There's this prevailing narrative that someday self-driving cars will eliminate all human-controlled driving and that these types of stories are like milestones in the path to that future utopia.

But I don't think that's the case anymore than the narrative that LLMs will eventually displace all software engineers.

▲

ggreer an hour ago | parent | next [-]

The invention of the automobile didn't eliminate all horses, but it certainly replaced the vast majority of horses.

In 7 years we've gone from GPT-2 to Fable and Astra. There seems to be no sign of slowdown, except from the top AI labs who want to slow down out of caution. It would be incredibly surprising to me if in another 7 years, my software engineering skills were still useful. AI capabilities will get to the point where even the best human is adding noise to the system, similar to today's chess engines and calculators.

Self-driving cars will take longer to roll out because they require new hardware and changes to laws, but I don't see how a career such as "truck driver" or "cab driver" exists in 20 years. At most it would be for a few niche uses and tourism novelty (similar to steam locomotives or horse-drawn buggies).

▲

eru an hour ago | parent | next [-]

I mostly agree. However, I do actually think that many of the skills you use as a software engineer will still be useful in 7 years. It's just that, as you suggest, no one will let us near any actual code.

▲

htrp an hour ago | parent | prev [-]

>It would be incredibly surprising to me if in another 7 years, my software engineering skills were still useful

This profession has always required reskilling/ changing based on new technologies. How many people are still meaningfully employed as pascal programmers?

	▲	ggreer 26 minutes ago \| parent [-]
		Almost every career requires staying on top of new developments to remain competitive. Even mechanics must learn about new systems in vehicles to be able to repair them. Not many people are still meaningfully employed as Model T mechanics. The difference with LLMs is that they are a substitute for more general cognitive skills. Right now they are also a complement (using them increases the productivity of humans writing software), but it's only a matter of time before they improve to the point where they fully substitute. And it seems quite likely that the timeline for that is significantly shorter than my remaining career. If you are confident that there will be specific tasks that LLMs can't accomplish in 2033, I'm happy to entertain such ideas. But considering the current rate of progress, I can't think of anything like that.

▲

eru an hour ago | parent | prev [-]

Self-driving cars and trucks will eliminate human-controlled driving in the same way that engines (ICE, electric, steam, whatever) have eliminated human and animal powered transportation.

Yes, there are a few people around who ride horses, have an ox-cart or pedal their own bikes, but in a commercial sense, those are rounding errors.

Compilers have already replaced software engineers producing native code by hand. (Apart from a handful of exceptions.) LLMs will and do have a similar impact one level higher.

I'm really, really happy when I see my less technical friends suddenly vibe-code things they have dreamed up!

▲

throwaway85825 an hour ago | parent [-]

And when will the self driving car be able to orthogonal park a trailer by backing down and alley?

▲

eru an hour ago | parent [-]

I don't know what that means. But presumably you are talking about some difficult technical feat that one can perform with a vehicle?

I'm fairly sure, if it's something mechanical and a human can do it, the computers won't be far behind. But even if not, we can redesign the vehicles to make it easier for a computer, eg we can make all axes or even all wheels independently steerable and control their velocity individually. A human would be overwhelmed with that, they can barely keep hold of a steering wheel, but a computer wouldn't mind too much.

	▲	throwaway85825 33 minutes ago \| parent [-]
		Not difficult just an edge case with a more questionable value proposition for implementation. Yes I considered independent wheels but a basic trailer is $3k, which doesn't leave much room for anything complicated. The point is there's a million of these edge cases for which autonomy will not economically compete for another hundred years.