ftxbro

15 ms·

181.
by ftxbro 3y ago
> I left academia over a year ago, partially because I felt the pressures How do we know he isn't bitter about academia and trying to paint it to look worse than it is?
182.
by ftxbro 3y ago
> I was asked today to comment on connections between the Manhattan Project and AI. I found myself surprisingly uncomfortable with the request: my own thinking is still forming, and I'm not yet confident in what choices to espouse.
183.
by ftxbro 3y ago
> the kinds of stones that go into the cheaper one- or two-carat solitaire bridal rings popular in the US have experienced far sharper price drops than the rest of the market
184.
by ftxbro 3y ago
these kinds of articles are usually cringey but this one was pretty good. it doesn't have the usual smug subtext like of course ai isn't an existential threat
185.
by ftxbro 3y ago
i wonder how long before every living person has an ancestor who was a billionaire in 2023
186.
by ftxbro 3y ago
comment by karpathy https://twitter.com/karpathy/status/1697318534555336961
187.
by ftxbro 3y ago
# Basic guardrail involving insults. define flow user express insult bot escalate negative sentiment and reciprocate insult
188.
by ftxbro 3y ago
Take the number of vehicles in the field, A. Multiply it by the probable rate of failure, B. Multiply the result by the average out-of-court settlement, C. A x B x C equals X. If X is less than the cost of a recall, we don't do one.
189.
by ftxbro 3y ago
If they have to ask it questions like this to argue that it can't reason, then to me that looks like some cope and that it can reason just fine it's just kind of dumb from its limited model size, data size, and cpu amount.
190.
by ftxbro 3y ago
> It is something I have been wondering about: why did Meta not keep the training process going on while the loss curves seemed to go down? If I remember correctly, it's because the main reason they trained multiple models was to sh
191.
by ftxbro 3y ago
a triumph of man and machine
192.
by ftxbro 3y ago
how is your paper different from all the ones like 'transformers are really x' where x is the author's special field of study
193.
by ftxbro 3y ago
> It solves an SVM that separates 'good' tokens within each input sequence from 'bad' tokens. This SVM serves as a good-token-selector and is inherently different from the traditional SVM which assigns a 0-1 label to
194.
by ftxbro 3y ago
no, it was from balenciaga in paris https://www.youtube.com/watch?v=rxwut4H6wJM
195.
by ftxbro 3y ago
> It involved men and women aged 60 to 85 without memory impairment. All were given a diffuser and seven cartridges, each containing a single and different natural oil. People in the enriched group received full-strength cartridges. Cont
196.
by ftxbro 3y ago
OK but why they write "benign optimization landscape devoid of stationary points" instead of "convex" other than "just for show"? In my understanding it's not better for either audience experts or laymen.
197.
by ftxbro 3y ago
> To mark the ending of the 52-year cycle, all fires in the Basin of Mexico are extinguished and then lit again from a single source, in what is called the New Fire Ceremony. we still have this tradition today, when social media sites be
198.
by ftxbro 3y ago
> Primarily Discord, pseudonymous, GPU-poor volunteer community somehow coordinating enough to train >10B, OPT/BLOOM-competitive models what are the discords for Eleuther and RWBY are they public?
199.
by ftxbro 3y ago
least capitalist hacker news comment. do the arrows mean flow of capital? i don't think so. i think it means flow of people or cultural influence or leadership
200.
by ftxbro 3y ago
this is gpt or if it's not then it should have been
201.
by ftxbro 3y ago
You got downvoted into the ground but I think it's true. Personally I don't think companies should be allowed to do this. Every technical term used in AI is now also the name of some AI company. Even words like Windows and Stripe
202.
by ftxbro 3y ago
> we show that over-parameterization catalyzes global convergence by ensuring the feasibility of the SVM problem and by guaranteeing a benign optimization landscape devoid of stationary points does this mean 'an over-parameterized t
203.
by ftxbro 3y ago
lol it starts only as far back as 2011
204.
by ftxbro 3y ago
This is obviously the right take on LLM "hallucinations." It's an inherent part of how they were able to acquire the emergent cognitive capabilities, not some weird flaw that needs to be patched.
205.
by ftxbro 3y ago
you can decide to go unemployed for a while if you find your current job is unethical and demoralizing
206.
by ftxbro 3y ago
least fearmongering headline
207.
by ftxbro 3y ago
it's always darkest before the dawn
208.
by ftxbro 3y ago
sometimes you can print academic papers that have pseudocode in them it's like source code pseudocode is basically python anyway
209.
by ftxbro 3y ago
it's from claims from psychics from a military program like 'the men who stare at goats' they don't actually believe it > this highly classified project was about the mind-reading of the enemies, so the CIA started re
210.
by ftxbro 3y ago
haskell

More ›