Hacker News Reader

I have to disagree about AI and doctors, until at least you can get much better sensing technologies. Computers are only as good as the data you feed them, and someone saying they have a pain in some vague region really isn't enough to diagnose anything accurately. It's why doctors tell people to not use the internet to self-diagnose: they quickly determine they're dying of some rare syndrome when really they just have heartburn. A human doctor can interact with human patients (most of whom are not very logically-minded) and figure this stuff out; a computer can't. Remember, a lot of diagnosing depends on the symptoms a patient self-reports, which is bound to be extremely error-prone. Now, if we get to the point where we have Star Trek-style medical scanners, then things will of course be different, but that's a long ways off.

However, I totally agree about the labor-based economy crumbing well before unemployment reaches 50%. If 30% of the workforce is displaced by automation, our economy won't work any more the way it's set up now. It's why we really need a Basic Income.

The trend is obvious, watson can already out diagnose a doctor.

Sure, but diagnosing is actually only a small part of a doctors job. We still need someone to get the input to feed into watson, interpret the output from watson and explain the practical consequences of that output to the patient, help decide on the correct course of action for that particular patient and then make sure that that course of action gets carried out.

Religious fanatics believe that coming off the messiah is inevitable. You don't have any more evidence than they do. This belief that general AI will show up soon and change the world is nothing more than a secular religion.

That's a straw-man. There's plenty of evidence that even limited non general purpose AI will harm the job market enough cause massive social issues; I was very clear, general AI is not necessary for these things to be problems. Self driving cars alone will cause massive unemployment across many sectors over the next decade as it becomes common and unemployment need only get so high before social change is forced.

As for the inevitability of general purpose AI, that's not an unfounded religious belief based on nonsense from the ignorant, it's an obvious and practical step that will be taken because there's every reason to believe it's possible and no reason to believe brains are magical and intelligence isn't achievable, and tons of the smartest people in the world are actively trying to do it. Informed and logical predictions are nothing like superstition and to say there's no evidence is just willful blindness.

They've been working on AI for ages now with little to show for it.

People have been saying the exact same thing about fusion, and it's always "20 years away", but we're no closer to having fusion power than we were 40 years ago.

The goal posts are continually moved with AI because AI never seems to look quite like human intelligence (though why should it?). AI and ML coupled with robots that are relatively simpler than Rosie from the Jetsons are cheaply out-competing human labor in many sectors today, and an avalanche of new sectors is on the cusp (within 2-5 years, not 20 years). Driverless cars currently outcompete human drivers, and a enormous number of people are currently employed as drivers. The remaining hurdles are regulatory. AI is intelligent enough today to outcompete a lot of humans in the work which is currently putting food on their tables. And I haven't seen the burgeoning industry that readily turns former taxi drivers into productive workers in other industries. The industries that can readily accept taxi drivers are generally shrinking as well (meaning, they've got all the workers they need, more than they need, in fact).

Maybe you're right, and it will all work itself out. Some huge growth industry to which humans are uniquely (at least at first) well suited may spring up in the next couple years. But the crisis is awfully close (taking into account driving occupations alone), and there's no market solution in sight, so it seems like prime time to start looking for non-market solutions.

>They've been working on AI for ages now with little to show for it.

That's just not true, it's just that every time the AI crowd makes something really cool and it becomes used, we stop calling it AI because it doesn't think like we do.

> People have been saying the exact same thing about fusion, and it's always "20 years away", but we're no closer to having fusion power than we were 40 years ago.

That's also not true. https://en.wikipedia.org/wiki/ITER

ITER isn't producing any power. If your fusion reactor isn't producing more energy than it consumes, then it's not a useful device. We've had fusion reactors since Philo Farnsworth made his "fusor", but what's always been a pipe dream is one that's actually useful. ITER is not. Go read your own link: they predict it might do something useful in 2027! Whoopee, I can make wild-ass predictions about things over a decade away too, that doesn't mean my predictions are worth anything.

ITER is designed to make more power than it requires, it's now just an engineering problem rather than a theoretical one. You said we've made no progress; that is progress, a plant is in construction, yes these things take time but to deny progress is being made is simply incorrect. Having useful fusion, i.e. producing more power than being input is a matter of scale, it can't be done without building a full sized reactor and that's what ITER is, the first full size reactor. That's called progress.

This doesn't sound right to me. The ITER is going to (supposedly, which of course is probably optimistic) take another 10 years to finish. When you have a workable design for something, it does not take multiple decades to build it, unless it's something on a really gigantic scale (like a continent-spanning highway system). It doesn't take remotely as much time to build a full-scale nuclear fission power plant producing many megawatts of power. That means this thing is not a real design, it's still an R&D project where they really don't know if it's going to work as planned or not, and they're going step by step, collecting data, seeing what works and what doesn't, etc.

So, sorry, you can call it "progress" if you want, but there's no evidence that it's going to work as planned until they actually can demonstrate it doing so.