gpm

17 ms·

331.
by gpm 6mo ago
By this logic I also can't fly across the ocean in a modern airplane, but I obviously can. We can get together and do great things. Whether that's science and industry or simply not murdering everyone who doesn't press red.
332.
by gpm 6mo ago
You literally can fix it, you just need at least half the word to press blue to fix it.
333.
by gpm 6mo ago
> Is this repeated or one time? I think it's meant to be one time. > Seems difficult to guarantee you don't kill a bunch of kids each button press cycle... Seems extremely easy to guarantee that I don't kill a bunch o
334.
by gpm 6mo ago
The people who chose blue in no way contributed to the peril you are in, thus you aren't justified in killing them in self defense.
335.
by gpm 6mo ago
The only option that leads to a statistical chance of murder though is red.
336.
by gpm 6mo ago
You have two buttons. If you press one, you're more likely to die. If you press the other, you might murder millions of people. Am I talking about the game, or a preemptive nuclear strike that has a good chance of knocking out the enem
337.
by gpm 6mo ago
> I don't think I want to live in a world in which they all died out. So the blue side would also include the people who are good at game theory...
338.
by gpm 6mo ago
An eager intern can remember things you tell beyond that which would fit in an hours conversation. A disgruntled employee definitely remembers things beyond that. These are a fundamentally different sort of interaction.
339.
by gpm 6mo ago
They look like a pretty small company. The damages they could recoup might not be high enough to justify the costs.
340.
by gpm 6mo ago
And then slow rolled support. And then flat out lied that they received "the correct" documentation justifying the transfer when they hadn't received any documentation, and denied the appeal. Frankly the whole thing is inexpl
341.
by gpm 6mo ago
We actually know that a "100% pass rate" is trivially possible: https://rdi.berkeley.edu/blog/trustworthy-benchmarks-cont/ Anthropic p-hacking the benchmark strikes me as cheating, and somewhat unlikely.
342.
by gpm 6mo ago
Oops, sorry, moved this portion of the comment to a top level comment simultaneously with you replying. Since the part of the comment that was replying to GP was addressed well in a simultaneous comment. https://news.ycombinator.
343.
by gpm 6mo ago
Curiously Opus 4.7 claims to have a 87.6% pass rate and Mythos claims to have a 93.9% pass rate... leading to the conclusion that it's actually possible to "solve" the problems that OpenAI claims are incorrect.
344.
by gpm 6mo ago
> that can be found around 1/20th the price of a brompton Where in the world are you finding a not-stolen bike for less than a hundred dollars!? > none of them come with big gears that I'd need to get home, with sections tha
345.
by gpm 6mo ago
It's rather obviously illegal to leak classified intel by taking public actions based off of it... that's practically the meaning of the word "classified".
346.
by gpm 6mo ago
> Imagine designing an SVG yourself without being able to ever look outside the XML editor! I honestly think I could do much better on the bicycle without looking at the output (with some assistance for SVG syntax which I definitely don&
347.
by gpm 6mo ago
I for one delight in bicycles where neither wheel can turn! It continues to amaze me that these models that definitely know what bicycle geometry actually looks like somewhere in their weights produces such implausibly bad geometry. Also mi
348.
by gpm 6mo ago
I think I'd go with something like x xxx x xx
349.
by gpm 6mo ago
Performing the charity work of discovering bugs before someone evil uses them to cause damage does not somehow obligate you to perform more charity and fix those bugs.
350.
by gpm 6mo ago
It's a statement based on the contents of the articles linked at the top of this comments section.
351.
by gpm 6mo ago
I'm not sure what you're referring to by "that" but I think you're right that it's 10B to not purchase or 60B to purchase, so as an option posting $10B for an option with a $50 strike price.
352.
by gpm 6mo ago
Definitionally both circumstantial and direct evidence are forms of evidence. No modifier is necessary. And incidentally you can be convicted in a court of law purely on circumstantial evidence, and that's the place in society where we
353.
by gpm 6mo ago
$1B per month on AI services does not seem remotely plausible to me... Engineers don't consume that many tokens... And on the AI development side they're the ones providing compute in the form of a "million H100 equivalent C
354.
by gpm 6mo ago
Paying $10B for the option is also crazy though. Paying $10B for the thing outright and not just an option would be absurdly high.
355.
by gpm 6mo ago
Simultaneous with GitHub copilot dropping support for the Opus models in their $10/month plan...
356.
by gpm 6mo ago
Yes and no. For a truly malicious actor, you're right. But it shifts it from "well we aren't obviously committing fraud by quantizing this model and not telling people" to "we're deliberately committing fraud b
357.
by gpm 6mo ago
> And if that’s the case, would it happen much quicker since technology has been speeding up year by year? I wouldn't expect this. Historically we've had a roughly exponential rate of shrinkage. If we keep that same exponential
358.
by gpm 6mo ago
> but is there an actual coordinated effort to compete with [...] the West Yes, absolutely. China regularly produces long term planning documents to coordinate efforts, and the latest ones have specifically prioritized technology like ch
359.
by gpm 6mo ago
Yes... but the agreement only says they won't train on your data if the law is already preventing them from doing so.
360.
by gpm 6mo ago
Considering the propaganda value in controlling the inputs to the machine that answers peoples questions, I rather expect them to be subsidized forever.

More ›