The end game in chess comes when there are few pieces left on the board, and checkmate—the final blow—is nigh. It’s a point in the game where, often times, there’s still technically moves to be played, but the game is already won. And when it comes to capitalism as we know it, it now seems that we may very well be in the end game stage. That is, it’s possible that although there are still moves to be played, the game is already won. Or at least it may be, definitively, as soon as one of the world’s technology giants cracks the ‘iPhone’ of robots and creates an all-purpose generally intelligent machine that can learn and master any task you throw at it.
The constant chatter about the issues surrounding automating the workforce crescendoed earlier this year with Elon Musk’s remarks regarding AI at the National Governors Association (NGA) Conference in Rhode Island. He noted that with AI “there will certainly be a lot of job disruption” and that “robots will be able to do everything better than us...” He added that when it comes to developing AI, companies are “racing [because] they kind of have to race.” And without government regulation, he made it clear that tech companies are going to go full steam ahead with AI R&D—which is already happening at an exponential rate.
Although it seems as if there’s a monstrous mishmash of robots already at our doorstep—sometimes literally, in the case of the Domino’s Pizza Delivery Robot—it’s likely that all these robots are, in the grand scheme of the evolution of AI, gimmicky. They’re narrow AI: capable of a specific task, within specific parameters. But the hypothetical iPhone of robots will be generally intelligent. It’ll be able to make your breakfast in the morning, play catch with you in the afternoon, and go off to its factory job at night. Exactly like Robbie from I, Robot. Or any of the million other machines in sci-fi who are generally intelligent.
Although this iPhone of robots will doubtlessly require hugely complex innovations in machine intelligence, the two core components of its existence are already being developed: neural networks that can learn how to execute a given task, and some type of physical machine that can manipulate an infinite assortment of real objects.
The most striking recent achievement in the development of Artificial General Intelligence (AGI) came last year when the neural networks developed by Google-owned DeepMind defeated one of the world’s best Go players. Its neural networks learned to play Go, then beat the world’s best. DeepMind subsequently announced that it wants to have its AGI learn to play, and master, StarCraft II next, and the company has explicitly laid out plans for ultimately applying its AGI to real-world tasks. (It seems that as long as something can be turned into a game, a machine can learn how to play it and master it; kind of like WOPR with nuclear war in War Games.)
As far as the real-world object manipulation component, we’ve all become familiar with Atlas from Boston Dynamics, a formerly Google-owned company that was sold off to Japan’s SoftBank (yet another tech giant). Atlas has demonstrated an incredible flexibility with manipulating real objects and moving through a real environment, and it’s a small logical leap to imagine a combination of DeepMind’s neural networks with a robot like Atlas. In fact, we get a little taste of just that in the OpenAI video below, where we see a clear demonstration of the two components coming together in the nascent stages of a “general purpose robotic system.”
When trying to picture a future in which robots can do literally every job there is—as Musk predicts—it may be reasonable to imagine our society suddenly populated with a single mass-produced robot model capable of learning, mastering, and obtaining superhuman proficiency in any given task, on blindingly fast time scales. This ultimately seems just as likely, if not more likely, than a motley assortment of robots each built specifically to vacuum your floor or deliver a package or act as the lamest security guard in history. This is because a machine that learns its task is necessarily flexible, and it can adapt and improve in response to its environment.
— Nige Willson (@nigewillson) July 22, 2017
All this means that from now on, whenever you tune into that constant background chatter about automation in the workforce, keep in mind the possibility of the sudden emergence of a single, generally intelligent robot. A single endlessly pliable and highly efficient employee-replacement that could make the necessity of work, and the economic model of capitalism, a thing of the past in the blink of an eye.
What comes next, how we plan to organize society once machines become our equals (at least) in terms of intelligence and productivity, is a question that we’ll have to seriously ponder with our simple wetware brains for now. Musk has already stated that he believes some type of universal basic income (UBI) will be necessary, but it seems that in the face of generally intelligent robots that can take literally any job (even CEO of Tesla and SpaceX, apparently), that may not be enough to stave off the need for some entirely new way of running an economy.
What do you think about the idea of a single robot being intelligent and pliable enough todo any job on the planet? Do you think this is how automation in the workforce will unfold? Let us know your thoughts below!
[brightcove video_id=”5522410273001″ brightcove_account_id=”3653334524001″ brightcove_player_id=”rJs2ZD8x”]