AI's singular uncertainty reflects scope, speed, and unpredictability

It is hard to think of any historical parallel for the technological revolution promised by the growth of artificial intelligence (AI). In both its speed and its possible scope, AI appears unprecedented. 

The real issue, however, is that few technologies of the past created such uncertainty. It is true that nobody could predict what the internet would do to society. But the variation in people’s expectations of AI’s eventual effects is particularly wide. Here are four areas in which predictions of the impact of AI have shown this stunning divergence.

  Global inequity: Many previous technologies have helped level the playing field between developing and developed nations. The information-technology revolution and the internet helped those in India, for example, compete on some services with companies and people in the West. The economic surplus created by the new technology — for consumers or producers — was shared between countries.

  We don’t know if the same will be true of AI. One view is that its spread will enable greater productivity for workers everywhere, that it will minimise problems caused by linguistic differences, and allow for more efficient provision of government services in capacity-starved developing countries. This might have an equalising effect on welfare across nations. But others point out that there is an embedded first-mover advantage to AI greater than demonstrated by previous technologies, which were more easily replicable. Those who can possess the most computing power, who can train their algorithms earliest, may retain their lead indefinitely. The patents and proprietary information that go into most AI models are controlled by pools of finance concentrated in only a few geographies, and they will benefit disproportionately.

  In other words, we can’t be sure if AI will reduce the welfare gap between developing and developed countries, or greatly increase it.

  Employment: All technologies, when introduced into industry, both create and destroy jobs. But with AI the predictions are particularly stark. Dario Amodei, chief executive officer of Anthropic AI, told Axios this week that his best estimate is that AI will “wipe out half of all entry-level white-collar jobs, and spike unemployment to 20 per cent in the next one to five years”. He said governments should prepare for the “mass elimination of jobs across technology, finance, law, consulting and other professions”. 

  Others claim, however, that the long-term effect is likely to be positive in terms of productivity gains, which will then open up new roles and raise wages all round. Most employers don’t agree with Mr Amodei. One World Economic Forum survey-based report predicts that AI will cause 92 million jobs to be wound up, but create 178 million new ones by 2030. Their survey of 1,000 large global employers found that more than half expected to increase their salary bill by 2030, while only 8 per cent expected the salary share of revenue to decline.

  Creativity: What will AI do to human creativity? Will we be more or less able to express ourselves once we have access to better and better AI engines? Few things are more divisive than this. Across social networks, AI-generated visual “art” has become ubiquitous. Most small businesses now rely on free AI image generation for basic mailers and graphic design. The creation of images — and, increasingly, music and lyrics — is only dependent on the quality and inventiveness of the prompts given to the AI, not the inherent talent and skill of the human creator.

  Human artists and designers, meanwhile, point out that this is not generative, but regenerative — an unimaginative remix of whatever real work by real artists an algorithm finds online. In the absence of new artistic feedstock, derivative AI-assisted artwork will wind up being an endless circle of replication. Only humans can create novelty, and if they are edged out of the creative process, creativity will grind to a halt for the first time in history.

  Human survival: The industrial revolution, based as it was on fossil fuels, eventually led to global warming with its existential risk for many human communities. AI’s risks are, according to some, more immediate and even higher. Even two years ago, some in Silicon Valley were warning AI-related risks were startlingly high. The Center for AI Safety said in June 2023 that “mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks such as pandemics and nuclear war”. Others in the field scoff: One senior MIT researcher said that there is “no more evidence now than there was in 1950” that AI would pose an existential risk to the human race. It’s hard to find a starker difference than this: One group involved with the technology thinks it may well kill us all, and another group thinks this might make us immortal.