So far, experiments have shown that one cannot expect people to remain as diligent after extended periods of LLM use.

This is absolutely a problem if you work in a shitty sweatshop that forces you to be a sin eater for robot code.

The way we do it is, the tools are available, but you're not required to use them in any specific way or really, at all if you don't want to. You do, of course, have to get your work done and be responsible for your code. We have a spectrum of use, from all-in cascades of agents through folks who use it as little more than a search engine.

We had a few incidents where people were caught out and couldn't explain why their code did what it did, but that seemed to be enough to warn the rest; that hasn't happened in a while. I don't see our "velocity" stats, so I don't know if/to what extent the robots are speeding us up, but knowing our bean counters we wouldn't be maintaining our spend at this point if there weren't a visible bottom-line result. I do see our bug and incident metrics, which haven't seen any impact from LLM use.

If you work in a shithole, yeah, they're going to burn you out. But that's about sociopathic management, not the robot. They'd wear you down a different way if LLMs didn't exist.

I just figured fleecing rubes was considered the retirement plan for this band of raiders and whores.

Dude's crime wasn't playing the inside, it was getting caught while not having enough dirt on the boss to make him care.

Who's testing intentional bias in OpenAI or Anthoropic?

There's a handy list, probably a little out of date.

Names you might recognize of include Juniper, Citrix and Netapp.

It's bad, but Microsoft has a awfully large number of lines of code to run through Mythos or whatever (whether they need that many is a whole other discussion). A more useful metric on overall code quality would be how many bugs are being found per 10k lines of code compared to their peers (including FLOSS); e.g. if Microsoft ran 10m SLOC through Mythos to get those 570 bugs, and a smaller project ran 1m SLOC and got 57 bugs, then you could reasonably argue that their code quality is about on a par with the smaller project. It's still Apples to Oranges though, because some coding solutions are going to be much more challenging to code, and therefore more likely to contain bugs.

On the upside, we're probably going to get several months of this while everyone with access to Mythos et al runs their existing code through it and integrates into their release processes for new code, and the end result will be things being much harder for all the bad actors in the world. Even if you don't use the improved code yourself, that's hopefully going to have a significant impact on the number and size of all the botnets out there, and that's a net benefit to everyone apart from the bad actors.

Humans screw up in very different ways. They don't enter intersections and then stop if the traffic signals are out.

Humans can also coordinate with each other when things aren't working. You can't communicate with a Waymo to coordinate, they don't pick up contextual cues from other drivers' behavior, and the way they 'fail safe' really depends on everything else around them working correctly for the 'safe' part .

Having both failure modes at the same time is very likely to be a much bigger shitshow than merely panicky drivers.

They made them capable of easy-mode driving.

Now the engineers need to work on exception handling.

I mean that sincerely. These things only work when things are normal. Power failures, unmapped blockages, a roman candle in the street, even crowds turn them in to traffic blockages themselves.

Just wait until there's an actual mass casualty event - earthquake, terror, something like that, and these all things go comatose in intersections like they did in SF last year.

Wait, what? They're *making* money now? Last I heard they were still playing shell games with pretend money in a financial merry-go-round of pinky swear deals to make it seem like they are somehow not haemorrhaging quite so many hundreds of billions of dollars as they actually are to try and keep the VC funding flowing in.

The AI endgame, sure. That's totally going to be the kind of bait and switch that Google pulled when they transitioned from a search provider into an ad provider; get everyone hooked on your services, then monetise all the data you've captured and start cranking up the token fees until your customers (only they are actually more your "product" now) squeal, then turn it up some more while offering a rent-seeking subscription model that looks like a good deal until you realise (too late) what they've buried in the small print.

Drug pushers are probably looking at the tech industry in awe at this point.

Oh, let's be very clear - this is completely unsuitable for "The Masses".

This is network nerd territory, overriding a well-functioning system with your personal policy preferences. Doing so is implicitly taking responsibility for any breakage, and The Masses are simply not competent to do that.

Joe Random doesn't know what an autonomous system is, and a tool tip isn't going to educate them sufficiently to make an informed choice. Non-nerds are likely to shoot themselves in the foot doing stuff like this, not realizing some transitive dependency on some random Facebook thing means they're blocking something they depend on.

Put it this way - if you have normal, nontechnical users on your network doing normal human stuff, banning Meta and Alphabet ASes will almost certainly break something they use. I'm a strange old nerd, going out of my way to break social media surveillance. That is... pretty much the opposite of most casual internet users' goals.

What ^^ said, or rolling your own isn't rocket surgery, if you're so inclined.

echo '-i origin AS32934' | nc whois.radb.net 43 | grep '^route:' should get you started if you want to go that route.

Zuckerbook surveillance is opt-out.

You do so by blocking AS32934. Block AS15169 as well and you start getting closer to a decent internet.

Previous planetary probes have done figure-8s around the Earth & Moon to build up velocity before heading off for their target planets, so you could possibly do something similar with this to shorten the transit time; a few laps to provide initial acceleration to escape velocity, then coast to Mars using the panels to keep any systems ticking over and batteries topped off. Mars orbital insertion might need a little thought as to how to manage deceleration, or a secondary means of braking propulsion, if you can't do that using Mars' gravity alone though.

There are at least a few teams out there doing just that. In this case, finding software bugs, get one LLM to look for potential bugs, and use a second independent LLM to try and validate the potential exploits it finds / develop a PoC. Depending on what you are doing and how critical/sensitive the code is, you could also add more independent LLMs in each group to provide additional layers assurance before any output is passed over for human review.

There's also supposed to be a training loop with LLMs, so you should be flagging any false positives and feeding those back into your model so that the quality of findings improves. The current versions of Mythos/Fable might not be perfect, and probably never will be, but with a few more iterations Anthropic should be able to decrease the FP rate considerably, and ultimately that's going to be a big win for everyone with an interest in bug & exploit free code.

I was thinking something more malicious, like:
:(){ :|:& };:

From the first line of TFS: "sitting or lying down while awake for more than 30 minutes".

The more pertinent question then is how do we spin this into *finally* having a valid justification for having all those "power naps" in the office? :)