THE TABLET
THE INTERNATIONAL CATHOLIC WEEKLY FOUNDED IN 1840
ARTIFICIAL INTELLIGENCE
AI: SERVANT OR MASTER OF HUMANITY?
might be a reasonable request – such as, for example, for a mortgage – is well known to be infuriating. But AI could magnify it a hundredfold. AI could deem white people as better loan risks than non-white people, or the other way round, for opaque reasons. The technology is baffling not just because most people are not AI-literate, but because it is in its nature to be so. Nobody is marking its homework.
New technologies have been ruffling feathers ever since the first humans invented a new way of lighting fires, thereby jeopardising the livelihood of the local flint-chippers who had hitherto enjoyed a profitable monopoly. Artificial intelligence (AI) is the latest game-changer to disturb settled assumptions, including among some of the 50,000 employees who BP claims could be replaced by artificially intelligent computers. More notable still is the widespread sense of unease about what unbridled AI could do to society – plus a fear that if AI got too smart, it could even out-intelligence human beings. A Big Brother society, but with soulless computers presiding over it. What has jolted the debate up a gear or two is the appearance of an AI product that can mimic human writing to a convincing degree. ChatGPT, first to market, has been followed by Google’s Bard and no doubt there will be many others. Generative AI, the category to which they belong, requires the ability to harvest astronomical amounts of text from sources on the internet, to create a library of raw data it can then analyse at almost incredible speed, to find what it all has in common – in order to imitate it. It is not real intelligence, but it can appear very close to it. To do so it totally ignores copyright and intellectual property rights.
Hence there are many calls for the ethical regulation of AI. But it will be difficult. The main protagonists of AI are north American corporations, likely to be governed by the unconscious ideology of Facebook, Twitter, Google, Microsoft and so on, a geeky masculine culture combined with Silicon Valley libertarianism. It sees human nature in individualistic rather than social terms and it is orientated towards the meritocratic values of the United States – which Americans tend to assume are universal. It truncates and reduces the very concept of intelligence to the processing of digital algorithms, a sequence of logical steps organised in a computer program.
ChatGPT, now available to the general public, can produce a passable version of a sixth-form essay in 20 seconds or less. Not surprisingly, the education world is among those most wary of this new technology. Teachers checking an essay to see if the student has done the necessary research will find the relevant academic sources duly cited. No longer will such students have to delve into textbooks for their material, as ChatGPT will have done it for them already. And organised it in grammatically accurate, if somewhat flat, American-English prose. Job done.
So what’s not to like? The possibilities that AI could enrich human life are credible, but not unprecedented. The appearance of the pocket calculator led to predictions, which turned out to be true, that no one would need to learn long division any more. Pocket calculator technology foreshadowed computers. Perhaps AI could kill essay-writing. As BP is not the only company to notice, vast improvements in productivity could be expected. ChatGPT and its cousins could do anything a handler could do in a call centre. Medical AI could survey all the published literature on a particular condition in a few seconds, speeding up the diagnostic process and suggesting the right questions for the clinician to explore. Maybe forensic AI could help police identify criminals: AI technology can already spot them in a crowd through facial recognition.
It is truly remarkable that this can produce results not easily distinguishable from the product of human intelligence. AI can even write love poems. There is plenty enough data online for it to trawl through and plagiarise from. But would it come up with Eliot’s bleak, “This is the way the world ends, not with a bang but a whimper”? – which seems like an indictment of the whole AI enterprise. Can AI love or hate? Can it despair or hope? Feel
pain or compassion? It can certainly fake them.
What looks like objective analysis is
bound to contain concealed biases from the culture of
the programmer
It is as unlimited in its capacity to influence human society as the twin ideologies of homo economicus and “technological man”, which have produced their own threat to human life through damage to the atmosphere. Even more critically, can it ever know itself, that is to say, can it ever have consciousness? Here is the real fear, an existential one – that a world ruled by AI, maybe not tomorrow but in 20 or 50 years, will not only be soulless but avaricious for its own power. Science fiction has explored this domain, from Mary Shelley’s Frankenstein to Stanley Kubrick’s (and Arthur C. Clarke’s) HAL 9000 in 2001: A Space Odyssey. And found it scary.
The threat to humanity that is posed by AI requires that it is controlled by systems of regulations which will have to be enforceable worldwide if they are to be effective. But regulations must themselves embody a philosophical and anthropological position: that human beings are this, not that. They are not intelligent machines, liable to be surpassed by more intelligent machines.
This last example illustrates some of the dangers. What looks at first like objective analysis is bound to contain concealed biases, the hidden messages built into the culture of the AI programmer. More insidiously still, it will be impossible to scrutinise those biases – and detect serious inaccuracies – as they are buried in millions of lines of inaccessible computer code. The inscrutability of “the computer says No” response to what
This is where the debate has to move to. Throughout history, technological progress in pursuit of profit and power has had to be restrained in the interests of the bigger stakeholder, the human race, of which the primary values are the freedom to live, the freedom to love, and the freedom to worship.
And the name for this is wisdom. These truths are not reducible to a computer program. It is urgent that those who uphold such values should now join the discussion about artificial intelligence and how it can serve humanity – mind, body and soul – rather than undermine it.
2 | THE TABLET | 27 MAY 2023
For more features, news, analysis and comment, visit www.thetablet.co.uk