t_mann

24 ms·

301.
by t_mann 3y ago
Not clear to me from when the chat statement is, but given that the presentation is from 2020, it would actually be funny if he went from 'this is going to get us 100% certainty' to 'well, we might never be fully sure, but at
302.
by t_mann 3y ago
Well, it appears that Kevin Buzzard hasn't fully made up his mind on this point (from the presentation by him that I linked): > If my work in pure mathematics is neither useful nor 100 percent guaranteed to be correct, it is surely
303.
by t_mann 3y ago
Mathematicians might not like the insinuation, but the best reason to formalize mathematics is indeed to check its correctness: https://www.andrew.cmu.edu/user/avigad/meetings/fomm2020/sli... tl;dr: cont
304.
by t_mann 3y ago
It arguably matters a lot less in this context - you don't need experience to see when Copilot has failed at Lean, the compiler will tell you anyway. But I'm not saying that understanding it yourself won't matter. In fact, I
305.
by t_mann 3y ago
Not sure what you mean by that - what I meant by formalization was to really deduce every statement in a proof from basic axioms and rules. So basically what Prof Tao did in this blog post (except that the hard rote labor has all been abstr
306.
by t_mann 3y ago
Exactly. There are multiple ongoing projects to formalize existing math, there seems to be some convergence on Lean4 at the moment, but the best analogy is probably programming languages, where there's a constant battle between the old
307.
by t_mann 3y ago
> I don't follow how the proof text is generated from lean? Or is it not, they are doing the verification in lean and the proof text is completely separate. My understanding was that the Blueprint tool that he linked is what's
308.
by t_mann 3y ago
You don't really need understanding if you ask (or train, or fine-tune) the AI to produce a formalized answer, and then write a loop that feeds back the compiler output until it's correct. That doesn't guarantee that it'
309.
by t_mann 3y ago
I find it almost scary how casual Terry is about all this. Comparing this to my own humble attempts at graduate math, I feel like a caveman trying to rock a basic wheelbarrow who just gets a visit from a guy in a helicopter. And then that g
310.
by t_mann 3y ago
Not sure I agree, particularly with the 'I don't want to hear about the story' part. I have a bit of a general chagrin with the 'no spoilers' obsession. Yeah, there may be works like Game of Thrones where the unpred
311.
by t_mann 3y ago
D-K claim a non-monotonic relationship, which simply isn't supported by that data, as you yourself point out: people rank themselves correctly (ordinally). I didn't mean to say that all self-assessments are the same, if that was t
312.
by t_mann 3y ago
I agree, the statistical analysis in the original post makes me very uneasy. I think it could be a case where the conclusion is correct, even though argument isn't necessarily. And yes, the fact that the slope is less than one is fairl
313.
by t_mann 3y ago
I was surprised by the figure from the original article, imho that's the strongest rebuttal: perceived ability grows strictly mononotonically with actual ability, no sign of the famous non-monotonic U-curve. Yeah, the slope is less tha
314.
by t_mann 3y ago
A bit surprised, and maybe even disappointed, that he didn't start his own company. Rarely in history did someone have a better shot at becoming a billionaire of their own making. But you have to give it to Nadella, that's quite t
315.
by t_mann 3y ago
Exactly. I think it would actually be very exciting if OpenAI uses this moment to pivot back to the "Open"/non-profit mission, and Altman and Brockman concurrently start something new and try to build the Apple/Amazon of
316.
by t_mann 3y ago
Current language models fail in all sorts of quantifiable ways, but I think that trying to discuss their merits away by reasoning about what it means to 'truly understand' something, or to be 'truly intelligent' is a com
317.
by t_mann 3y ago
> So it doesn't really ever make sense to compare net worth, market cap etc to GDP. another nitpick, stock-to-flow comparisons can indeed make sense, eg "return on X" measures. they are also fairly common in the context of
318.
by t_mann 3y ago
That's also how I would read it, and I think it's perfectly understandable why someone who was considered a Jew by the Nazis and had to flee from Germany, despite holding a prestigious university chair position and being a scienti
319.
by t_mann 3y ago
If you condense every text passage into a one-dimensional sentiment score, then all texts are going to look like graphs of rising and falling sentiments to you. Nothing unexpected there. I actually liked the charts, but I didn't see mu
320.
by t_mann 4y ago
The onus is generally on whoever claims the presence of an effect, harmful or otherwise. The null hypothesis is always that there is no effect.
321.
by t_mann 4y ago
Evidence means scientific studies, not 'people talking'. A hundred years ago people would have told you that corporal punishment is good for kids, whereas studies now clearly show the opposite. [0] [0] https://www.ncbi.
322.
by t_mann 4y ago
Let me check if I'm getting this right - it sounds like the more fundamental way to address this wouldn't be to make a new Tumblr/OnlyFans/..., but to make a financial services provider that doesn't discriminate aga
323.
by t_mann 4y ago
Is there any evidence that seeing topless women is bad for kids in some way, or is that just a US cultural norm?
324.
by t_mann 4y ago
right - just removing .txt at the end should be enough even
325.
by t_mann 4y ago
Yeah you're right, I just tested it on my site. Before I just tried it in Codepen, but I forgot that it autowraps everything inside <html> and <body>, haha
326.
by t_mann 4y ago
ok, but wouldn't the tags still be resolved by browsers, even if they're not wrapped inside <html>? or is the point that it's not meant to be viewed through a browser? the overhead of the tags in terms of bytes seems fa
327.
by t_mann 4y ago
Are anchor tags frowned upon in the small web movement? It feels like making it harder to follow hyperlinks goes against the "spirit" of what the web was built for.
328.
by t_mann 4y ago
The key point (in the end) seems to be that the odds for a streak like Niemann had are about 1 in 80k. Statistically speaking, I'd say that's a long shot from a smoking gun. Here's a good rundown of a case where a cheater is
329.
by t_mann 4y ago
" I had the impression that he wasn't tense or even fully concentrating on the game in critical positions, while outplaying me as black in a way I think only a handful of players can do. " That's a mere gut feeling, how
330.
by t_mann 4y ago
Frankly, isn't that just what you signed up for as a mobile developer? The WorldAnimals app would probably have worked perfectly fine as a web app, but then the author would have had to figure out payments and discoverability himself.

More ›