Timnit Gebru says harmful AI systems need to be stopped

The field of artificial intelligence (ai) cycles through what are called AI summers, epochs where every other news headline seems to be about AI and there is ample funding for the field, and AI winters, which come from the disappointment of undelivered overpromises during the summers. We are currently in perhaps the most intense AI summer ever, where just the mere mention of “AI” gets startups 15-50% more funds in investment.

But just like past summers, even the current hype cycle is an “AI summer” only for those profiting from building these systems or the researchers who get funding to work on the dominant paradigm of the day. For many people in the AI pipeline—from the exploited workers supplying and labelling data that power these systems and the content moderators who filter out toxic content, to the marginalised groups who live in apartheid states being overpoliced because of ai—it is a nightmare that shows no signs of abating.

A number of countries around the world are scrambling to propose regulation pertaining to AI, and some have passed laws. Many are feeling the pressure to act because of the current fascination with AI and daily headlines about the utopia that its boosters promise, or the doom that, some predict, it will bring to humanity.

Groups parroting unfounded claims about the impending AI utopia or apocalypse have brainwashed students at some of the prestigious universities that supply Silicon Valley’s engineers and scientists, and have influenced multilateral organisations and governments. One group they have not influenced so far, however, is the labour movement.

Those working on the repetitive task of providing examples to train or evaluate systems like ChatGPT or Dall-E do not expect an all-knowing machine on the horizon. They clearly see how hiding the extent to which their labour powers these systems helps multinational corporations sell the supposed power of their technology, while exploiting millions of people around the world. These workers are organising to improve their working conditions and curb the development and deployment of harmful AI systems.

For example, in 2023 Kenyan workers employed by third-party outsourcing companies for the likes of Meta, OpenAI and ByteDance, established the first African Content Moderators Union, and one of them sued Sama, an outsourcing company, for union-busting. As noted by Adrienne Williams, a former Amazon delivery driver who campaigned for better working conditions, the less labour that companies are able to exploit, the less they are able to develop harmful AI systems, because it would not be profitable to do so.

Higher-paid tech workers are also organising with their lower-paid counterparts, not only to advocate for better working conditions but also to stop their organisations from developing harmful AI systems. From Google workers protesting against the company’s involvement in developing computer-vision technology for drone warfare in partnership with the American government, to the NoTechForApartheid campaign started in partnership with Google and Amazon employees, tech workers are protesting against the use of their labour in creating harmful technology.

The labour movement’s pushback against the proliferation of harmful AI systems is not limited to tech workers: many industries that are affected by the potential uses of AI systems have joined the fight. AI was a key topic of contention in the historic strikes by writers and actors in Hollywood in 2023. Concept artists hired lobbyists and filed class-action lawsuits against companies that generated “AI art” using their work as training data, without consent or compensation. Creatives refused to accept studio terms stipulating that their material could be used to train generative-AI systems that could then put them out of work or devalue their labour.

Given widening inequalities around the world, the climate catastrophe pushing more people into the margins, and the growing number of refugees, which is projected to rocket while tech billionaires amass more money than ever, the labour movement is only going to grow in importance during 2024. It has a vital role to play as it becomes one of the key ways in which the development of harmful AI systems can be curbed.■

Timnit Gebru, The Distributed AI Research Institute