fuzzbazz

If the marketing naming is to be believed, in 1.4nm vs 4nm you'd be able to fit ~twice the transistors in your chip. That's twice the cores, twice the cache... That usually makes it faster.

> on campus between the hours of 10 p.m. and 8 a.m

The intent may be also seen as pro-education: "they're free to protest elsewhere without annoying other students"

It turns out 248.5 days is ~2^31 hundredths of a second.

Was this the uptime of the Air India plane? Or it's just speculation?

> won't hold up in court

Does that even matter? because at 4:50 on the video you can clearly read:

    "You agree to give up your right to go to court to assert or defend your rights under these Terms" 

... in the "binding arbitration and no class action" terms that you need to Agree to.

There is no way to decompile an LLM's weights and obtain a somewhat meaningful, reproducible source, like with a program binary as you say. In fact, if we were to compare both in this way that would make a program binary more "open source".

These names show a "right to forget" notice when you google them. Crashing or completely bailing out when they're about to be generated by the LLM sure looks like a glitch, or somebody forgot that multiple people can have the same name. Anyways, these are now being Streissand efected due to this behavior, the opposite of the intended result of the legislation.

Alternative title: Nobody with an email ending in .ru left in the MANTAINERS file.

We'll know for sure the next time twitch's source code leaks

How can you even have democracy without freedom of speech?

How can you freely choose who to vote without free exchange of information?

Please can we rollback everything to scenario #1 already? I'm starting to think that #2 and #3 are actively counter-productive and the pendulum is going to swing back too hard.

65173 rulings in 230 days by 11 people. That's 25.76 rulings per person per day if we ignore holidays and weekends.

They can't possibly read the cases, is this a kangaroo court?

I mean, it works properly, but a bit slower. The driver just detects if the kind of application currently running (i.e. a game) needs to be confined to only one CCD.

The core to core latency between the two CCDs seems to have doubled from the 7xxx series according to some latency benchmarks I saw yesterday.