[sorenn111]

Sure there is a lot of hype in AI/ML right now, but this post reads like there is an axe to grind with all ML. it ignores true progress made in a lot of areas and denigrates the whole field.

to me it did not read like an objective post, but more like just a "all AI is bullshit" style blog post

[m463]

I agree. Some people are hyping things, but this is expected.

All new technology is overestimated short term, but underestimated long term.

We might not have autonomous driving that takes our kids at school, but we have cars that can recognize lanes and other cars and complain if something is dangerous. We also have facial recognition, voice recognition and almost turing level chatbots to sell you stuff.

I suspect AI might be analogous to computer graphics. Purists in the beginning said the holy grail was ray tracing. However, people still worked on the problem, marching the state of the art forward with smaller building blocks, and now that ray tracing is appearing, a practiced eye is needed to see the difference.

[AnimalMuppet]

> All new technology is overestimated short term, but underestimated long term.

Well, no, some technology is overestimated short term and also overestimated long term.

For example, flying cars. Nuclear fusion (though that one could still come through). Gallium arsenide (still one of my favorite names for a speed-metal band, and still available as far as I know).

The question is, which category is AI going to be in? AI for specific tasks seems likely to be underestimated long term. AGI? My guess is that it's overestimated long term, because it isn't going to happen. That's a guess. Evidence? Don't have any. Guesses are like that.

[riyadparvez]

I think we need a new word for this kind of posts anti-hype hype. Lot of people try to ride on anti-hype train to fame without bringing anything new to the table.

Anytime there's new progress in AI, you will see many comments or posts some variations of "but humans do it more efficiently" (in arbitrary dimension) or "what about the other problem AI didn't solve". More often than not these are just some lazy layman criticism that makes the posters feel smart without offering anything new or substantial.

[lonelappde]

pg calls it "middlebrow dismissal" and tells HN commentators to avoid it.

[jturpin]

Yeah that's where I'm at. There's this general sentiment that if AI does not solve everything immediately, then it is worthless and hype. Especially the part about not being able to deal with corner cases, forgetting that all AI needs to be valuable is to deal with such cases better than your average human, which isn't a very high bar.

[andrepd]

No, I feel like that's the exact other way around. There's a general sentiment that AI will solve everything immediately and lead to a massive breakthrough in every single aspect of our lives, when in fact what has been accomplished so far in this "new AI" boom is improvements brute-force statistical fitting.

[AnimalMuppet]

Hmm. That means that in order for a human to get paid, they're going to have to be able to do something better than an AI can...

[eli_gottlieb]

It's a "Reverse AI Effect". If the AI Effect is that anything we actually understand cannot possibly constitute artificial intelligence, the Reverse AI Effect is that nothing can possibly constitute useful AI until it rises up and kills all humans.