Hacker News Reader

Comment by ben_w | Hacker News Reader

On the contrary, the loop is:

1. "Computers will never be able to X, as that requires imagination and intuitive thinking"

2. "X is at least 50 years away"

3. Press release: Computers do X (last I saw: Diplomacy (the game not the job), interpretation of medical scans)

4. "X isn't real AI, it's just brute force search/a big database/a glorified Markov chain/linear algebra" (LLMs and diffusion models go here, self driving cars)

5. Computers better at X than most/all humans, but still improved by having human collaborators (Go is either here or 6, IIRC also some variations of Poker, Starcraft, half the Amstrad back catalog)

6. Computers only made worse by having a human trying to help (Chess goes here)

View on HN

That isn't the loop. A lot of predictions about computers have turned out wrong.

The biggest one in my lifetime was the belief that social media would bring the world together, create world peace, and spread economic prosperity.

Usually the technologies that surprise us were predicted by nobody. They just show up.

Go watch "2001: A Space Odyssey". Then go watch some music videos from 2001. You will get a better idea about how accurate these predictions are.

Even GPT is not at all what anyone predicted before the late 2010s, when we started seeing Transformers scaling up and improving rapidly with more data.

Now that we've cottoned onto that trend, it's all anyone talks about. Which is going to make it extra funny when it doesn't deliver what so many predict.

This kind of gets me, too. I'm seeing a lot of this "skeptics said we couldn't do this, but we did," but really? Predictions seem to have largely been wildly unrealistic. The creator of the first hardware perceptron in the late 1950s predicted it would be sentient with a decade. Clarke and Kubrick thought we'd have computers that resisted being turned off and could autonomously control spacecraft (and that we'd have human missions to Jupiter and a permanent station on the moon) 22 years ago, which we still don't have. Back to the Future II thoguht we'd have antigravity in 2015.

I recently did a full X-Files rewatch and the first episode involving AI comes up in season one, when a computer programmed to handle internal building security and temperature control becomes sentient, decides it doesn't want to be turned off or do what its system administrator tells it to any more, and decides to start killing people in the building, via cleverly calculated means that make the deaths look accidental. Scully, the skeptic in this situation, insists this kind of technology is decades away. Here we are, decades later, and I don't see anything with a capability anywhere close to defying its programming and carrying out murder plots that look like accidents.

It feels like people are being fooled into a single decade's worth of legitimately rapid and unexpected advances in autonomous game play and text generation and now acting like the entire history of computing and AI has been tech consistently exceeding expectations, but that has definitely not been true over any appreciable long term.

Most of your examples are literally works of fiction.

The possible exception:

> The creator of the first hardware perceptron in the late 1950s predicted it would be sentient with a decade.

I can't find a reference for (just tried searching, a few variations, no hits where they made that claim).

I've never seen back to the future but the concept of a hover skateboard still sounds pretty cool.

It never happened, but we got the Segway and those little balancing single-wheel skateboards.

>That isn't the loop.

Unfortunately, you're right, that isn't the loop. Any new tech is also the novel weapon of tomorrow. I can't even think of a counter-example.

I'm jealous of your username, zero.

> social media would bring the world together, create world peace, and spread economic prosperity.

I don't it'll happen, but it's a little early to declare yourself the winner.

Fair enough. Give it a couple more decades to play out completely.

Now do flying cars, self driving cars, brain interfaces, 3D printed organs, fusion power, humanoid robots...

For every piece of tech that exceeded people's expectations, there are five more that never got out of science fiction books despite decades of continuous investment and constant media hype.

> flying cars,

Helicopters. Also a bajillion startups that charge 100k, currently iterations are basically human sized quadcopters.

Not sure how relevant any of that is though, as I was specifically writing about AI…

> self driving cars,

I tagged #4 already

> brain interfaces, 3D printed organs, fusion power, humanoid robots.

All #2, IMO, with a few that might count as a #3 technically but not being good enough to deserve it.

Although, again, not sure it's relevant, as they're not AI.

(Have any of those been dismissed as absolutely impossible rather than just hard? I'm not a student of mass media, so I don't know where to look. But brain interfaces seem to me to be what the very first use of galvanic cells on frog's legs were pointing towards, and artificial fusion has been obviously possible in principle since the 50s because of the fusion bomb…)

I'd say that none of these have been proved to be impossible but they're hard in a different way than a lot of the developments we're used to because of Moore's Law.

You conveniently left out the media hype cycles

The hype cycles are not relevant.

Reply on news.ycombinator.com