There's a bunch of what you say that I agree with*, but then you start going to crazy-town with your talk of "crypto-communists". Especially right after you proclaim "welfare for life" as a solution for the displaced people, which is the very essence of "to each according to their needs".

* In particular, I see "fewer jobs" as an intrinsically good thing. Yes, we all understand that leads to a wealth distribution problem. There are multiple possible outcomes.

He's clearly talking about evolutionary algorithms. There is a human involved, yes. But that's moving goalposts. Scribing for copying text was replaced in part by the printing press, then further impacted by photocopiers and later printers. Yes, a human still fed texts into scanners, or nowadays types "10" in the # of copies field of their word processing document, but you can hardly call that human a scribe.

When it comes to 'software replacing teachers', we really haven't made many fundamental advances since Gutenberg(who at least substantially increased the percentage of the world's books that weren't produced by students taking lecture notes in class, which presumably meant that you at least had the option of reading the textbook and skipping the class). If you just need information, technology has done quite well, and continues to make improvements; but if you aren't ready to turn information into knowledge all by yourself, there isn't much on offer.

There is a certain amount of irony; but it's those years of expensive and supply-limiting training that are precisely what make such an attractive target.

It's not an easy target; the computer system that ends up replacing your radiologist or your lawyer or whatever will likely have cost far, far, more to develop than the human it replaced did to raise and train(even if you count the human's recreational spending); but the computer's ability to do work will just keep increasing if you buy more silicon, while the human doesn't scale. If you could hire a single radiologist and make him more productive just by buying additional office chairs, you probably wouldn't bother with the robot.

Based on my (layman's) reading of CDC data from the Morbidity and Mortality Weekly Report; it would appear that we may have settled on a solution for the middle aged and surplus.

It's just not a terribly nice solution.

Isn't a corporation a robot that exists specifically to hold assets? Some of them are better at being rich than others; but surely a corporation that owns the expert system that runs the corporation would be a rich robot; were it successful...

Depends on the political climate: if some bleeding heart is calling the shots, sure; but if it's tough-on-crime time, then the rapidly maturing world of combat robotics will be tapped to provide low-cost 'treatment' solutions to these populations.

'Real' empathy would require a strong AI, more or less by definition(and a relatively human-like strong AI at that). Conveniently, though, there's no externally visible difference between real and fake empathy, and faking it is on the level of passing a Turing test, which is hardly trivial; but likely to actually happen in the comparatively near future.

The best diagnosticians might actually be the ones who see the chopping block sooner. Traversing decision trees, crunching patient statistics, and doing machine vision on whatever comes back from radiology and histology are all things that computers are either already good at or improving and plausibly expected to continue to do so at a reasonable clip. "Getting a patient's report of their symptoms and making them feel as though they've been duly listened to" or "calming some screaming brat long enough to innoculate it" are not things computers are terribly promising at. However, they are things that can be done, even done well, by basically the cheapest category of went-to-less-school-than-the-doctor-or-some-of-the-fancier-types-of-nurse medical workers you can legally get away with using for the task. Somebody will still have to do medical research; and it'll likely take a while for the public to accept that ResectXact(tm) software is a better candidate than some well-reputed surgeon to chop them open and do some maintenance; but attrition is likely to be brutal among the relatively expensive people whose specialized skills are amenable to expert systems and whose bedside manner and basic patient interaction are no better than a much, much, cheaper nurse or tech of some flavor.

Lawyers, in the same way, are going to require some people who are sharp enough to not fail during oral argument, who know how to work a jury, who can project a besuited air of consummate professionalism when dealing with clients who are paying well for the services of Somebody, Somebody, and That Other Guy; but it's hard to imagine that humans are going to last long against glorified search engines when it comes to "Traverse the entire law code and case law, give me the top hits, flag anything from things that the presiding judge has cited in decisions he has written in the past". Until computer generated text stops sounding so much like markov chain word salad, they'll probably still need some peons to stitch things together; but that will be a dead-end, unbearably soul crushing paralegal sweatshop of misery, not even an entry level job.

