The current climate in AI has so many parallels to 2021 Web3 | Hacker News Reader

The current climate in AI has so many parallels to 2021 Web3

twitter.com

14 pointsGavCo8 comments

The big difference is that AI applications like ChatGPT and Midjourney have demonstrated actual value (if only for entertainment purposes for now), while Web3 hasn't been able to convince the market that it offers any value at all.

That’s addressed in the source.

Except there’s actual demonstrably useful value coming out of the “AI” buzzword because these are technologies that can actually be used to do things, whereas “web3” was never able to demonstrate a single actual use case or valuable result.

So yes “AI” is a buzzword, but no “AI” is not web3: it’s the buzzword name for a set of technologies that can actually do something of value.

Now whether “AI” is actually “intelligent” in any real way is a different discussion.

hmmm this makes an interesting point. Web3 has disappointed me in it's adoption so far. And I remember being told how fundamentally changing it would be very soon.

Objectively speaking, AI has delivered on much of its hype going into 2023. Generative AI applications actually exist, and consumers love them. There are a variety of B2B products doing various kinds of classification and prediction which have grown popular.

On the other hand, The burn rate in the field is insane - I'm not certain that even a small portion of "AI" businesses are on the right track or delivering something other than vaporware. With ad-revenues potentially declining as data evaporates - even the idea that 1 out of 10 will make it and make the rest worth it seems dubious.

A veterinary clinic local to me tried an AI service that attempted to diagnose issues found in x-ray images. It was a pretty slick set up, they just had to wire up their image archiving system to the service's endpoint, and then sending it to the service was as easy as a button press. I think they said it cost $10 per shot.

They had it for awhile and then dropped it. Said they didn't feel it was really delivering any value. It wasn't catching anything they missed, and they didn't want to trust it too much in case it one day it missed something they missed.

I can't remember the name of the company, but I remember thinking maybe they might find better use repackaging it as a product to help train human students.

Replacing high skill professionals always seemed like a bad bet to me. Augmenting experts such that they can do more, or replacing low skill work is much more promising.

I’d love a cheap phone app which could diagnose skin conditions/recommend otc treatments. If I’m talking to a vet I’d rather a person look at the issue.

If there's one thing I've learned with a career in tech, it's that people who finance this stuff rely almost entirely on signaling from peers to make their decisions. They are fundamentally incapable of evaluation based on merits and in my experience are rather deaf towards "obvious" signs of progress like the release of DALL-E, until they see it shows up frequently on Twitter.

As it turns out, having a highly privileged upbringing where you are evaluated purely on "success" (procuring more wealth) leads to a lot of "this is like studying for a test and passing". But, you haven't developed a sense of intrinsic reward and care way too much about social reward instead. Sometimes these will align.

Having said all of that - I think this tweet thread is really only applicable to the funding side of things (and I imagine it's quite true in that regard). But comparing the real results of machine learning over the past few years with...Web 3? That's a stretch. This caveat is cleared in the thread but it bears repeating. Web 3 was a grift designed to allow major players to cash out their "coins" back to real USD by rounding up a final batch of greater fools.

Reply on news.ycombinator.com