, that particular question is "probably 0, but with 100% variance". This is because I think most of the things
124.
by ben_w 1mo ago
> The current government is maximally evil and interested in using technology to keep power and crush dissent. You don't need to introduce hypotheticals! No no, I assure you, it's quite capable of getting very much worse.
125.
by ben_w 1mo ago
A flat distribution is still a distribution, and correct exploration would have revealed that the distribution is flat. The agent appears to have gained the false belief that it has learned something and done some exploring, when in fact it
126.
by ben_w 1mo ago
Mine is around 10%. There's lots of moving parts and we all have to input our best-guesses as to how they interact. Some are predictable (e.g. "military will want capabilities, want them able to choose targets"). Others are n
127.
by ben_w 1mo ago
The paperclip maximizer is also at the intentional direction of a human or humans. It's not "AI surprises everyone by having a thing for paperclips", it is "idiot tells AI to maximise paperclips no matter what, and then
128.
by ben_w 1mo ago
I'm old enough to remember when people dismissed all the doom coming from these companies as "marketing". (A thing many of them have been entirely consistent about since GPT-2, or indeed earlier given the founding documents).
129.
by ben_w 1mo ago
Talk is indeed cheap. Thing is, while there are many historical examples of a highly organised bunch of people with a lot of resources who engage in campaigns of violence for various reasons, many such attempts fail disastrously. Furthermor
130.
by ben_w 1mo ago
There's a lot more than one chip fab. A while back Yudkowsky wrote that a ban would only work if was enforced by airstrikes. By a game of telephone, some people read "bomb", but there's a very big difference between &quo
131.
by ben_w 1mo ago
I think you misunderstood my point, just as the other commentor misunderstood one level up. Perhaps a different approach to explain the problem here: "It ain't what they don't know, it's what they know for sure that just
132.
by ben_w 1mo ago
> Apropos of TFA, if one is to believe LLMs can carve a threatening path to humanity end, how could one not see the same for our much older extant threats? Which "much older" extant threats? Global warming and nukes are industr
133.
by ben_w 1mo ago
> You believe in AGI, but not ASI then? I have observed that these initialisms do not have generally agreed upon meanings; they are not useful for discussions with people who have not already stated what precisely they count. There is di
134.
by ben_w 1mo ago
> I don’t think ai models come close to nuclear weapons or to run-of-the-mill, everyday carbon emissions in terms of danger to humanity. Even today, AI is the biggest discrete threat to bringing carbon emissions under control. Almost eve
135.
by ben_w 1mo ago
I think that mainly changes the nature of the threat rather than anything else. Even if we didn't already have any wealth inequality or consolidated power, they're already starting to become capable of creating those things for th
136.
by ben_w 1mo ago
LLMs can already control robots. Today's LLMs are inept at it, but they can do it. Things will be bad, unless someone does something to stop it. Will anyone do that? We don't know. https://xkcd.com/2278/
137.
by ben_w 1mo ago
I suspect you two are discussing different AI. I think you (and I, and everyone who wants AI development to pause while we catch up with the implications at least) is looking where the ball is going. I have the general (non specific to anyo
138.
by ben_w 1mo ago
Yes, but Venus has 93 times the atmospheric pressure of Earth, and it's overwhelmingly CO2. Earth isn't likely to get a true runaway like that until the sun gets another billion or so years on the clock, and when it does it will
139.
by ben_w 1mo ago
While true, the impact from global warming is likely to be less crops rather than no crops.
140.
by ben_w 1mo ago
All human behaviour since no later than the Polynesians turned the Pacific islands into a network of sailing trade routes. "Humans move" is one of the easiest things to predict. The biggest problem today is that they generally mov
141.
by ben_w 1mo ago
None of those are "all"/extinction. Bad, sure; I wish they were not present to threaten us, but they are not close to "all". Humans are extremely resilient omnivores. We are able to eat a huge range of things from a
142.
by ben_w 1mo ago
> Taking away the race/ethnicity layer for a moment, it might be that an LLM develops a predisposition to emit positive terms (like "accept") when the prompt contains "banananow", and negative terms when it conta
143.
by ben_w 1mo ago
I think they're saying that while it doesn't matter, the agent and human "do not actutally know" that it does not matter. Philosophy sometimes says that knowledge is a "justified true belief"*; in this experime
144.
by ben_w 1mo ago
Mm. The job ads I see are a bit like that. Roles that are explicitly AI-driven-dev want you to have experience orchestrating multiple agents at once, almost everyone else has job ads written as if LLMs had not yet been invented. One or two
145.
by ben_w 1mo ago
https://en.wikipedia.org/wiki/International_reactions_to_the... Rasmus Jarlov: If the US attempts a military takeover of Greenland, “then we are at war” - https://www.newstatesman.com/internation
146.
by ben_w 1mo ago
Different kinds of failure. The Russian weapons (at various points) were missing critical components, maintenance, fuel, etc. The US weapons appear to have functioned as specified, but the commander-in-chief appears to know marginally less
147.
by ben_w 1mo ago
We know it's possible to put backdoors into LLMs, we don't have reliable ways to detect them without direct support from whoever inserted it. Europe is less-worse-off with open weights than with… I guess it's weights-as-a-ser
148.
by ben_w 1mo ago
Same. Calling them that stood out positively, like the gull wings and the party mode; unfortunately, the same mind for that then gave us the Cybertruck and the not-legally-a-department DOGE.
149.
by ben_w 1mo ago
I remember at the time it was a refreshing change from all the stuffy corporate nonsense. As with so many, their weaknesses are also what made them stand out initially.
150.
by ben_w 1mo ago
69, surely? This is Musk we're talking about, after all.

More ›