The four big risks from artificial intelligence

Throughout history, technological breakthroughs have created new opportunities for invention, adaptation and progress while inflicting irreversible damage on many lives and livelihoods. They have tested the remarkable abilities of human beings and societies to adapt to the turmoil of transition and to survive what economists call “creative destruction.”

The world must now prepare for a technological breakthrough whose implications are vast and which are unfolding at a speed that has frightened even the men and women who have spent their working lives preparing the ground for this upheaval. Artificial intelligence will transform our lives – for better and for worse – so thoroughly and so quickly that we have no choice but to prepare ourselves and one another for the fallout.

Without question, there will be medical and scientific breakthroughs that transform the labor of decades into the work of days. Those with access to the most powerful AI tools will have an opportunity to live longer, healthier and more prosperous lives than human beings have ever experienced.

But there are also risks that we must think through and prepare for. The four most consequential of these involve disinformation, proliferation, displacement and replacement.

Disinformation

Without continuous access to accurate, verifiable information for citizens, consumers and investors, there can be no democracy or free market capitalism. The advent of social media and the tidal waves of distorted information it generates has already poisoned public attitudes toward institutions of all descriptions. The mainstreaming of AI will add a vast chorus of pre-programmed non-human voices to the conversations that shape political life in every country in the world.

The ease with which malicious political actors, criminals and terrorists can create video illusions that fool even the most sophisticated viewer will make it far harder for political leaders and those who report the news to build and sustain credibility. China, Russia and other authoritarian states will develop more effective forms of digital propaganda that undermine freedom in profound and unprecedented ways, and they will sell these technologies to any government willing to pay for them.

Proliferation

In recent years, the technology problem that has most preoccupied political debate within democracies is data collection from online activity and its impact on privacy. But artificial intelligence is a democratized technology. The powerful tech companies that have come to dominate our online lives can set rules and guidelines for the use of the products they create. To some extent, they can enforce those rules. Still, AI models that are nearly as advanced – and more powerful than the algorithms in general use even a few months ago – are already available to anyone with marginal programming skills and a laptop computer.

I personally know a number of people who are now running their own large language models that use publicly available information to produce large amounts of text. In a field with an open-source culture and very few barriers to entry, that availability will spread far and wide quickly and easily. Millions of people will soon have their own GPT running on real-time data available on the internet.

It’s a powerful tool that individuals can use to create useful new things that break new scientific and artistic ground. It’s also a weapon that rogue political actors, criminals and terrorists can use to code malware, create bioweapons, manipulate markets and poison public opinion. It’s true that authorities can deploy AI to create more effective tools to police these crimes, but governments have never faced a threat this diffuse.

Displacement

We know the explosion of artificial intelligence will displace untold numbers of workers as machines replace people, even in knowledge sectors, on a scale that most of us until recently thought impossible. It’s true that we have seen these upheavals before. Most recently, the surge in global trade of recent decades killed millions of manufacturing jobs in countries where workers earned relatively higher wages as factories moved to developing countries. Then, automation displaced manufacturing jobs more broadly.

In both cases, these tech disruptions yielded much higher productivity and wealth globally, and they eventually created more jobs than they destroyed. But it takes time and resources to retrain workers and to establish sustainable social safety-net protections for those who can’t adapt. And the displacement triggered by the expansion of artificial intelligence will hit more workers in more places much more quickly than in any workplace disruption the world has seen before. This workplace revolution will create economic and political turmoil on a scale that national governments and multinational institutions are not prepared to manage.

Replacement

Finally, there is the most personal aspect of this revolution. Humans will soon become much more accustomed to direct communication with machines. Instead of turning to simple bots for weather reports, we’ll rely on complex AI-driven machines for complex interaction and even companionship. We already know that excessive social media use can produce anxiety, depression and even self-harm for teenagers and isolated adults. This problem is about to become much larger as more people with anti-social tendencies build relations with increasingly sophisticated machines. This is AI’s most profound challenge, and it’s the one that policymakers are least prepared to meet.

Nothing separates us from these risks except easily solvable technical obstacles and time. Each of them will have to be addressed within families and communities, among public- and private-sector decisionmakers and across borders – and the revolution has already begun.

*      *      *

Ian Bremmer is the president of Eurasia Group and GZERO Media and author The Power of Crisis.