Hacker News Reader

Comment by paganel | Hacker News Reader

It's interesting because "selfdriving" cars themselves were seen as a form of "AI" at the height of that related goldrush, I'd say 2017-2018 or so.

View on HN

It's absolutely AI. Perhaps this is one of those weird redefinitions that happens when a gold rush takes over a word, like how "crypto" now means digital currency instead of merely "encryption". But what does "AI" mean now then? Generative AI?

Or when "crypto" meant encryption rather than simply "hidden". (For example crypto zoology)

Self driving cars were supposed to be around the corner because AI was advancing so fast.

Instead of autonomous cars, we get generative AI out of this last rush of AI advances. Today, when people talk about "AI" as a startup they are talking about using this already existing technology.

Self-driving was always a speculation that the science would advance into engineering within a sm timeframe, which is always a risky bet. I was hugely skeptical, told everyone that would listen that I didn't believe for a second that Musk would ever deliver on self driving. But I'm far more bullish on AI, we actually do have something new here, it just takes a lot of skill, talent, and discovery to deploy it in useful ways.

The industry (OpenAI, DeepMind, et al) are not seriously working on generate AI. Thats a phase. They are working on AGI. AGI definitionally includes all human cognitive capabilities and that includes vehicular operation and real world interaction (in form of embodied robots). It is likely possible that even self driving cars can be done safely without needing AGI with sufficient scale, and I don’t think it’s that far off from happening.

There are those who are working on AGI, who say they do not know when or if it will show up. Those who do think AGI will appear (if you can get them to define it) think it will come from a gradual increase in capability , just exactly the same way current advances have happened.

Then there are those who are using existing generative AI to book revenue today. This includes OpenAI, and other generative AI providers, but it also includes companies like Accenture, which booked $600M in AI consulting last quarter and expects it to continue growing. There's GitHub co-pilot, which is booking revenue.

> Self driving cars were supposed to be around the corner because AI was advancing so fast.

To be fair, that was almost true. Waymo is doing well, Mobileye is looking good in demos, and even Tesla has a solid L2 system now.

It seems inevitable that this will hit a really dramatic exponential growth phase within the next 2-3 years.

Inevitability has nothing to do with it at all. And there's zero evidence we'll see a "dramatic exponential growth phase...". That's just the same hucksterism we've heard from Musk for a decade.

Unless you limit "Self driving" cars to L2, there's no supporting evidence indicating anything like a breakthrough.

I think your response implies that I'm referring to Tesla as seeing exponential growth in a robotaxi.

Tesla is currently at zero on that dimension. They might get it above zero and grow rapidly in 2-3 years, but I'd give it very low odds.

Waymo OTOH, has increased to 3 fairly difficult areas. It seems extremely likely that their growth rate will increase dramatically sometime within the next 2-3 years. If they don't, there are other competitors with strong technology that might move ahead.

"To be fair, that was almost true."

No, it was not even close to "almost true" except for hype from the various hypesters.

Waymo currently operates in TWO areas, not three. Phoenix and San Fran are not "fairly difficult." Come to Michigan in the winter for difficult. Come to Denver in the winter. Waymo has (wisely) selected areas that are either very simple in layout, that have sympathetic governments, or both. And what other competitors are out there with "strong technology"?

Robotaxis are a novelty. Just like the monorail at Disneyland.

As someone else pointed out in the threads, the Apple Car division was around 2000 people, most of whom (AI folks?) have been redeployed, resulting in a remaining 600 being fired - so it is presumably a set of skills that Apple does not think makes sense to redeploy elsewhere such as other manufacturing plants or retail stores or PR department or finance department, etc.

It's still interesting that self driving cars are framed as being something other than "AI".

But I guess they laid off not the "AI people" working on self-driving cars. They most probably laid off experts in the automotive domain (I don't know, mechanical engineers working on creating parts for the prototypes, for example).

AI is a capability, self driving cars are a product for a vertical; AI is useful for many things beyond self driving cars, and self driving cars need a lot of ingredients beyond AI.

If we have to declare that self driving cars “are” a single thing, I’d say it’s manufacturing, not AI.

It's a promising sign. Technologies stop being called AI when they actually work.

Or they’re people that refused to be redeployed because they didn’t like the other projects.

Reply on news.ycombinator.com