nearbuy

21 ms·

301.
by nearbuy 3y ago
There's a lot of concern about instrumental convergence ( https://en.wikipedia.org/wiki/Instrumental_convergence ). It's the idea that an AI that isn't rogue or evil might still kill us. It could kill us
302.
by nearbuy 3y ago
Concerns about bias and racism in ChatGPT would feel more valid if ChatGPT were even one tenth as bias as anything else in life. Twitter, Facebook, the media, friends and family, etc. are all more bias and radicalized (though I mean "r
303.
by nearbuy 3y ago
This is unfalsifiable speculation though. I could make that claim about anyone (eg, I could say actually it was H. G. Wells planted the seeds of techno-optimism in the present day AI community) and you would have no way to prove me wrong. R
304.
by nearbuy 3y ago
This makes it sound like he called for terrorist attacks. He said something along the lines that countries should make an agreement to stop AGI research, and if a rogue nation starts developing advanced AGI anyway, other countries should be
305.
by nearbuy 3y ago
I think you got that backwards. There's a quote here[1] (sourced from court documents) from the director of admissions at Stanford that he was admitted: > Dear Elon, > As per special request from my colleagues in the School of En
306.
by nearbuy 3y ago
No, this is the proper way to handle accusations from someone who may not be credible. You investigate privately, and only report after, if the accusations seem credible. You don't blast across the media that so-and-so is accused of se
307.
by nearbuy 3y ago
The term "temperature" has been used in machine learning for a long time and came from using it as a parameter during training, analogous to physical temperature in https://en.wikipedia.org/wiki/Boltzmann_dist
308.
by nearbuy 3y ago
An LLM absolutely doesn't respond the same way each time if asked the same question three times in a row, with temperature (randomness) set to zero. It responds the same way only if you start a new chat, which is a clean instance with
309.
by nearbuy 3y ago
> If such a coalition grows large enough It won't. People have never defeated a useful new technology that destroys jobs. People widely like using these tools. You'd need to ban their use worldwide. If the US bans AI, China and
310.
by nearbuy 3y ago
We don't have to be happy about it, but we can't stop this new technology any more than we could stop the invention of the steam engine or the printing press. Technology always displaces jobs; that's largely the point of inve
311.
by nearbuy 3y ago
It is a tricky question, especially if you don't remember the precise definition of combustible/flammable. I believe only the fuel (the thing getting oxidized) in the reaction is considered combustible/flammable, not the oxid
312.
by nearbuy 3y ago
I think the stat was that more than half of public school graduates in California couldn't read at a grade 11 level . The parent comment to yours seems to be saying these kids literally can't read, and don't even know the al
313.
by nearbuy 3y ago
They seem to be fudging it a bit. In the first example with the circle, anything better than 45%-55% (as measured in my graphics editor) is reported as a perfect 50-50 split.
314.
by nearbuy 3y ago
They do appear to be fudging it. At least for first practice example with the circle, it appears to show a perfect 50-50 split when you're up to 5% off (in area) in either direction.
315.
by nearbuy 3y ago
This argument can be generalized against anything police or prosecutors use to convict people. Fingerprints can falsely implicate someone in a crime. Witnesses are notoriously unreliable. If we allow prosecutors to present fingerprints or w
316.
by nearbuy 3y ago
Yeah, I'd say expectancy is unitless, value is in something like "utils", and expectancy * value is the expected value. I would put impulsiveness in something like amount of "meh" per second (e.g. for each second yo
317.
by nearbuy 3y ago
They have about a 38% market share of e-commerce in the US, far ahead of any competitors, but also far from 90%. https://www.statista.com/statistics/274255/market-share-of-t...
318.
by nearbuy 3y ago
If you think Unity is slow to load, wait until you try Unreal...
319.
by nearbuy 3y ago
You're right that if you're replicating an existing copyright image, the process doesn't matter. Legally, if you lived in a cave your whole life and never saw any art and by amazing coincidence you just happened to paint and
320.
by nearbuy 3y ago
> Once there are millions of images, you end up with just tiny blobs of pixels being copied from many different images. This is not how these neural nets work. They don't copy pixels from anywhere. They learn features. The features
321.
by nearbuy 3y ago
> well-resourced students are more likely to get accepted into, and graduate from, elite schools. Happy to back that claim up with some research if you like. We agree on this, but it's also a major confounder for observational word
322.
by nearbuy 3y ago
What Ramanujan's potential could have been is a different question than whether Einstein's equals are dying undiscovered in cotton fields. By definition, we know of zero examples of geniuses who were never discovered due to povert
323.
by nearbuy 3y ago
If we're trying to help our kids, it would be nice to know which interventions actually affect adult intelligence. The word gap research isn't a randomized trial. It mostly shows that children from poorer families in the US hear f
324.
by nearbuy 3y ago
Srinivasa Ramanujan would be a counterexample. He grew up poor with no formal training in mathematics but was nonetheless one of the world's greatest math geniuses.
325.
by nearbuy 3y ago
Without randomized controlled trials, there's no easy way to separate the effects of increased educational spending and tutoring from hundreds of other factors that might affect someone's admission rate. Clearly we expect it helps
326.
by nearbuy 3y ago
In Australia, similar laws were implemented and Google and Meta eventually made deals to pay a number of media companies for links. I assume that's the outcome the government is hoping for. Not sure what will happen this time.
327.
by nearbuy 3y ago
Without getting nitpicky to the point where it's impossible to say anything at all without complaining it's not a perfectly complete and true statement, their statement still seems correct to me. Especially given that this is a st
328.
by nearbuy 3y ago
You misread me. I said that no one said a p-value is the chance the result is erroneous. They said nothing about errors whatsoever in the quoted statement. Apologies for the double negative in my last comment.
329.
by nearbuy 3y ago
They could be p-hacking, though getting 6 sigmas through p-hacking without outright fraudulent data is actually some pretty impressive level of p-hacking. I'd say it's more likely they either made an error or there really is a cor
330.
by nearbuy 3y ago
Unless I'm missing something, that statement appears to be correct. They don't say their result isn't erroneous. They say there is statistically only a very tiny chance that the correlation is due to chance. If the correlatio

More ›