Hacker News Reader

Comment by xmprt | Hacker News Reader

With the increasing adoption of AI into people's workflows I think this is something that's really important to be thinking about. AI is a paradigm shifting technology whether you like it or not.

View on HN

Or is "AI" just a kind of a fax machine (from the post) which Big Tech is trying to morph into something profitable (e.g. loading it up with ads/commercial responses "optimised" just for you)

The question is "What if AI is a platform like the smart phone?" and then "What are the apps that make multi-billion dollar businesses?".

What ifs are quite bad questions to predict anything. Also, AI is not. A platform like the smartphone needs building blocks and distribution systems for those apps. Whatever you build with the current llm is not an app, is an interface to a black box you have absolutely no control over. A black box that, on top of that, people can access directly with a chat interface. I don’t think that’s a sustainable way to build a platform, nor it is a good idea to build a business on top of such a liability that you can't control in any possible way.

>Whatever you build with the current llm is not an app, is an interface to a black box you have absolutely no control over. A black box that, on top of that, people can access directly with a chat interface.

so that means there are no apps built on top of search?

An app in this context is something that allows you to do something much quicker than trying to putter around and figure out what prompt to use.

There are apps that are essentially interfaces with a few frills on top.

However I do agree that based on the platform the apps you get will not be multi million dollar apps, they will be more like browser plugins. Low value propositions that does not make you rich, but maybe does well enough that you can spend your time on it.

LLMs are like the capacitive touchscreen that the iPhone introduced. A novel way to interface with computers. It’s being hyped like it’s the iPhone itself. But it’s really just one piece of it.

It also feels kind of garbage at some major tasks we actually want to use it for - just like typing text on a capacitive touchscreen.

Maybe like the Pentium back in the day?

Yes, exactly. Which is why I still think LLMs are here to stay (in local-only form, none of these hosted solutions). We worked around capacitive touchscreen's limitations and quirks. But the hype and valuations around LLMs at the moment is about as ridiculous as if we were to hype whatever company was behind the iPhone's touchscreen back in 2007. And now these touchscreens can be bought by anyone off Aliexpress.

Or they're like the 3D TV screen. Something revolutionary and awesome which eventually fades into irrelevance.

LLMs don’t feel like a transformative piece of tech. It’s more like the CD changer system you can install in the trunk of your car - “oh yeah that’s cool” but still CDs and sometimes it would lock up.

That's to be expected. We are not plugging anything to it.

Try to do that with a capacitive screen.

Isn't this the sort of "wrong question" the author is talking about? If you're stuck thinking in the frame of smartphones, you're going to ask about apps and not the more interesting ways AI can generate revenue?

Reply on news.ycombinator.com