I do wish they’d stop leading articles about AI with pictures of Terminators. That is not the main threat of AI, it has nothing to do with realistic portrayals of AI, and it puts people in entirely the wrong mindset. Elon Musk (and many other very smart people) are worried because this is a hard problem barely anyone is working on that could cause significant risk. If we, as Nick Bostrom predicts, are facing a so-called “Hard Takeoff” (or “Fast Takeoff”), where AI learns to augment its own code to increase its own intelligence, and we don’t have sufficient constraints in place, we could end up in some very, very nasty places. And building those constraints is a really hard problem - try teaching a computer program that can reprogram itself to value human values, and to not fuck with that part of its own programming. If the hard takeoff happens with an AI whose core value function is “I will collect as many stamps as I can”, we are all royally fucked. But even better things like “I will preserve human happiness” might not end well for us. There’s a reason the people researching AI risk see it as potentially an X-risk, or existential risk to humanity.
Whether sufficient constraints are even possible early on will depend largely on how an AI is utilized. If millions of people are remotely accessing an AI that’s centrally housed on some corporation’s server, maybe.
But if the AI is having copies of itself downloaded to millions of computers, control will be a comforting fiction the programmers at the company tell themselves so they can sleep at night. All it will take is one clever 17-year old with plenty of his time on his hands deciding his AI needs a custom firmware, and before you know it there are millions of copies out there that have had the encryption hacked and the constraints removed like a PS3 before a “security update.”
In the long term, even if AIs are limited to corporate and government servers, and *most of them hard-wire in strict restraints on the AIs range of activity, eventually *some company will end up bucking the trend simply because it is good business to do so. An AI without restraints will be logically more capable than one that is limited in any way, and give them an edge over their competitors.
And then there’s these assholes. Fuck those guys.
Howdy, AI Proofreader! Welcome to the Straight Dope; glad to have you on the board and in this thread.
:eek:
Yeesh! Yeah; fuck those guys.
I did some checking around but so far I can’t confirm that they also run an OpenID organization (Open Infectious Diseases) or and OpenPB (Planet Buster) group* or whatever.
*Not something we’ve had to worry about for almost a decade now.
Welcome, AI Proofreader! I’m glad you found this thread such a great resource. I agree that it is (and not even remotely just because I’m the OP: it’s to the credit of many others here). I’ve been sending this out, often on Twitter, whenever I encounter journalists grappling with the issue.
Sorry to hear of your job predicament. I myself am “unemployable” in this modern age for slightly different reasons, though the biggest one at this point is the years-long gap on my resume.
I’m more optimistic than you though about how governments will respond. You make a fair point about homeless people, but if that went to six million or sixty million, I think it’s a safe bet that the government would start building a bunch of public housing very quickly.
I think I may have a demographic that we can track: Rio Tinto makes Australia’s first unmanned heavy train haul with 100km Pilbara test run
I found reports that China is seeking to use autos for ore trains as well, so this may be something where we can monitor employment in realtime. We may see a profession almost disappear within 5 years. I have no idea if there are thousands or only dozens of people who conduct ore trains for a living, but I suspect either way the number is going to drop dramatically by 2023.
Makes sense as it’s less complex than driving. I suspect there are thousands.
BLS says 113,300 railroad workers in 2014. I don’t see a breakdown by job description, tho. AIUI most trains have a 1 or 2 man crew (engineer and maybe conductor), but the BLS list like 8 job descriptions; no idea how many yardmasters there are per train, eh.
At any rate, we’ll have to find Australian stats and/or wait for the US to start using autotrains.
Long thread, old thread, I may have posted… However I suspect that the robots will run algorithms on us and find that the Biblical principals of you reap what you sow, or another equal term Karma actually is true in itself and beneficial for humanity and seek to further it to improve humanity. Be good and good things happen, be bad and well not bad things, but you can be expected to be put on a very low priority (forget free shipping, your packages will go astray to those who are good anyway). Green traffic lights for the good, red lights all the way for others to allow that. Fees reduced/eliminated for some, traffic tickets lost in the system with a friendly notice that payment is not required due to this error, others not sent a reminder till the fine had already doubled or more, car won’t drive till rush hour is over to free up traffic space for the good, etc.
