Not being an artificial intelligence “expert”, an egghead, or a mad scientist, I can sometimes find the trajectory of AI confusing. In my reading, I’ve come across evidence to convince me in entirely contradictory directions, and there is certainly a lot of tension in the public conversation.
Are we in the run-up to an AI apocalypse? A technological utopia? Or, even with all our computational power, is AI still as dumb as a rock?
Despite the many “ifs”, “buts” and “maybes”, the most compelling arguments I’ve read are those written by individuals much cleverer than I, who do an excellent job at busting some of the myths so beloved by the media. Many (but not all) of those included below are attributable to this fascinating article by Professor Luciano Floridi of Oxford University. I thought I would distill and share for those with less time on their hands.
Five (arguable) AI misnomers:
- Superintelligence is coming
Though some commentators believe that “strong” AI or “superintelligence” is right around the corner, others have convincingly argued that there’s no reason to believe anything resembling truly intelligent machines will emerge from our current and foreseeable understanding of computer science and digital technologies. AI is yet to even pass the Turing Test, devised back in 1950, and we still have little idea about how our own brains and intelligence work. There is a long way to go before AI matches our kind of “general” intelligence, let alone surpasses it.
- But systems continue to improve all the time…
This is true. Nevertheless, exponential growth now doesn’t mean exponential growth ad infinitum. Ray Kurzweil famously argues that the law of accelerating returns means that machines will pass the Turing Test 2029 (and then presumably accelerate beyond it). Others (like Floridi) are more skeptical, and point out that growth can – and often does – slow or plateau. The fact is, there are no guarantees on the rate or consistency of the AI advance. The past isn’t always the best reference point for the future.
- If true AI happens, it will present a threat
Elon Musk has previously warned that we are “summoning the devil”, and Hollywood has taught us that sophisticated AI is likely to be hostile. Such skepticism and fear often accompany progress and complex change, but even experts who believe superintelligence is inevitable rarely voice concerns about our destruction. On the contrary, many believe that AI will be intelligently adapted in such a way as to understand, protect and preserve humanity and its surroundings. Others have even speculated that sophisticated AI may turn its attention to frontiers to which it may be better suited than humans, like space.
- We humans need to stay “top dog”
We’re probably too late. Even without superintelligence. As Floridi elegantly puts it, having learned we aren’t the center of the universe (Copernicus), the biological kingdom (Darwin), or of rationality (Freud), post-Turing’s achievements it is already true that we are no longer the center of the world of information processing and agency. There is AI that can outperform us on scores of previously human-centric tasks. We must graciously learn to live alongside intelligent machines, and use this time of development to set them on a course which leads them to solve, rather than shore up, future problems. But we should no longer expect to always be the smartest and the best.
- Superintelligent or not, AI is the writing on the wall for work
There’s no doubt that the development of AI will lead to the automation of many roles, but that needn’t spell disaster for those of us dependent upon work to earn money. Firstly, entire new domains are likely to emerge from an economy bolstered by AI, and these will generate jobs that cannot easily be automated. Secondly, the thoroughness of AI might well create the need for greater human intervention: for example, AI that audit may identify many more cases that need complex (i.e. human) reflection. Lastly, as with most technological innovations from the plow onwards, AI (at least in the near future) will most obviously be used to augment the existing workforce, making many tasks less laborious.
Sensationalism is usually more beguiling than pragmatism, and though it’s fun to speculate about what may well be some extraordinary changes afoot, we should keep our heads. Not least because within them we still have the most precious, conscious, and complex intelligence of them all. At least for now.