For once, the dour director-general of the BBC was effusive with his praise. For you, Herr von Ribbentrop, I would gladly fly the swastika from the top of Bush House, John Reith promised the departing German Ambassador in 1938, at a gala BBC event.
A year later, after appeasement collapsed in shame as the tanks rolled into Prague, the BBC’s first boss was still praising Hitler’s “magnificent efficiency”. Reith’s biographer daughter later confirmed her father had revered the Fuhrer. Regarding jazz music as “hot” and a “filthy product of modernity”, the Scot ensured it was banned from the airwaves.
Lord Reith usually got his way, but not every time. If he had, the corporation’s golden age of the 1970s, on which its laurels rest, would never have happened. Reith fought the end of the BBC monopoly with everything he could muster, and was still regretting the creation of ITV years later. “Almost everything I had done has been destroyed – the wrecking of a life’s work,” he sobbed to Parliament in 1962.
In an age when so many historical figures have been cancelled, it’s curious that the first director general of the BBC has so far escaped the mob’s attention. But today, blissfully ignorant of history, self-proclaimed “Reithians” wear their label with pride in their Twitter bios. Like a Waitrose bag for life, this is a signifier of their own good taste and superior moral virtues.
The Reith Lectures continue, too, and the BBC modestly bills this year’s subject, artificial intelligence, as “The Biggest Event In Human History”. Bigger than Alexander, Christ, the Mongol invasion, or the atom bomb? Apparently so.
And the 2021 Reith lecturer, Professor Stuart Russell, ventures even further: “Success would be the biggest event in human history and perhaps the last event in human history.” Readers will spot this is the premise of the movie franchise, The Matrix, and many other pieces of speculative science fiction.
It’s a little trite, but not altogether unfair, to note how the collapse of faith and the end of the Cold War coincided with the rise of fantasies about machines. Internet utopianism was succeeded by wild speculation about the advent of “artificial intelligence”.
Alongside this trend runs a casual belittlement of human achievement amongst our intelligentsia. Contemporary psychology now views us as either as rats, incapable of agency and merely fodder for government prodding and nudging, or lumbering algorithmic vessels for genes.
The dominant political ideology of environmentalism is no better, regarding every human activity as sinful, creating harm to “the planet”. With machines being magical, and humanity a harmful virus, this is fertile ground for wild speculation about artificial intelligence and how much better things might be if they were to take over.
In this context, the childlike, slack-jawed awe that runs through the 2021 Reith lectures becomes understandable, if not excusable. Thanks to the explosion of the public sector technology quangos under the Conservatives, many jobs now depend on these AI fantasies too. Who cares if the story is more complicated?
Russell spends much of his four hours presenting himself as a sober scholar of computing. But he is also smart enough to figure out why he has been chosen for the occasion, and so obliges.
“It’s entirely plausible, and most experts think very likely, that we will have general-purpose AI within either our lifetimes or in the lifetimes of our children,” he claims.
General purpose AI means a machine is conscious, self aware and capable of reasoning with the depth and subtlety of humans. Russell’s assertion has been made many times before. It is not supported by many serious and distinguished practitioners or scholars of AI and neuroscience today. The techniques do not exist, nor is there any promising line of enquiry.
A number of recent and very readable books for the layman describe how AI is falling far short of even much more modest expectations. Deep learning is proving to be a useful addition to the arsenal of data analysis, and does a lovely job at cleaning up old movie clips. But the systems are “brittle”, which means a technique is not transferable to a very similar situation without starting from scratch.
In the field of robotics, cheap sensors and communications are proving to be far more useful, and are much more likely to give us those elusive productivity gains that the economy needs. If AI plays a part at all, it is often a redundant luxury.
Perhaps speculating about machines is part of human nature. The term “artificial intelligence” itself was a marketing concoction, created in the 1950s by an MIT Professor, John McCarthy, in order to obtain funding for his obscure and shoestring journal on mathematical automata.
Once he rebranded, the money flowed in. He was transformed into an academic celebrity. Similar fame beckons today. So the Reith Lectures this year strive not to tell us the full story – and in that, they succeed.
The von Ribbentrop story above is my favourite BBC anecdote and can be found in Tim Bouverie’s splendid history, Appeasing Hitler. Bouverie’s book is not only a story of misplaced pacifism, but is also a study in groupthink that possessed the chattering classes, and indeed a nation, for many years. “Appeasing AI” would make an excellent companion.
Andrew Orlowski tweets at @andreworlowski
