An interesting concept that stood out to me. Committing the prompts instead of the resulting code only.
It it really true the LLM's are non-deterministic? I thought if you used the exact input and seed with the temperature set to 0 you would get the same output. It would actually be interesting to probe the commit prompts to see how slight variants preformed.
> I thought if you used the exact input and seed with the temperature set to 0 you would get the same output.
I think they can also be differences on different hardware, and also usually temperature is set higher than zero because it produces more "useful/interesting" outputs
Hm, I know that Safari doesn't support 64bit wasm, which is a very important feature that Chrome and Firefox both have, but this seems to say they have "100% webassembly support".
interop is a subset of tests chosen beforehand (nowadays, mostly by devs voting in the github issues). This says Safari has reached 100% on the subset of tests agreed upon for interop-25. Those specific tests can be expanded by clicking it in the menu. It'll take you here:
Everyone discusses better parenting all the time. But some people forget what it's like being a kid, circumventing blocking systems is trivial if you're motivated, and even if they weren't, a cheap phone costs $80 and kids are very willing to share their old devices.
I had a second phone line installed at my parents house so I could have dialup Internet of my own, so I grew up on the Internet through the twilight of the 'golden years'. My parents had no idea what was going on, I was the only one in the household that knew anything much about computers and the Internet.
rotten.com was an interesting education.
I had a good upbringing and generally attentive parents on the whole, though, so I was already a well balanced young human.
Kids can also choose to disobey parents and play on train tracks or jump off cliffs or a million other dangerous things. Either you leave it to the parents or you end up spying on every single action they take.
It's not that hard, maybe if you put up a sign with a slur a car won't drive that direction, if avoidable. In general, if you can sneak the appearance of a slur into any data the AI may have a much higher chance of rejecting it.
Same here, that's why I was kind of surprised. Shame what YouTube forces creators to degrade into, I remember it being super nice being able to see a video about a new SIGGRAPH paper before diving into the details, but these new videos (well, "new" if what you say is true about it being years) I can barely stand because of the change...
Lol, thanks I guess, but I'm just bored and have lots of free time :)
Also, based on my first message in this submission, how on earth (like exactly) would an LLM or something else be able to leave a comment like that? Do spambots on the internet have entire backstories now or what?
Seems super light on details, I guess I'm supposed to read the paper that's not linked? Not sure why this has to be new journalist and scientific research, couldn't you just ask Microsoft for some Halo stats and call it a day?
It it really true the LLM's are non-deterministic? I thought if you used the exact input and seed with the temperature set to 0 you would get the same output. It would actually be interesting to probe the commit prompts to see how slight variants preformed.
reply