In his first of his four (impressive) Reith lectures on living with artificial intelligence, Prof. Stuart Russell of the University of California at Berkeley began with an excerpt from an essay written by Alan Turing in 1950. The title was Computing Machinery and Intelligence and in it Turing introduced many of the core ideas of the later academic discipline of artificial intelligence (AI), including the sensation du jour of our time, so-called machine learning.
From this amazing text Russell drew a dramatic quote: “Once the machine-thinking method had begun, it would not take long to outstrip our weak powers. So at some point we have to expect the machines to take control. “This thought was formulated more emphatically by IJ Good, one of Turing’s colleagues at Bletchley Park:” The first ultra-intelligent machine is the last invention that man will ever have to make, provided the machine is docile enough to tell us how to maintain it. “Control.”
Russell was an inspired choice to give a talk on AI as he is a world leader in the field (z The current approach to building “intelligent” machines is deeply dangerous because he holds the prevailing concept of intelligence – the extent to which actions to achieve certain goals can be expected – for fatal mistakes.
One expert joked that he was just as concerned about super-intelligent machines as he was about overpopulation on Mars
AI researchers build machines, give them specific concrete goals and judge them as more or less intelligent based on their success in achieving these goals. That’s probably fine in the lab. But, says Russell, “When we start to get out of the lab and go into the real world, we find that we cannot fully and properly specify these goals. In fact, it has proven extremely difficult to define the other goals of self-driving cars, such as the balance between speed, passenger safety, sheep safety, legality, comfort and courtesy. “
That is politely put, but it doesn’t seem to bother the giant tech companies that are driving the development of ever more powerful, relentless, purposeful machines and their ubiquitous installation at critical points in human society.
This is the dystopian nightmare that Russell fears if his discipline continues on its previous path and succeeds in creating super-intelligent machines. It is the scenario that is implied in the thought experiment “Paperclip Apocalypse” by the philosopher Nick Bostrom and is entertainingly simulated in the computer game Universal Paperclips. It is of course also warmly derided by the tech industry and AI researchers as implausible and alarming. An expert in the field famously joked that he was just as concerned about super-intelligent machines as he was about overpopulation on Mars.
But for anyone who thinks that life in a world dominated by super-intelligent machines is a “not in my life” perspective, here is a salutary thought: We already live in such a world! The AIs concerned are called corporations. They are definitely super intelligent as the collective IQ of the people they employ eclipses that of ordinary people and even often governments. They have immense wealth and resources. Their lifespan far exceeds that of humans. And they exist to achieve an overarching goal: to increase shareholder value and thereby maximize it. To achieve this, they will tirelessly do everything regardless of ethical considerations, collateral damage to society, democracy or the planet.
One such super-intelligent machine is called Facebook. And to illustrate that last point, here is a clear statement of its overall purpose, written by one of its chief executives, Andrew Bosworth, on June 18, 2016: “We connect people. Period. Therefore, all of our growth work is justified. All of the questionable contact import practices. All the subtle language that helps people be found by friends. All the work we have to do to bring in more communication. The work we will probably have to do in China one day. All of it. “
As William Gibson famously noted, the future is already here – it just isn’t evenly distributed.
What i have read
Choose a side
There Is no “Them” is entertaining online talk by Antonio García Martínez against the “othering” of West Coast tech billionaires by the East Coast elites.
Vote of confidence?
Can Big Tech Serve Democracy? is a great review article in the Boston Review by Henry Farrell and Glen Weyl on technology and the fate of democracy.
Follow the rules
What parking tickets teach us about corruption is a nice post by Tim Harford on his blog.
