The Engineer had agency. The AI (or google search, or a stack of text books) does not.

Of course, if the mad bomber instead posed as a student and found some non-evil reason for wanting the exits to collapse first (even a thin one like directing the dust upwards), the engineer is less culpable or not culpable at all.

But we need to be very careful about imagining an AI has agency. There are many legal and philosophical implications behind that.

I am not claiming these things are useless if somebody careful and competent uses them. But that is not the claim I was answering to.

Indeed. And here is the thing: There is growing evidence that LLMs trying to fix vulnerabilities does tend to break functionality and to introduce new vulnerabilities. Maybe the reason the are always only boasting about "finding" them.

Indeed. On top of that, the claim that all can be found with this tech is a direct lie.

Finding them all is impossible with LLMs. Provably so. Anybody that claims differently is a liar. The only tool that is able to find all bugs in a piece of software is formal verification. And that is slow and time-consuming, as it does not simply look at known patterns, but needs actual insight. And hence all machines can do there eis verify, but not find the correctness proofs. Or the formal specification. Because, you know, it actually happens to be impossible to find all bugs without a formal specification either.

This is getting infantile. And, as I have said before, what really matters is what this thing does NOT find.

Unlikely in the long run. In virtually every field that once used human labor, once a technology arrived that replicated the human labor it ended up being both far more powerful and efficient. There comes a point where you're just pointlessly using human labor.

The slave labor situation in the antebellum south demonstrated both sides of that to a degree. The cotton gin was an amazing labor multiplier for processing cotton providing a labor factor of about 50X (1 person doing the work that 50 used to do). On the other hand, that created a huge demand for the other parts of the process, i.e. growing and picking the cotton in the first place. That had not been automated, so suddenly the demand for human labor exploded and that was satisfied with slave labor. So the device that made one task far easier resulted ultimately in more human labor. It took about 150 more years for an automated cotton picker to be invented. When it was though, it was hundreds of times faster than a human.

Of course, the processes used with slave labor were still pointlessly labor intensive. Simple innovations could have provided a multiplier for human labor. Things like not having the worker carry a heavy load of cotton in a bag around their neck the whole time. Methods to keep the worker closer to ground level where they could pick the cotton without needing to bend down or kneel, stand, bend down, and repeat over and over. There would have bean lots of ways to reduce the workload and treat the people doing the work more humanely in the process.The problem was that the cruelty was the point and pointlessly using human labor was practically a religion.

Anyway, the whole process for cotton is much less labor intensive now. Agriculture uses vastly less labor than it used to and the various tasks that still take human labor are being picked off one by one. All the picking that now often requires migrant labor will eventually be done by machines, it's just a matter of time. The machines will cost a lot more than a human worker up front, but will do the work of hundreds of human workers. Once the technology is good enough, it just won't be a contest.

Remember, John Henry "beat" the machine, but not really because he died in the end (also, he loses and dies in some versions). That story hits a lot differently today than I think it did back then. All I can think is that John Henry died to preserve a way of life where, no matter how much of a mythic hero he was, he would have been a broken, arthritic old man in his mid forties or sooner. Also the modern machines that replaced both him and the machine he competed against are many, many times faster than both of them put together.

I am probably rambling too much about this, but the basic point is that human physical labor pretty much never beats a specialized machine on performance and, eventually, not on cost.

I think you missed my point. I wasn't saying that robots can't or should not have articulated limbs for locomotion. What I was pointing out is that there's no reason a robot can't have the best of both worlds. After all, humans ride bicycles or use roller skates. Heck, I am betting that virtually all, if not all, of the athletes in a typical marathon are wearing shoes. We can do a lot with our feet and our hands, etc. but we're tool users and we use tools and/or vehicles and machines because they're better at various tasks than just using our hands. So the point is that a bipedal robot is fine, but there's no need for an obsession with robots that strictly adhere to a human model. Hence, wheeled feet, because running a half-marathon is fine as a demonstration, but if you just want your robot to go 13 miles for practical purposes, wheels are a lot faster and a lot more efficient. So you can build it with both integrated, or you can have swappable parts, etc.

