[Meganet]

Thats just not a good analysis.

Blockchain and NTF are and were stupid. A lot of people knew about this and the hype was more of a news hype because we had nothing else to talk about (until ai came).

I have seen so many really good and helpful ai demos/features internally, its impressive.

With AI / ML we are getting self driving cars, robots (talking, listing, walking), agents etc.

LLMs are not crazy good because they can generate stories, they are crazy good because they are a very very good interface to humans.

Facebooks Segment Anything ml model basically solved segmentation problem. Alpha Fold solved protein folding. Nvidias Omniverse with Robots solved the robot motion problem.

AI is not a hype, AI delivers left and right every week there is something really cool new.

Instead of writing uneducated blog posts or just blindly rant about it, at least try to follow up on AI news, you will be amazed how much it solves. And Until we are seeing ANY slow down or ceiling, until then i do believe that this right now is what the iphone was or the internet just crazier.

Its frustrating that people are not even able to understand AI, Blockchain and NFT good enough to be able to separate them. Just because something gets hyped doesn't mean its the same thing as the other thing which got hyped.

And no you were not able to talk to a computer system as fast and good as you are able to do that today with OpenAIs voice input. And no you never had a system which was able to answer that many questions in such a high quality.

[darby_nine]

> LLMs are not crazy good because they can generate stories, they are crazy good because they are a very very good interface to humans.

Do you have an example of a tool that uses an LLM as an interface? Seems like that'd be the fastest way to show people this is a superior interface.

We're obviously a long ways away from star-trek style natural interaction with computers, so I'm curious what you're doing that can work today. Aside from straightforward content generation, of course.

[dartos]

https://www.openinterpreter.com/

Its great for people with mobility issues, since speech to text is so good now.

v0.dev kind of has a hybrid traditional interface mixed with LLM content generation. May not be exactly what you were asking for.

[darby_nine]

This is very cool and I could have seriously used this when recovering from RSI! However, it's not exactly a great argument that this is better than a keyboard and mouse for those who are abled enough to use them fully.

[dartos]

Well tbf, we’ve had decades of improving mouse and keyboard interfaces, but beyond better speech to text, natural language interfaces have been the same for like 15 years.

The mouse was controversial on release as well, since most computers weren’t graphical at the time.

Let the LLM backed interfaces cook. I don’t think they’re a replacement for graphical UIs, but that doesn’t mean they won’t be better for some applications.

Braille, for example, can be read by blind people AND non-blind people in dark rooms. Not strictly better than regular text, but far from useless.

[apwell23]

> at least try to follow up on AI news

your comment seems to based on reading AI clickbait 'news' that claims everything is 'solved' by ai .

[paulgb]

> Its frustrating that people are not even able to understand AI, Blockchain and NFT good enough to be able to separate them. Just because something gets hyped doesn't mean its the same thing as the other thing which got hyped.

I agree with your main point, to those of us on HN there is obviously more substance to the current AI wave than (say) web3.

But I can hardly blame people who aren't actively following tech news from believing that it's more of the same -- many of the VCs and tech media boosters are the same every cycle. If anything, I think it's more of an onus on those of us who do follow closely to sound the alarm on the bullshit. (And there is bullshit this AI wave, too, on top of the obvious substance.)

[some_furry]

Which AI startup do you work for?

[Meganet]

The GenAI Stuff you don't like/hate helps blind people to navigate the world because generating images also allows to analyse images.

There are plenty of other areas which will help a lot of people left and right including you. Cancer research, material research for energy.

Cancer for example can be detected in xrays thanks to GenAI stuff. Its a lot easier to get an x-ray machine somewere around the globe but an expert is hard and we do not have enough radiologists anyway.

[angoragoats]

I agree with the general gist of the article that generative AI is bullshit for many use cases. But I'd rather point out why this comment is misguided for a very specific reason: it's a prime example of why we, as technologists, need to be careful with our terminology around some of these innovations.

You are lumping together many different things in this post. For example:

> With AI / ML we are getting self driving cars, robots (talking, listing, walking), agents etc.

The "AI / ML" part of this sentence is telling. I am aware of exactly zero self-driving cars that are powered by LLMs (what the general public almost always means when they say "AI" these days).

Self-driving cars are enabled by physical sensors in combination with various ML algorithms which have been around in some form for literally decades. I'm not an expert in this field, but my understanding is that what's actually happened in the last ~decade which has allowed them to flourish is the development of better _hardware_, that is, hardware that can run these algorithms fast enough, at a large enough scale, and still be small and cool enough to fit into a car.

Ditto to some extent with your other examples, though maybe a general-purpose robot could be made better by interfacing with an LLM.

I realize this may not be your intent, but by writing in this way, you are confusing the layperson into thinking that all of these innovations were enabled by ChatGPT-style "AI," when in fact some of them have nothing to do with that type of tech at all.

I really wish we'd all be more honest, and not conflate transformers/LLMs with other "AI" algorithms. In fact, I think it'd be good if we stopped saying "AI" completely, though I realize this will never happen given that term's stickiness with the public at large.