scarmig

11 ms·

151.
by scarmig 8mo ago
See e.g. Richard Sutton, who, although not at a top lab, is certainly a very important figure in the field. Or, if you want someone with concrete influence at a top lab, Larry Page.
152.
by scarmig 8mo ago
At least in the case of the researchers I mentioned, they have a deeply held, genuine belief that AI will, in the very near term, exceed humans in all intellectual capabilities, and that poses a bigger risk to human existence than humans si
153.
by scarmig 8mo ago
More on the periphery than an insider, but I personally know researchers in all three major labs who were there long before GPT-3. They all care about existential safety, a lot. In the sense that they believe there's a meaningful chanc
154.
by scarmig 8mo ago
I agree that all our theories of consciousness are deeply inadequate. And it it were purely a scientific question, I'd be fine holding off on this question. But consciousness plays a huge role in most theories of ethics, and agnosticis
155.
by scarmig 8mo ago
It would be smarter to take that approach. Google's core competency is technology, technical infrastructure, and research. More mundane things like manufacturing and customer service are... shall we say, less of a core competency. Take
156.
by scarmig 8mo ago
There's a coherent story that straddles both lanes, by assuming that the human economy is in some weird place where the vast majority of humans don't create real economic value and mostly get employment through inertia and custom,
157.
by scarmig 8mo ago
Imagine two bright high school students. One goes to a tony private school taught by people with PhDs; the other is in a developing country. They both want to understand, say, Morse theory. In a world without AI, only the former has access
158.
by scarmig 8mo ago
> especially for biology where it doesn't refuse to answer harmless questions Usually, when you decrease false positive rates, you increase false negative rates. Maybe this doesn't matter for models at their current capabilit
159.
by scarmig 8mo ago
We do know it's not all cars constantly, though. The PGE outage in San Francisco proved it, as anytime a Waymo came across a unpowered traffic light, it was configured to ask for assistance. This led to disaster, as there weren't
160.
by scarmig 8mo ago
I'm convinced my last groan will be reading a thread about Google paper clipping the world, and someone will be moaning about Google Reader.
161.
by scarmig 8mo ago
"Exactly" is impossible: there are multiple Lidar samples that would map to the same camera sample. But what training would do is build a model that could infer the most likely Lidar representation from a camera representation. Th
162.
by scarmig 8mo ago
I read this recently and wanted to post it on August 4. You jumped the gun!
163.
by scarmig 9mo ago
It depends on the situation, and we need more data/video. But if there are a bunch of children milling about an elementary school in a chaotic situation with lots of double parking, 17 mph is too fast, and the Waymo should have been dr
164.
by scarmig 9mo ago
It's a matter of different protocols, not exclusivity. An asker going into a guesser culture is like a client that doesn't respect congestion backoff; the guesser protocol is meant to ensure fairness for clients. The way to deal w
165.
by scarmig 9mo ago
All files related to MKUltra were destroyed in 1973, at the orders of CIA director Richard Helms. So, not so much. Guess the CIA really didn't want their experiments with Charles Manson to come to light. (Yes, yes, there only exists ci
166.
by scarmig 9mo ago
People also forget ARkStorm scenarios, which involve rains akin to 1861-1862, submerging the whole of the Central Valley. Likely several times worse in damages than the biggest earthquake possible in California.
167.
by scarmig 9mo ago
Not really, but diversification often useful; it reduces variance, which is a common goal. And there have been decades (e.g. 2000-2010) where international stocks outperformed American ones.
168.
by scarmig 9mo ago
Data science is not fundamentally about data or science. It's about either justifying decisions that have already been made or delegating decisions to an unbiased casting of bones to let the gods decide.
169.
by scarmig 10mo ago
If I were diagnosed with Alzheimer's, I would seek out assisted suicide. But I think it's more complicated than that: its existence incentivizes pushing people toward assisted suicide. The government finds a way to help with bloat
170.
by scarmig 10mo ago
Cancer. The worst types of it have the advantage of killing quickly. Alzheimer's destroys the self, and you survive a long time with it, leading to much more suffering, both to you (to the extent you continue to exist) and to your fami
171.
by scarmig 10mo ago
> Pieper emphasized that current over-the-counter NAD+-precursors have been shown in animal models to raise cellular NAD+ to dangerously high levels that promote cancer. As someone who's seen both cancer and Alzheimer's up clos
172.
by scarmig 10mo ago
The dynamic seems to be identifying some phrasing as a secret neo-Nazi dogwhistle and haranguing people who use it. However, that doesn't effect any change, so a new word has to be identified as secretly neo-Nazi. Repeat ad infinitum,
173.
by scarmig 10mo ago
Universities believe that constantly putting out pieces that sound like some research is revolutionary and will change everything increases public support of science. It doesn't, because the vast majority of science is incremental and
174.
by scarmig 10mo ago
You can double your company's value by saying it's an AI company. Easiest, simplest way to create value.
175.
by scarmig 10mo ago
If you watch the video, the cat is unfortunate, but not particularly concerning. It was night time, and the cat ran underneath the car from the sidewalk. It's very unlikely that a human would do any better (though, I'd welcome Way
176.
by scarmig 10mo ago
Depends on if the market can stay irrational longer than Gemini stays solvent.
177.
by scarmig 10mo ago
Agreed. It's essentially a giant gamble with a big payoff, and they're both talking their books.
178.
by scarmig 10mo ago
His argument follows almost directly, and trivially, from his central premise: a 0% or 1% chance of reaching AGI. Yeah, if you assume technology will stagnate over the next decade and AGI is essentially impossible, these investments will no
179.
by scarmig 11mo ago
It's at least a little bit amusing that, five or ten years ago, if you opposed big corporate tech allying with government to impose undemocratic political programs, then you were a fascist, while all good thinkers supported that partne
180.
by scarmig 11mo ago
Pretty sure the article is AI slop, so it's kind of connect the dots Ilya has appeared to shift to closer to Yann's position, though: he's been on the "scaling LLMs will fail to reach AGI" beat for a long time.

More ›