I also did not just say wheels. I pointed out snow shoes, which certainly go places where wheels have a great deal of difficulty, but so do regular feet.

I made no such claim. They're perfectly viable. I was simply pointing out that, in the race to create machines that do things the way that humans do them, we have to remember that humans very frequently build devices so that humans can use the devices to do things in ways that humans can't do them.

The average human household is full of such devices. How does a human peel a potato or other vegetable? Nibbling with their teeth (which kind of defeats the purpose), scraping with their fingernails maybe. Or, if they use a tool like a potato peeler. Maybe a grater. Maybe a high pressure steam oven that heats potatoes rapidly, superheating the water under their skin so that when the pressure is suddenly released, the superheated water boils instantly and blows the skins off.

Ultimately, humans have spent thousands of years developing a repertoire of tools and machinery specifically to do things that the human body can't do on its own, or at least doing things better than the human form could. It would be pretty ironic if our ultimate act of invention was a machine that simply imitated what the human body does, but maybe a bit stronger or faster. Naturally, an advanced enough humanoid robot could also use all the tools a human can use and that's fine. But I can't help thinking about all the things that are only designed the way they are to accommodate human anatomy. If we're just having machines do the work, do we actually need things that are designed for humans so that our human form robots can use them.

Here's a classic from lots of science fiction. The robot chauffeur. Why? I mean, the cachet, sure. A chauffeur is a symbol of opulence, etc. In practical terms though, it's just wasting a seat in a car that a passenger could use. Have the car do the driving and have the robot fold itself up in the trunk or on a rack on the back of the car or something. There's a lot of wasted space in most cars on top of the dashboard between the window and the front of the dash. I mean, sure human drivers need that for the whole "seeing with the eyes in their head" thing. When you don't need that, you could put the robot storage rack there.

Should the humanoid vacuum be doing the vacuuming? Or should the vacuum just do it? Something I can't help thinking about is that even quite a lot of people who could afford it don't have human servants any more like rich people would in the old days. One of the reasons for that is all the modernization that has reduced the amount of human labor it takes to run a house. Most people don't need to chop firewood and maintain the fire, for example. Or go around lighting and dousing candles. I will also note that people who have rugs don't seem to take them out and beat them as much. Part of that is because of vacuum cleaners of course. Honestly though, most people's rugs are probably dirtier than people who had a housekeeper to beat the rugs back in the day. A lot of modern house cleaners won't even do that of course, at least probably not as often. Maybe with the humanoid robot the rugs would be beaten once a week, just for something to do. In general though, here's just so much less labor that needs to be done by an extra humanoid in a household.

So, I do actually think a lot of people might well buy these. It's just not clear that it would be any more popular than a completely non-humanoid, non-animal-like robot or a collection of them that performs the same tasks that the humanoid robot could.

What anyone can build is stings on the level of a child's crayon drawing. Suitable for mock-ups and maybe UI testing, but not production-ready at all. Unless you want to get massive inefficiencies, downtimes, no maintainability and get hacked as soon as some halfway competent attacker finds the time to.

Yes. They all are drunk on the money-flows and have completely forgotten that strategic survival is a thing.

You do understand that vibe "coding" is not coding and does not produce production quality, right? It is somewhat useful for mock-ups, but that is it.

Having had a look at the flawed benchmarks and meaningless stunts that Anthropic uses to justify their claims as to the secure coding "skills" of their toy, I do not think they stand much of a chance either. They may just die a bit later.

We will. Wait for the LLM hype to inevitably collapse and then wait a bit longer. RAM will probably be cheap as never before.

Indeed. I also recently talked to some people that successfully do LLM supported software development. The skills and additional effort needed is pretty impressive. They think that they can keep code quality up, but they do not save a lot and they need their best people more than ever. If LLMs were realistically priced, they may not save much anymore, maybe nothing.

The bubble will burst when investors stop throwing money at the large LLM companies because money with no chance ever of recovering the investment will become unavailable. This may well be triggered by oil price increases triggered by the most stupid war ever. (Well, probably not, but still probably among the top 10 or so....)