The one major complication to keep in mind is that robots/automation almost never literally 'replace' you. Rather, they allow for a different way of doing things that no longer requires you.

Robots built to replicate human capabilities are, despite continued effort, relatively pitiful. Competent bipedal locomotion, a couple of dexterous hands, fallible but very, very, adaptable image recognition, etc. are a fairly tricky package to put together on a reasonable budget. Outside of tech demos, that's why you don't bother to build the robot to resemble the worker, you restructure the task to play to the strengths of the robot(see basically all contemporary manufacturing processes). This task restructuring can also involve the user: replacing a telephone operator, say, would have been impossible until relatively recently; you need speech recognition software good enough to do the job and computers cheap enough to run it. So we didn't: Pulse code dialing allows line switching to be done with relatively simple electromechanical devices, which is why operators were on their way toward the exit more than a century ago, despite AVR 'agents' still being considered lousy and terrible to work with today.

You will almost always be misled if you try to predict odds of replacement based on 'what the job requires' rather than 'what the job produces'. Beating the people currently doing a job at the skills that the job requires is difficult, frequently impossible or uneconomic. Achieving whatever goal their job exists to fulfill(or achieving something else that eliminates that goal); is almost always how it gets done.

What I'd be interested to see is if, and how aggressively, they take action against image collections that are not of any use for their desired purposes.

They obviously can't be too capricious and unpredictable, or they'll spook users; but you can't offer 'unlimited' storage without making some provision for 'that guy who hacks together a FUSE filesystem that uses images uploaded to Google Photos as a storage medium' or the 'Cool, this will make my next time-lapse video project way easier' cases.(and, of course, if you are feeling particularly uncreative, /dev/random just needs a dash of formatting information to be as many bitmaps as you could possibly desire.)

Are they just going to go with the ISP-style 'I said unlimited; but I actually meant X photos or Y GB of traffic per month; apparently I'm allowed to get away with that, so STFU', are they going to have peons manually examine accounts whose size gets out of hand and decide what to do?

I did, and here is the definition they came up with:

http://www.merriam-webster.com...
http://www.merriam-webster.com...

Didn't see your (inaccurate) definition there, so maybe they already fixed it.

We can only hope that Uber's notoriously...risk tolerant...approach to just ignoring regulations that they don't like will result in a lot of spam that is actually 'spam' for the purposes of the CAN-SPAM Act of 2003 being sent out.

That particular law is more or less a dead letter, given how easy covert or extraterritorial spamming is(and, of course, it's assorted gaping loopholes); but there are theoretical penalties that could stack up fast if you actually fuck up.

In this case, if grabbing people's contact lists doesn't count as 'email address harvesting' in the context of the prohibition on sending to harvested addresses, I'm not sure what would.

Honestly, it's downright impressive. Uber has managed to get markedly sleazier since they did their "Oh, 'god view' and threatening to stalk reporters who piss us off was naughty; we promise to be good..." charm offensive bullshit.

Given that the store doing the advertising presumably stocks the banned goods, and it's a lot harder to hide physical merchandise in a hurry; I'd assume that the plan depends on the costs of being discovered being less than the value of the advertising, with the cute little trick being there to make it newsworthy, not to fool the cops(even if there were somehow zero machine vision errors or plainclothes cops in Russia; it can't be that uncommon for off-duty cops to wear street clothes; or for the families of police to talk to them).

There is probably a small but nonzero risk that, thanks to the buzz, some humorless enforcer will throw the book at them; but barring that the plan would appear to depend on the actual penalties for 'banned' goods being pretty toothless.

I'm going to need evidence for this one. The USA can "easily" accommodate 320 trillion people with "existing technology"? More than the number of ants on Earth???

Put another way, 1 million times as many people means the entire population of Canada in a single square kilometer. Or 33 people per square metre. I get that you want to build vertically, but we categorically do not have this technology to do this.