The Real Problems With Artificial Intelligence

a collage of some of my favourite Midjourney AI images, none of them my prompts; to view larger size, right click and open in a new window/tab and then simultaneously press CMD and + to enlarge (CTRL + in Windows)

My thoughts on AI and on Artificial General Intelligence (AGI) are evolving as I use publicly-available AI apps more and more, and see how they are being employed.

My sense at this point is that AI/AGI is neither a new problem nor a solution to anything. The actual problem is humans’ propensity to misuse technologies, usually with the best of intentions. AI/AGI is just another tool that neoliberals can use to advance their vision, militarists can use to advance their vision, and technotopians can use to advance their vision. None of that is of any value in dealing with the polycrisis predicament at hand, which, like all predicaments, is insoluble. And in playing with these new toys we are likely to make a lot of messes and cause a lot of damage, as we have done with essentially every new technology we have ever invented.

Trying to ban or ‘freeze’ development of AI/AGI is, I think, akin to, and as futile as, banning or ‘freezing’ the development of arrowheads, or cars, or the printing press, or letter openers, or any other kind of technology throughout our history on the basis that it could easily be misappropriated (accidentally or deliberately) to destructive ends. As John Gray put it in Straw Dogs (before AI was a thing):

If anything about the present century is certain, it is that the power conferred on ‘humanity’ by new technologies will be used to commit atrocious crimes against it. If it becomes possible to clone human beings, soldiers will be bred in whom normal human emotions are stunted or absent. Genetic engineering may enable centuries-old diseases to be eradicated. At the same time, it is likely to be the technology of choice in future genocides. Those who ignore the destructive potential of new technologies can only do so because they ignore history. Pogroms are as old as Christendom; but without railways, the telegraph and poison gas there could have been no Holocaust. There have always been tyrannies, but without modern means of transport and communication, Stalin and Mao could not have built their gulags. Humanity’s worst crimes were made possible only by modern technology.

My skepticism about the use of AI/AGI as a vehicle for problem-solving is that AI/AGI is inherently devoid of the capacity for imagination. Its most interesting ‘work’ happens when it uses its clever data crunching capabilities to barf out random concatenations, like ChatGPT’s poetry or the sometimes-stunning images that come from Midjourney’s misunderstandings of (mostly badly-worded) prompts. The genius of randomness. Its most compelling outputs are largely accidental.

None of what it produces is really art, but some of it could well inspire art, by provoking our rusty human imaginations to think in ways or about things we hadn’t thought about before. But that’s mostly dumb luck when it happens. AI/AGI will never be imaginative because it is intrinsically incapable of metaphorical, lateral, inductive or abductive thinking — it can never acquire the vast rich human, uncategorizable slurry of content-in-context that would be needed to enable such thinking, and in any case these ways of thinking are non-analytical processes that are not strictly intellectual and cannot be programmed. Only in human-written sci-fi will AI/AGI be able to look at the pigment-free colouring in a butterfly’s wing and ‘independently’ imagine how that ‘technology’ might be commercially applied to aeronautical coatings or noncounterfeitable banknotes.

Living in an age of staggering imaginative poverty at exactly the time when imagination is most desperately needed to help us cope with the polycrisis, we are inevitably going to be disappointed with the inherently stale, derivative, clichéd and prevailing-narrative-reinforcing ‘intelligence’ that AI/AGI comes up with.

Since AI/AGI can only ever do what it’s told to do (by humans or by other AI/AGI bots), its use for anything other than mundane commercial and military applications (and misapplications) is inevitably going to be limited. It might precipitate the end the world (most likely by military or geoengineering accident), but it will never produce anything genuinely novel. That is the difference between creativity and imagination.

I think our impoverished imaginations are mostly a result of lack of practice. I used to invent games, conjure up imaginary friends, daydream about going into other dimensions etc. All of that is done for us now, constraining our imaginations to what Hollywood and the gaming companies can manage with CGI, and the hackneyed, trite, warmed-over myths that they reinforce.

When I look at the Midjourney ‘showcase‘ of most-upvoted images, it is kind of depressing. Anything in the world that you can imagine could theoretically be constructed and displayed from the prompts, but 99% of what is presented looks like posters or cels from Hollywood cartoons, comic books, violent action films, sci-fi and horror movies, or disturbing incel fantasies. Part of that is that the Midjourney AI can’t imagine, but most of it is due to the fact the prompters can’t imagine either.

So, yes, I’m worried about how humans will continue to abuse new technologies for nefarious purposes, such as producing fake videos indistinguishable from real recordings, to the point we will have to be skeptical of everything we see on our screens (if we aren’t already). And I know this is a slippery “guns don’t kill people…” argument (though some technologies like weapons are basically Moloch Tragedy technologies, and the less use we make of them the better).

But I’m far more concerned about how, for example, we’re using new kinds of underwater explosives, guided missiles and drones to ‘anonymously’ assassinate people we don’t like, and to blow up pipelines, dams and potentially nuclear power plants, creating political havoc, social and ecological disaster, and accelerating the risk of nuclear war.

In the meantime, AI has its uses, and I look forward to seeing continuous improvements in its very useful capacities for information-gathering and synthesis, and for increasingly high-quality image production. For example, I now have ChatGPT installed on my Google search page, and its responses to my searches, which appear beside Google’s, are so superior to Google’s that I only bother to look at the Google results now when I’m asking about something that happened recently (ChatGPT’s knowledgebase only runs up to September 2021). It’s that much better.

A caveat, though: I must admit that it’s required me to up my game in learning how to word and phrase my chat/search queries, without which it’s often just a GIGO exercise. It took me years to learn how to use the Google search bar effectively. And now I’m back to square one with the chat box. It’s like a conversation with someone you don’t know — ChatGPT and I learn from and teach each other how we communicate and understand, and only when we’ve got that understanding down can we start to craft sentences we know will be understood by the other.

And these days, instead of using Creative Commons licensed images (which have been absolute lifesavers for unpaid writers like me for the last two decades) on my blog posts, I’m now using mostly Midjourney-produced images. No more worries about copyright, and I have far more control over the types of images I can produce. And it’s a lot more fun.

Maybe I’ll be more concerned about the evolution of artificial intelligence if and when it becomes, um… intelligent. Y’know, like, not just processing data (often suspect data at that) really quickly, but actually coming up with something useful for addressing and coping with some of the challenges of our time. So many of our recent ‘smart’ technologies are focused on creating new (largely artificial) ‘needs’ (and doing so strictly to make a profit). It would be nice to have some that actually addressed some real existing needs instead.

But that can’t and won’t happen until we shake the false computer-as-brain metaphor and start to understand how nature and its creations actually adapt to changes in the environment in ways that enable them to survive and thrive. That entails far, far more than mere ‘intelligence’. Invented technologies let you do the same old things faster/cheaper/better etc. But evolution lets you do new things.

It’s taken the natural world several billion years to evolve that astonishing capacity. Small wonder our bewildered, bumbling species is still at the starting gate. Still playing with fire, and still not cleaning up after our messes. And still, and more than ever, unable to imagine.