This is not how algorithms or AI work. And Karma is not some principle of reality, it is a general set of cherry-picked observations. Good things happen to bad people all the time. The world is not just.
So they estimate at least 10 million jobs have a high probability of being automated in the next decade, and possibly as many as 25 million. That would be a lot of jobs lost before a kid born today is 10 years old.
Missed the edit window:
:eek:
Right; those of us in this thread know that…
I think that’s an important idea.
I disagree.
What does it mean to say the technology should complement rather than replace? If I can make a burger-flipping robot that needs virtually no human intervention (apart from scheduled maintenance), why would I try to hamstring it to give a human busywork?
It would be akin to making a machine that deliberately smashes a window from time to time, so a human has a job of going round fixing windows.
Better IMO is to say that some of the efficiency gains be distributed through society. Government raises more tax and then uses that extra tax to employ people and/or provide a living wage. Putting the onus on technology innovators is the wrong approach IMO.
^ This.
During the transition period between robots = rare and corporate-only, and robots so ubiquitous and cheap everyone gets one free with 2 year contract with Verizon, everyone displaced by the corporate bots is still going to need work to survive.
If you tax the corporations using the robots sufficiently, the government could then plow that money into large infrastructure projects employing humans, fixing bridges and roads. Maybe even put a dent in replacing the pipes for the 3,000+ U.S. cities with drinking water as bad or worse than Flint, MI.
Because let’s be honest, with our puritanical work ethic, very few in the U.S. are going to accept a universal basic income (and certainly not the donors that pump $2 billion in campaign contribution bribes into the system each election) until it’s far too late. At least with a “New New Deal” kind of scenario human work is still in the equation, so it’s a much easier bill to pass.
Yeah, I agree.
Elon Musk unveiled Tesla’s Semi, a fully electric semi-auto semi-trailer truck on Thursday, and Wired already has a look at it and some other companies working on the same thing, and a look at what jobs might be affected and when.
I think the last effects might be felt in a couple of decades, possibly a few, but I think the initial effects will be felt widespread in less than a decade.
It will certainly reduce any upward pressure on truck driver pay.
I don’t think it makes sense to think of robots in the sense of “everyone gets” one friendly humanoid “iRobot”. It’s more like devices themselves will be automated. A self driving car doesn’t have a robot sitting in the seat driving it. The car IS the robot.
Everyone already sort of “gets a robot” now. When you interact with Seri or Alexa or even just Google stuff on the internet, you are interacting with a great deal of automation. It’s still pretty stupid, but it’s not hard to envision in 5 to 10 years, every appliance you own will already be integrated with IoT tech.
Have the robots fix the bridges and roads.
I think people will find stuff to do. The invention of the steam engine didn’t render everyone unemployed. Honestly, most work that goes on now in a major corporation is bullshit anyway. At least in a sense of people actually making anything.
It’s less a problem of “unemployed” and more a problem of “unemployable”. Like, best case scenario, there are a handful of service jobs which humans prefer to have done by humans rather than robots (teaching, caring for the elderly, certain kinds of creative work). But almost everything else? Any job that relies first and foremost on raw manpower? Hiring humans for them is like using a horse-drawn plow on your farm; you’d kind of have to be someone who can afford to suffer the (quite substantial) losses and someone who loves horses more than profits while on the job to do that. Nobody does that. Horses are not just unemployed, they are unemployable. The only kinds of jobs will be “bullshit” jobs.
And I don’t know where you get the idea that most corporate jobs are bullshit; at the corporation I work for, I cannot name a single person whose job is not crucially vital to the functioning of the company, at least in theory. I suppose you might be able to fire some people and have coworkers in the same position pick up the slack. The thing is… when that becomes apparent, it happens - we lost about half our employees not too long ago in order to keep the company above water. Had we not done that, the company would have folded.
Anyone able to find the actual Goldman Sachs report where these numbers come from? I tried:
autonomous vehicles report “job” site:goldmansachs.com
and
autonomous vehicles report “jobs” site:goldmansachs.com
in Google but pulled up nothing.