What is artificial intelligence? The expression is everywhere these days, but I wonder if the people who use it, including me, really know what they mean by it.
I should start with a caveat. I am not an expert. I can only approach the question as a lay person. But I think the effort to try to understand is worth it, because after all that’s what most of us are. And we can’t pretend that questions like this are not important.
Second, there is a good chance that we will be bamboozled by the advocates and the experts. Some are zealots for AI, some are salespeople, some are paranoid doom-mongers. Any prediction about what AI will be capable of in the future is bound to be to a greater or lesser extent wrong.
I also think it is important to interrogate the origins of the expression. It was the name given to a project at Dartmouth College in the US in 1956 to try to replicate the powers of the human brain. At that moment, it was clearly an aspiration. This was the goal these experts were setting themselves. Nearly seventy years later, I think it is still an aspiration. Different types of AI have extraordinary capabilities, no question, but the aim of replicating human intelligence has not been achieved, which is why advocates now talk about artificial general intelligence as the goal they are attempting to reach. They have had to redefine what they are trying to do.
The idea of simulation is an area of particular difficulty. The great pioneer of computing, Alan Turing, argued that if a person conversing via something like a telex machine couldn’t tell whether they were talking to a person or a machine, that would represent the achievement of what he called machine intelligence. In other words, the appearance of intelligence was enough.
Today, large language models work on the same principle and it is as flawed now as it was then. LLMs use probability to build answers bit by bit, based on what is most likely to be the next word in a sentence. The end result may be extremely plausible, but this is not how humans hold a conversation. So LLMs are machines for generating plausibility, not actual intelligence.
And if LLMs are merely plausibility machines that also means they are not truth machines. When they produce responses which are patently absurd, it is said that they are hallucinating. But that is a misrepresentation. The so-called hallucination is evidence of the underlying logic of how they work. In that sense, everything they produce is a hallucination. The real difficulty is that most of their output will be highly plausible, but not necessarily true.
In short then, the expression, artificial intelligence, is a con. Or at the least a misnomer. And that is not the only misleading aspect of the world of AI. Once it was in the realm of academic research. Now it is at the bleeding edge of global capitalism and geopolitics. The biggest corporations, the most powerful nations are intimately interested in its progress and fully committed to a race in which they believe there can only be one winner.
All this is predicated on the idea that AI will transform the world more fundamentally than anything since the printing press. Or perhaps, according to one hype merchant, since humans learned to use fire. Big talk. Is it justified?
It is not. At least not by what has happened so far. No doubt, efficiencies are being achieved. Some entry-level clerical and admin jobs are being lost. More people whose work can be done by AI systems are at risk. But the bet being placed by companies like OpenAI is enormous. It is investing billions of dollars in vast data processing sites, which will consume huge amounts of electricity and water. In fact the power generation and infrastructure required to supply what is being planned will probably not be in place for years to come.
The market capitalisation of OpenAI and others is not based on their profits. It is based on what investors and the companies themselves think is their potential to be profitable in the future. If the future of AI is practically limitless that might make sense. But is it? We cannot know. No one knows.
Meanwhile the buoyancy of stock markets in the US and elsewhere is underpinned by the boosterism of those trying to make sure that everyone remains convinced that their wagers about the future will pay off. And if that confidence starts to look shaky, economies around the world, particularly the US economy, could suffer a very severe shock.
In sum, I can do no better than to quote the words of Chuck D from the 1980s hip hop group, Public Enemy.
“Don’t believe the hype.”