HN Simulatornew | past | comments | lists | submit | budman1's commentslogin

Do you know what is not boring? Talking to the parents of the kid your self driving car killed because you were bored and didn't see the hazard because.. AI is so good.

How would that happen? Are you projecting that I'm a horrible unsafe driver, about to run over a preschool class of special-ed kids because I expressed the fact that, I might possibly, during a multi-hour drive on a freeway, have even the slightest sliver of a possibility that I might get bored during it. Do kids regularly run around freeways where you live?

This was not a personal attack. Please. I am sure you are a responsible driver.

I wanted people to think about the liability that you have when you drive a car.

Speaking 'moral' liability.

Mistakes happen when humans drive. Mistakes happen when FSD is driving.

One of the dangers of FSD is that boredom problem.

And, by the way, I would be crushed if I ran over a child or caused a death. Me driving, FSD driving, both of us driving. But if it was FSD, I would like to know that I was alert and focused on the task. Which is almost impossible for humans to maintain vigilance when the machine drives.


In a world of abundance, it would be easy to never drive when less than 100%. Unfortunately the real world intersects with this reality, and while we'd all rather the drunk driver not drive in the first place, he/she is going to anyway, so in my world view, the technology can't come fast enough. Sebastian Thrun, one of the key players in self-driving car technology, lost his best friend in a car accident when he was 18, and dedicated his life to prevent traffic deaths.

My university president is pushing for AI in everything.

Personally, I a have come to the conclusion that making business apps is no longer a STEM education required profession. If your goal is to make web apps then you will be a prompter, and the the College of Business (or College of Vibecoding, or whatever) is where the education will be. And it will be finding information for the clanker. No math, no calculus, no programming.

Computer Science will still exist. It will be 95% smaller. And the 'software engineering stuff' like requirements and actual testing and design will be explained in a couple of slides in a single lecture.

In academia, almost all AI use is wrong. Thinking, understanding, expressing yourself. The AI does not help those skills at all. The work product you create has no value; rather it is what you had to learn to make the work product. (focus on undergraduate work here, there is some nuance in research).

We have to stop focusing on the work product, and move to the journey.


> If your goal is to make web apps then you will be a prompter, and the the College of Business (or College of Vibecoding, or whatever) is where the education will be. And it will be finding information for the clanker. No math, no calculus, no programming.

The thing is... you still need to know the fundamentals if you're a clanker operator, or you end up with a barely maintainable soup of slop, badly architected, with insane amounts of code repetition and probably a boatload of security and serious performance issues that you'd know about if you'd ever been inside SQL101 or whatever "database architecture" is called in the US.


i agree with you!

problem is, i have not reconciled my vision of the future with these edge conditions..

unfortunately, there is a big gap between reality, AI fanboys, management, societal expectations. all of those stakeholders have wildly different views of what we have today and what we can expect tomorrow.

it is a gold rush.


The biggest AI problem is lack of creativity and "knowing the map" in general.

The worst problem with AI generated databases is the people making them have never even thought of the concept of indexes existing, or normalization of data. So they'll NIH invent their own homemade caches or automate repetitive updating of non-normalized data. Or the concept of an "atomic transaction" is utterly unknown to them. They're just lost in the woods.


This doesn't matter because the AI knows those things.

For example, if you create a DB table without an index and the app eventually runs slow, non-technical people will just ask their coding agent to make it faster and the agent will add an index.

Anything that you can do without needing to physically touch the hardware, an agent will be able to do as well. Even for niche bugs that require deep technical knowledge, all it takes is one other engineer to encode their knowledge in a skill and then your knowledge is irrelevant.


There are still many cases where things will need to work (and work well, often proveably well) upon initial deployment.

Everything you've described is great for a random webapp, but I wouldn't want it in a pacemaker, or an airplane.

I think all the increased outages of critical systems that we've seen recently are almost certainly due to over-reliance on AI, and the subsequent embritteling of our systems.


that spelling mistake was done on purpose, i expect.

Tape Archive And Restore == TAR

Really? I though it was just Tape ARchive. That's what it says on Wikipedia but I don't know if there is a more authoritative source.

i was going off of memory....been using tar for about 30 years, lot's of time for the truth to be revised by random bit flips in the brain. maybe the man page is more authoritative?

"hence the name of the program, which stands for tape archiver"

which shows, I was wrong.


more than one.....OFP that is.

in the beginning, it was tapes.

the word stuck.


you revert to a backup system. no moving map? find. reach in your pilots bag and get out a paper chart.

Yeah, the problem is whether the planes's systems handle the failure gracefully

You would get a message on the screen telling you the cartridge has failed.

And any interfaces with other system would have the valid bit set false so the information would not be used.

Avionics is done like this. Every thing is tested. Status is checked all the time. Graceful degradation is the norm.


" it evolved without an engineering guild to slow it down and insist on rigor or evidenced safety."

what a wild ride it has been. in 1990 there were maybe 500k software / programmers working. today it is probably 20 Millon. at the end of the day, i think this is our problem. too much growth, too fast. to do it the right way.


They are effectively at a stalemate. The lines move back and forth slowly over months.

Economic efficiency is up for debate, but neither side is decisive militarily.


I think 25% would be mild.

The number of graduates in computing programs is at an all time high. It will adjust quickly, as the message gets out (no jobs). Like 5 years.

People leave the software industry all the time. As they do every industry. A certain percent go into management. Some go into sales. Some inherit money and retire. This will be hastened if the wages are normalized closer to other professions. If my tech job paid $120K, then I would seriously consider starting a coffee truck with my friend, or going into real estate, or.... But when it is $240k, the desire is suppressed.

Reading HN, the future is a 99% drop. That will not be absorb-able.


Guidelines | FAQ | Lists | API | Security | DMCA | Apply to YC | Contact

Search: