c7b

14 ms·

211.
by c7b 3y ago
Seems like a good collection of standard ML techniques, introduced with a fairly unified mathematical notation and quite a few proofs. Quite the Herculean effort (600 pages!). It just seems to me like they're putting the emphasis on th
212.
by c7b 3y ago
In retrospect, how do you so far value the utility of the data you got? Did you take any actions based on them, do you think you will be doing so in the future?
213.
by c7b 3y ago
I think the point of the study was precisely to see how well the original model without adaptations (my understanding is that they only use prompting, which does not affect weights) can perform. I think that question is arguably even more i
214.
by c7b 3y ago
I'm sure lots of people have been thinking/working along that direction. I think the idea of combining LLMs with formal verification/proof assistant tools (Lean, Coq, Isabelle,...) is particularly interesting. Anyone is aware
215.
by c7b 3y ago
> Either you’re born an artist or you’re not. Either you’re gifted at art or you’re not, which is true of everything else, sports, coding. Either you’re gifted as a coder or not. What a strange claim to be opening your case with. Elitist
216.
by c7b 3y ago
It's correct that he is a professor at UCLA, but it's also worth mentioning that he's regularly called nicknames like 'greatest mathematician alive' (just try googling that phrase): https://academicinflue
217.
by c7b 3y ago
It could fine-tune / train new LLMs to incorporate new knowledge and then, given that it has CLI acces, launch new agent instances using those models. And we're talking about sessions without a defined end here, that's the pr
218.
by c7b 3y ago
While it is a joke, I assume, I think it hints at a crucial problem: it's relatively easy to imagine an Auto-GPT-style agent with some sort of CLI access on an internet-connected machine turning into a paperclip machine, no matter how
219.
by c7b 3y ago
Great engineering. Since the article asks about analytical tips for performance improvement: looks like you should be able to substantially get the complexity down using something like least angle regression (ie what's used by efficien
220.
by c7b 3y ago
> "You are ChatGPT, a large language model trained by OpenAI, based on the GPT-4 architecture. Knowledge cutoff: 2022-01 Current date: 2023-10-11 Image input capabilities: Enabled" That's surprisingly short, also compared
221.
by c7b 3y ago
Pretty sure that pretty much all of their business practices are GDPR violations. So we do have a choice - we could either end privacy, or have legislation that ends a potentially profitable startup.
222.
by c7b 3y ago
> To survive the flood of revenue optimized content..., you need an active [software] agent working on your behalf How long do you think it will be before that software agent's output will be just as riddled with commercialized cont
223.
by c7b 3y ago
There's math to show that the viability of a secret decreases rapidly as the number of people involved grows: https://journals.plos.org/plosone/article?id=10.1371/journal... The NSA couldn't even keep th
224.
by c7b 3y ago
Title: United flight descends 28,000 feet... General question: why is it that for such events, it is usually the airline name, or sometimes the flight number, that is mentioned first, and not the plane model? That would seem more relevant t
225.
by c7b 3y ago
> I have to question to the value of looking over existing papers given the current replication crises, is using an LLM to review existing papers just going to enable us to do more bad science faster? It might be worse than that, not onl
226.
by c7b 3y ago
Not sure the current generation of confidently-wrong language models will be of overall good use in summarizing literature, as suggested in the article. Sounds like the perfect disaster recipe for the "citogenesis" ((C) xkxcd) of
227.
by c7b 3y ago
Generally curious - any particular reasons you chose sqlite and gnuplot for this task rather than, say, Python?
228.
by c7b 3y ago
> Imagine essentially an archival service where all kinds of content (web pages, mp3s, videos, etc) are indexed according to their SHA-256. Isn't that pretty much what IPFS is?
229.
by c7b 3y ago
Probably worth adding: there's also one that tracks Cramer's bets, and as of today, its performance since inception (both March this year) is (minimally) better than the inverse one: https://www.marketwatch.com/inv
230.
by c7b 3y ago
> resistance to voter coercion and vote selling Thanks for mentioning that issue, it's something that I don't hear mentioned enough in online/distance voting debates (maybe it just means that I'm not involved enough,
231.
by c7b 3y ago
What I find interesting about this wave of tech is not just how useful, but how accessible it is. Image generators like SD pretty much work out of the box on a lot of consumer-grade hardware, LLMs might be a bit more of a stretch but still
232.
by c7b 3y ago
It's technically not necessary, but it's also not necessary that the authors give away their samples while they're still running tests (apparently they never even wanted to go public at this stage). The claims will be checked
233.
by c7b 3y ago
Does this count as an instance of 'don't roll your own crypto'?
234.
by c7b 3y ago
> I can't believe that the most important discovery of my lifetime is surrounded by this cloud of uncertainty. I don't get it, what else would you expect? Any claimed scientific breakthrough should generally be met with a healt
235.
by c7b 3y ago
Tbh, I had Design Thinking down as a buzzword and never really engaged with it or even knew what it really was. Then someone recently suggested using DT to revamp the UI in a place where most of us weren't quite happy about the current
236.
by c7b 3y ago
Obligatory xkcd: https://xkcd.com/331/
237.
by c7b 3y ago
Like a lot of things in math, I think Monty Hall becomes easier to see when you look at some extreme cases. 3 doors is the lowest amount of doors where the paradox appears, ie no way to go smaller. So try a much bigger problem. I like this
238.
by c7b 3y ago
You didn't once mention creating better products/services for customers, fostering an enjoyable work environment or even creating value for shareholders. Your whole argument is about (a very narrow Machiavellian view of) individua
239.
by c7b 3y ago
According to the same source, 2022 was the second-most VC deals since 2006, and ca the 4th worst for IPOs according to stockanalysis.com. There's clearly a massive upward trend in VC deals, while IPOs are much more stationary. Eyeballi
240.
by c7b 3y ago
I don't think semver is meant to condition users into blindly trusting updates. I always thought it was meant to condition developers into being mindful about what the changes they make will mean for their users. You can maybe never be

More ›