Why Isn't Everything Different Yet? (AI, where are you?) | Hacker News Reader

Why Isn't Everything Different Yet? (AI, where are you?)

davegriffith.substack.com

15 pointsdxs8 comments

A lot of words to not say much of anything to defend the actual point - what can AI actually do that we couldn't do before? Is it just that we can make new tech startups faster? Is that a good thing?

> A lot of words to not say much of anything to defend the actual point - what can AI actually do that we couldn't do before?

The article indirectly considers the point:

What could an electrical motor do in a factory that a steam engine couldn't? - There is hardly anything.

But: by tranforming the factory from a central steam engine into one electrical motor per factory machine over a long time, the production process(es) got more efficient - and this is hoped for AI, too.

Of course the idea of many motors powering many machines didn't escape anyone during that period. It wasn't practical until motor technology improved.

Given that we're in the "inefficient motor era", per the analogy, what is the blatantly obvious next business innovation that is currently held back by needing a small improvement (say, faster token generation) from AI?

> what is the blatantly obvious next business innovation that is currently held back by needing a small improvement (say, faster token generation) from AI?

I would personally say: at least the capability for a fast generation of images (designs) - even if not perfect - does have potential considering how much professional designers charge for generating concept designs.

Even if you hire a professional designer to create the final version, the fact that I can generate, say, 10-20 design concepts to see which ones I like, and then go to a professional designer to make a final version based on the AI concept saves a lot of time and money.

(1) We already automated the critical path of a great deal of processes, where that made economic sense. The long tail exists, but it's not replete with as much juicy low-hanging fruit that makes for splashy use cases. Rather, AI will likely produce a tremendous number of marginal improvements over time, which are likely to aggregate and eventually show up in improved top line performance. This could certainly accelerate if agentic AI becomes a true 1:1 replacement for certain types of labor.

(2) Outside of automation, AI is faster search. The information was there and now we can find it more quickly. This helps a great deal, but it's not fundamentally transforming access to information, which was already free and effectively limitless. But there's still value here. I think one key advantage of AI on the search side (for now, prior to meaningful degradations that might ensue) is that it can help push back against exploitative information asymmetry in insurance, consumer goods, health, etc.

Someone standing at a station wonders why the train isn't here yet.

It's a soda station.

This is trying to counter "anti-AI arguments" (term they used), but I suspect a lot of AI proponents would see the idea that changes might take decades as being an anti-AI argument itself.

> Most people consider it to have genuinely transformed commerce sometime around 1999–2001.

Because that is when tax and "munition" laws finally caught up. Internet commerce was already well understood before that time. The major online commerce players we think of today had already been in business for years. But an outdated legal system made scaling a challenge.

What's getting in the way of AI? It seems lawmakers have actually been afraid of holding innovation back here like they did with the internet and are allowing AI to do just about anything, even things that would normally be under intense scrutiny (e.g. copyright violations).

Reply on news.ycombinator.com