Forgot your password?
typodupeerror

Comment Re: At least half true. (Score 1) 58

At some point being faster is a liability if you feel compelled to "make the most of it". Impossible to apply that 'lot of judgement' when you get a gigantic ball of code submitted within a short window of time. I think pointing fingers at the patch is *absolutely* warranted if it turns out that the slopped up patches end up kneecapping the code more often than it happened before. Yes tighter testing discipline is *absolutely* needed, because the patches have gaps that no one is used to expecting, and certainly not going to find in a multi-thousand line merge request submitted by some entry level guy that asked Claude to add some unasked for feature that he thought was 'neat' that touches just tuns of unrelated code for no sane reason (most recently, the guy seemed to induce it to make every call to an external library only be called from a single place and rework all calls in a file to call a 'helper' to call the function from a library, impossible to review volume of code for zero benefit, and offtopic for the alleged meat of his request).

Comment Re:It Gaslights in Natural Language (Score 2) 58

It largely is, and it can be mitigated in many scenarios where an objective test can be run quickly to prove/disprove whether a change worked as intended. This is a critical ingredient in having codegen work, otherwise it will *look* reasonable but have screw-ups. I highly suspect the Microsoft change was likely slopped up and looked reasonable but had a huge gotcha.

But either way, in a vulnerability seeking context, this isn't a huge liability, though it is an inconvenience. Security team started using one of the models on a private codebase and it flagged 60 "issues". Of which, 58 were ultimately not actually issues upon human review, but it found two that had otherwise gone undetected, and two is enough.

Comment Re:Anthropic PR (Score 1) 49

Note that based on what I've seen, while it is certainly true that foreign companies don't get as much value out of their contracts as they would have liked, when they go with the "handshake, money comes later" they can get even more screwed.

Recently saw one such situation where one part of the company was livid that some team had been pushed to do work with neither up front payment or a contract, and the sales team saying "that's just the way things are done culturally here" and surprise, they never got payment.

The "no contracts, just vibes" approach should be an impossibly huge red flag when you get to this scale of operations, even in that region.

Comment Re:Anthropic PR (Score 2) 49

I agree with that, though huge chances the circular funding won't care which CEO is in charge either. To the extent they think it fuels their bubble, they'll keep at it.

I think there was some indication they started shrinking the circlejerk of funding a few months ago though, causing some of the darlings to have pretty catastrophic crashes when their 'business model' of income just being nVidia money to buy stuff back from nVidia evaporated and they had not proven any viable *actual* business yet on their part.

Comment Re:How much of this is due to reduced code quality (Score 1) 42

Well the FAA thing, at least if they are being truthful, sounds something like either a BGP screwup or a broadcast storm (we only get the answer after the telephone game through politicians). Which doesn't rule out AI bullshittery (they could have asked Claude for a good router config and taken it at face value, I know one person who did do that and sent it to a technician and I was able to interject because it did explicitly connect broadcast domains that were supposed to be distinct).

Microsoft and banking are squarely in the realm where AI codegen is likely a component.

In terms of what to look for, for the bullish AI management, it's damn near impossible. Someone submitted a pull request of thousands of lines of code ostensibly for a very minor change. Management is questioning the decision to block it because it's too big to plausibly review and that much change is extremely bad smell for the goal. However 'Claude audited it' so it *should* be good, and are we just being luddites? In my case, we still win in practice by rejecting it, but from public reporting it sounds like some other organizations may not be so fortunate.

Comment Re:Safety last (Score 1) 77

It's not really the automation that's the problem, it's the lack of follow through.

Good, it's automatically monitoring the chemicals. Bad, it doesn't actually do anything about it except send an email.

If it was fully closed loop automation complete with clear and audible local alarm, then that would have been great. Though one could say it should still be backed up by a human doing a periodic manual check, just like my local cheap pool does every 30 minutes or so.

Comment Re:Using AI bug discovery and AI patches is great. (Score 1) 43

It doesn't matter how cheap they are when their use is effectively unbounded. Slop purveyors open 1,000 bug reports or PRs?

These models typically cost on the order of 10-20 cents per MILLION tokens in (assuming cache misses!) and ~50 cents per MILLION tokens out. 1/100th the cost of Fable or Astra, for a task that's far less token-hungry than coding. 1000 bug reports or PRs is perfectly handleable. It simply is not an issue.

Secondly, if there's evidence that a bad actor is involved, a model is perfectly capable of asking you if you want to ban them or banning on its own if you give it permission. And if they're behaving in a way that makes them hard to ban (sockpuppeting, etc), likewise asking you / acting on its own to report the attack.

Where do we send our funding requests?

Right here. Note I'm not going to write your harness for you, but if you come back to me with a harness and a serious proposal, I'll totally look at it.

Comment Re:Shifting expectations (Score 1) 43

I can only find a single example in all of history of a person who was fooled by ELIZA, and only in a "he thought it was a remote-working employee taking the piss" way, not in a "he thought a computer was alive" way.

This notion that ELIZA mass-fooled people is simply not in line with the actual reality we live in. Weizenbaum pushed it heavily in his later life because he was an anti-AI zealot (driven in part due to anger about the use of computers for targeting during the Vietnam War, but also because some people started seriously considering the use of a tool like ELIZA/DOCTOR for talk therapy), with his main example being claims that because his secretary wanted to continue using it and asked him to leave the room, she had been stupidly duped into thinking it was alive. Of which Weizenbaum's own daughter later commented:

I always heard that just really kind of soaked in arrogance. Like, how could she be so stupid to think that this is actually a meaningful communication? Here you have somebody who so needs a landing pad for her feelings that she’s willing to embrace it, right? And he totally misses that, he totally misses the human need and instead speaks of just evidence of this person’s stupidity rather than evidence of their humanity.

Literally, before you ever mention ELIZA again, actually try the thing. For more than 30 seconds. The illusion falls apart really, really quickly. It's not fooling anyone. What actually happened is that people tend to break into two groups. One deliberately tries to get the bot to say ridiculous things to prove what it's doing and show it off as the cheap artifice that it is. The other does just the opposite, "role-playing" and deliberately trying to avoid saying things that would cause the bot to break the roleplay, to try to get the conversation to last coherently as long as they can. Neither group is "tricked".

Slashdot Top Deals

Receiving a million dollars tax free will make you feel better than being flat broke and having a stomach ache. -- Dolph Sharp, "I'm O.K., You're Not So Hot"

Working...