Is this not one way of doing it? And it's a way that will ensure future code contributions are much less likely to introduce stupid bugs (assuming you agree that Rust does just that).

I don't understand all the naysayers claiming it can't be done. Of course it can be done!

Weird stuff in the code (example above was a char array used to hold binary data) can be examined for how its used to deduce the correct implementation - one of the things that LLM's are actually good at. And since it's talking about large C codebases, they're likely well established and completely devoid of things like tests... so have the Rust output C-compatible libraries and run them through the exact same tests! Any remaining issues are just as likely to exist in the old codebase, but now you can sick a more junior dev (or an LLM) on those issues without worrying about introducing classic C errors.

IMO, the real question is if it's worth doing. Where are the arguments against doing it? The rust folks have some strong talking points on its advantages. What's the benefit to keeping a large C codebase in pure C if/when this option is available?

Nutters? It wouldn't be the first time he's sought out immunity not provided to anyone else (... and got it) - SCOTUS 23-939 Trump v. United States https://www.scotusblog.com/cas...

No, Android Auto (https://play.google.com/store/apps/details/Android_Auto?id=com.google.android.projection.gearhead&hl=en_SG)
is not the Gemini App (https://play.google.com/store/apps/details?id=com.google.android.apps.bard).

The statement from TFA may simply be wrong.
IE: "Pichai is talking only about people who are opening the Gemini app or visiting the Gemini web interface to enter a prompt or access Gemini Live."

That statement wasn't made on the actual announcement that Pichai had released: https://x.com/sundarpichai/sta...

But, if that is accurate, then Android Auto usage doesn't count towards this 1 billion active monthly users stat, nor do Google Search results that include their AI results, nor does clicking on said results to get additional output from the LLM. But it's just some statement made by Ryan Whitwam from Ars Technica with no attribution, and Pichai's announcement lacked any of those details, so who's to say?

You kicked off this thread while noting, "people who went to Google search". I hope you can agree that those are not included in the stat (or, at the very least, that's not what the stat claims to include). FYI, that's not the "Gemini web interface".

WTF is this "again" bit referring to? That was my first reply on this thread.

Where in, "or visiting the Gemini web interface to enter a prompt or access Gemini Live," does it say that, "people who went to Google search," are included in the stat?

FYI, that's the context of the discussion that the AC was pointing out as missed by you... your own comment. Good job highlighting the part that contradicts your initial statement.

Not according to TFS: "The 1 billion-user metric has nothing to do with any of that, though."

But what you claimed was NOT about going to the Gemini web interface. You explicitly said, "Google search" and its AI search result, which is NOT counted here.

That didn't include drilling down on the AI results or going into AI mode - which may or may not be counted in this stat (I read TFS as that NOT being counted, but maybe someone considers that to be part of "Gemini web interface" that counts for this stat).

Sorry, but AC got it right here. No need to double down on dumb.

IE: The 1 billion user stat is ONLY counting "people who are opening the Gemini app or visiting the Gemini web interface to enter a prompt or access Gemini Live". Y'all don't seem to be reading past the first half of sentences.

As clearly stated in TFS, that wouldn't be included in this statistic.

IE: "Pichai is talking only about people who are opening the Gemini app or visiting the Gemini web interface to enter a prompt or access Gemini Live."

Thanks! Those seem like valuable estimates and some decent info. Makes me wish they included some actual benchmarks even more :-)

I might have to give it a try just to see how slow it is on just CPU + memory - I've got a spare HP Z something-or-other with 128gb RAM, but the CPUs are pretty ancient.

Feels like such an announcement should include a few examples and their relative performance stats. Instead, the main linked article includes nothing of the sort, and Meta's announcement only notes it would fit in 20gb of memory, so you'd only need a 24gb or 32gb card - normal consumer level stuff, right? /s

That sounds like a kids toy I wouldn't be surprised to see in the store at any point in the future. Reminds me of those toy phones that would play canned messages, but pop in a small LLM for the replies.

IMO, the scope appears to be different. I think it's:
* Flock 1 in a million - one false positive out of all reads, including false positives, positives, negatives, and false negatives. Furthermore, that's by character, rather than by plate.
* LAPD 32.3% error rate - of the Flock positive alerts, 32% are wrong in some way (which could include errors in the NCIC database).

Flock picked the stat that appeared to be most favorable to them. Like this comment points out (https://yro.slashdot.org/comments.pl?sid=24065856&cid=66281440), 1 in a million character misreads would equate to about 1 in four plate reads being wrong, which gets pretty close to that 32%. The additional 7% could easily be due to other tangential errors (ex. number was in database but should not be).

More-so, it highlights the misinformation used here. They start with:

... which equates to about 1 in 4 plates read wrong. Then state:

Well, that should be WAAAAY higher. They're doing an absolute shit job of catching the errors then! That stat isn't saying what they think it says. It's almost like they're not even trying and just assume it's right! Why, there might even be lots of examples of innocent people getting abused by the police because they didn't do their due diligence, like this also from TFS:

Why would that EVER lead to someone getting mauled by a police dog?!?! Even if that WAS the right plate, that shouldn't be the outcome.

From TFS:

I hate the Flock situation as much as the next guy, but this guy wasn't mauled by a police dog because of a Flock mistake. Look elsewhere for that blame and fix that shit before we move forward.

Agencies that have, and continue to, actively prove that innocent people get caught up and abused by their system should not get access to this in the first place.

How many ALPR's are also giving out tickets automatically? What's the acceptable false positive rate for those? (It should be zero false positives, and false negatives should be acceptable; The many false positives could be put through human review, but we can see the issue with that right here)

FYI, the median age of TV viewers was NOT posted, despite what was implied. That was the median age of US residents.
Aso, two of the categories I posted were categories that would be considered streaming viewers. Their median age was still, according to him, high.
Lastly, there was no counter to the income disparity. IE: households that rely solely on OTA TV had significantly lower incomes, which was the whole fucking point.