gwern

22 ms·

361.
by gwern 2y ago
Since these sorts of things usually only get more and more complex over time, I would guess that it's all still true, but much more so.
362.
by gwern 2y ago
They obviously did not write OP and 'just' machine translate it, because no one writes like that (and if they were being honest, they would have disclosed that upfront). A LLM came up with most of that... and maybe all of that..
363.
by gwern 2y ago
Bingo. OP is obviously LLM-written and is nothing but clickbait. (Did anything but that tweet even happen? Is any of it true?) And HN is falling for it hook, line, and sinker exactly as calculated. Look at the author's previous blog po
364.
by gwern 2y ago
> Apple should not be the arbiter...I would strongly oppose Apple unilaterally deciding which apps can-, and which cannot-, use the geofencing feature... Apple to restrict certain apps...but for them to do that on their own is completely
365.
by gwern 2y ago
> This is really the fault of the insurance company (they geo-locked the app, Apple is just respecting that). This sounds like a verbose way of saying "this is really the fault of Apple".
366.
by gwern 2y ago
> Most technologies do not follow Wright’s Law – the prices of bicycles, fridges, or coal power plants do not decline exponentially as we produce more of them. I would bet Roser is wrong here and he can't provide a reference for his
367.
by gwern 2y ago
Who will be the first AI-first company to close the loop with autonomous coding bots, and create self-fulfilling prophecies where a customer support bot hallucinates a policy like OP, it gets stored in the Q&A/documentation retriev
368.
by gwern 2y ago
> For the first time, Liquid uncovers a scaling law that performance drop unavoidably brought by the unified training of visual and language tasks diminishes as the model size increases...No prior work has explored whether LLMs retain th
369.
by gwern 2y ago
It was not 'strictly philosophical' in the way some things like the Chinese Room argument are; in Chinese Room, it's stipulated that the Room is pragmatically capable of responding like a native Chinese speaker (it is somehow
370.
by gwern 2y ago
https://www.reddit.com/r/MachineLearning/comments/1jsft3c/r_... I'm still not quite sure how to think of this. Maybe as being like unrolling a diffusion model, the equivalent of BPTT for RNNs?
371.
by gwern 2y ago
That's not 'little known'. That's the only thing almost anyone knows about hat-making.
372.
by gwern 2y ago
Well then - the cost of hiring a LLM compared to hiring a human is irrelevant if you are going to deny that hiring in general is irrelevant, now isn't it? So either OP is making an idiotic comparison because he is wrong and using a L
373.
by gwern 2y ago
> Forty-six percent of the global population lives on less than $5 per day. In some countries, more than 90% of the population lives on less than $5 per day. If agentic AI code generation becomes the most effective way to write high-qual
374.
by gwern 2y ago
One of the frustrating things about the manicule is what I think is the reason you don't see it used online, and it is "everywhere" only in physical media: cluttered pages/GUIs and low-DPI screens mean that almost all of
375.
by gwern 2y ago
Google Docs PDF mirrored at https://gwern.net/doc/anime/2020-hoover-theevolutionofjimdav...
376.
by gwern 2y ago
Could it be date of the original / date the edited version was posted?
377.
by gwern 2y ago
One thing to note here is that the closure is a costly, credible, verifiable signal of commitment to their principles. Unusual or questionable principles, perhaps - few other Christians feel the need to close on Sunday - but principles neve
378.
by gwern 2y ago
In the time it takes you to do one good vote, the low-taste slopmaxxers can do 10 (or 20...?).
379.
by gwern 2y ago
> That came out in 2009, correct? I wonder how much was spent on LLMs up to that point. Quite a lot. Look back at the size of the teams working on language models at IBM, Microsoft, Google, etc, and think about all the decades of languag
380.
by gwern 2y ago
The WWII POW camp paper discussed is a lot more interesting and worth reading than this article winds up being: https://gwern.net/doc/economics/1945-radford.pdf
381.
by gwern 2y ago
Language models repeatedly delivered practical, real-world economic value at every step of the way from at least n-grams on. (Remember the original 'unreasonable effectiveness of data'?) The applications were humble and weren&#x27
382.
by gwern 2y ago
It can be easily gamed. The users are self-selected, and they have zero incentive to be honest or rigorous or provide good responses. Some have incentives the opposite way. (There was a report of a prediction market user who said they had w
383.
by gwern 2y ago
> Interestingly, for models up to 1 billion parameters, the loss actually starts to increase again after reaching a minimum. This might be because as training continues, the model develops more complex, non-linear representations that ou
384.
by gwern 2y ago
[0]: ...they're having a "beat the tariff sale". 5% off everything.
385.
by gwern 2y ago
Preprint: https://www.celiamoore.com/uploads/9/3/2/1/9321973/wakeman_y... A useful concept here is the 'incompleteness of contracts' ( https://en.wikipedia.org/wiki&#x2
386.
by gwern 2y ago
> the site design was and how clearly communicated the problem they were trying to solve I don't agree with either. Even after I enabled JS (no warning) and then after reading the whole page, finally realized that the implementation
387.
by gwern 2y ago
That's unfortunate about Mr Zadunaisky, but it does suggest there's no hope of a recording unless someone should stumble across it in the Borges papers (although given how his estate has been abused, little hope of that being usef
388.
by gwern 2y ago
I see no hint of an audio recording having been made, much less surviving & digitized, in the original article's description: https://www.cambridge.org/core/journals/journal-of-the-histo... It's too
389.
by gwern 2y ago
Yes, this has been done in many forms with other algorithms. You score each generation with a model like CLIP, for example, and then you can evolve 'Mona Lisa made of triangles', say. A constraint like 'exactly 13 circles&#x2
390.
by gwern 2y ago
Suggestion: change the 'missing teeth' form line to include an ordinal count or something. I was surprised to learn when filing my form that I had "Missing teeth: 30"!

More ›