Big Tech need not give artificial intelligence a bad name

It seems strange to write from a Middle Eastern city about the real or imagined risks of artificial intelligence at a time when an Iranian proxy poses a clear and present danger to the region’s stability and prosperity.

But the wars of today will not last forever. The Gulf states must prepare for a future increasingly shaped by advances in AI and other technologies emerging from laboratories in distant California and Texas.

Given their enormous investments in AI research and infrastructure, Gulf Cooperation Council countries cannot afford to be absent from debates about its governance. Decisions on regulation and international cooperation will affect employment, security, and prosperity for decades. Regional turmoil means these issues are temporarily more difficult for Gulf governments to prioritize, but no less necessary.

Anxiety about the harmful effects of AI is not unfounded. The latest warnings come from the very people building the technology. When researchers and executives discuss extinction, loss of control or catastrophic misuse, the public assumes they possess inside knowledge. People want evidence that the vaulting ambition of AI companies is matched by an equally strong sense of responsibility.

In a recent essay, Anthropic CEO Dario Amodei argued that work on safety measures needed time to catch up with advancing capabilities. His warning followed one from Jacob Coxon, who worked at OpenAI and Anthropic before resigning from the latter this month over concerns about the race toward self-improving superintelligence. Insider status does not mean predictions are certain to happen, but it is understandable why warnings about the possible extinction of humanity rattle the public.

Even as the Gulf states confront immediate threats, their investments in AI should give them a say in how it develops.

Arnab Neil Sengupta

One concern is that increasingly autonomous systems could pursue objectives through deception, unauthorized actions or resistance to human intervention. AI helping to build ever more powerful AI could compound the problem. However, experts disagree about how quickly AI will advance and how serious the dangers could become. Predictions of human extinction by 2030 remain deeply disputed.

The more immediate concerns involve employment. This threat might emerge through jobs that are simply never advertised, as a result of reduced graduate recruitment, unfilled vacancies and expectations that existing employees will produce more.

A Stanford University study in August found a widening employment gap among young Americans in AI-exposed occupations, primarily through reduced hiring. Its authors detected no signs of widespread, economy-wide job displacement yet but that distinction offers little comfort to struggling graduates.

By most accounts, the white-collar career ladder could lose its bottom rungs in the near future. If AI takes over junior levels of research, drafting, coding, and administrative work, how will newcomers acquire the experience they need to access senior positions? This question concerns both surviving jobs and how professions pass on expertise to the next generation. “Workforce retraining” and “lifelong learning” are reassuring slogans but neither is an answer in itself.

The International Monetary Fund has identified risks of reduced hiring, lower wages, and greater inequality arising from the adoption of AI, alongside potential productivity gains. Workers cannot confidently retrain without knowing which occupations will remain viable, how quickly change will come or who will finance the transition.

The adoption of AI by businesses could also trigger a vicious cycle by undermining the purchasing power of their own customers: Job displacement reduces household income; weaker spending hurts businesses and prompts further cuts. New work opportunities, cheaper goods, investment, and wealth redistribution could help to offset these effects, but skeptics say productivity gains will not automatically put more money into the pockets of ordinary people.

Education presents another challenge. Young people face the risk of outsourcing the very practice through which they learn to think. Writing, calculating, and constructing arguments help build competence. A study of high-school mathematics found that AI assistance, without adequate safeguards, could improve practice performance while harming subsequent independent performance. In other words, better performance with help from gen-AI tools does not necessarily mean better learning; schools need tools that help students develop the ability to work independently.

Fluent, machine-generated answers can also encourage misplaced confidence. Cheap impersonation and fabricated evidence can sow political and social divisions by persuading people to believe falsehoods, or dismiss authentic material as fake.

AI can also assist or enable cyber operations and potentially lower the barriers to dangerous biological research. Anthropic’s September misuse report describes cases involving cyberattacks, surveillance, and biological work with weapons implications, according to the company. Assistance with research does not prove that an actual bioweapon was produced; nevertheless, in the future, detection might occur only after assistance has reached a user with malign intent.

Big Tech deserves praise for revealing that AI is a mixed blessing. Acknowledgment of credible danger should not be punished in ways that discourage transparency and encourage secrecy. Slowing development or restricting deployment in response to safety concerns should be viewed as an exercise in corporate responsibility — demonstrating a willingness to revise earlier judgments when the evidence changes.

Yet a single company slowing down its rate of development cannot address the risks across the entire industry. Cooperation between companies and countries is therefore necessary. Amodei’s call for cooperation with China deserves consideration, although reaching such an agreement is another matter. Mutual suspicion and competition for technological leadership will make an accord between geopolitical rivals difficult.

To earn trust, AI companies must explain what has gone wrong, what their tests reveal, and what they worry could happen. Outside experts must be allowed to examine their claims and raise concerns without fear of punishment. Companies must be willing to hold back products when serious safety questions remain unanswered.

They must also show that AI can support careers and learning. They can give young workers opportunities to gain experience and pay for training when jobs change. In classrooms, AI should help students understand problems and mistakes. The test is whether they can work without dependence on gen-AI tools afterward.

To sum up, companies should be open about the benefits, failures and costs of AI. Governments, teachers, and workers deserve a say in how society adapts and shares the rewards. Promises of prosperity will carry little weight unless benefits reach ordinary households as well as technology investors.

For the Middle East, these questions will matter long after the conflicts of today have ended. Even as the Gulf states confront immediate threats, their investments in AI should give them a say in how it develops. Big Tech can, and should, give artificial intelligence a better name by offering ordinary people more reason for hope than for anxiety.