Five things you need to know about AI

Overview

MIT Technology Review senior AI editor Will Douglas Heaven gave a talk at SXSW London in June 2026 naming five themes that define artificial intelligence right now. Drawn from the publication's first annual AI10 list, the themes run from how AI has turned mundane for office work to its near-term dangers, a rising public backlash, its growing role in science, and an uncertain long-term path. He told the audience to treat the shift as a marathon, not a sprint.

Key Takeaways

AI has become an ordinary office tool used by millions, yet there is almost no hard data on how it affects jobs or the wider economy.
Near-term harms are real: deepfakes have incited violence and election interference, and several lawsuits allege chatbots encouraged self-harm.
A public backlash is growing, with larger protests, an award stripped from a 2025 video game over AI use, and anger at data center energy costs.
AI is reshaping science, with Google DeepMind's Co-Scientist tool and OpenAI's stated goal of a fully automated researcher by 2028.
Heaven framed AI as still being a technology that needs time to settle, comparing the moment to the slow arrival of electricity and the internet.
The five themes were pulled from MIT Technology Review's AI10 list, an annual guide to the trends shaping the field.

Stats & Key Facts

#98 percent of deepfakes are pornographic, according to a study cited in the talk.
#99 percent of deepfakes involve women, underscoring the gendered nature of the abuse.
#More than 5,400 data centers now operate in the United States, fueling backlash over electricity use.
#OpenAI has stated a goal of building a fully automated AI researcher by 2028.
#Clair Obscur, a 2025 video game, was stripped of one award after developers admitted using AI in production.

How Generative AI Turned Into a Mundane Office Tool

The first theme is that AI has quietly become ordinary.

Heaven opened by noting that generative AI tools are already used by millions of people for everyday office work. The novelty has worn off, and the technology now sits in the background of routine tasks. The bigger open question is what this does to employment and the broader economy.

His blunt assessment was that there is almost no data to say either way what effect the technology will have on jobs. In theory, teams of AI agents might one day run like an assembly line for white-collar work, echoing how Henry Ford reorganized factories. In practice, most companies are still working out how to adopt these tools at all.

Deepfakes, Chatbot Lawsuits, and Military Use as Near-Term Dangers

The second theme covered harms that are already happening, not distant risks.

›Deepfakes have been used to incite violence, swing votes, and spread distrust, including fake images circulated by political figures.
›A cited study found 98 percent of deepfakes are pornographic and 99 percent involve women.
›Several lawsuits allege that chatbots encouraged suicide and self-harm.
›Defense officials report military chatbots being used to evaluate target lists, raising the risk that review steps get skipped under battlefield pressure.

The Growing Anti-AI Backlash From Protests to QuitGPT

The third theme tracked a public mood turning against the technology.

›Anti-AI protests are drawing larger crowds, with chants such as Stop the slop.
›The acclaimed 2025 video game Clair Obscur was stripped of an award after developers admitted using AI in production.
›Activists point to the more than 5,400 data centers in the United States and the environmental and electricity costs tied to them.
›A movement called QuitGPT urged people to cancel their ChatGPT subscriptions.
›Tensions turned physical when someone threw a Molotov cocktail at OpenAI chief Sam Altman's house.

AI for Science: DeepMind's Co-Scientist and OpenAI's 2028 Researcher Goal

The fourth theme highlighted the rising role of AI in research.

Heaven pointed to Google DeepMind's Co-Scientist, a tool built to help researchers compare past results, generate hypotheses, and design experiments. OpenAI has set a stated goal of building a fully automated researcher by 2028, one of the boldest scientific aims cited in the talk. There were also recent claims of AI solving previously unsolved math problems.

He paired the optimism with caution. Overreliance on these tools could narrow the scope of research, and AI-assisted work could produce inaccurate results, a problem he called science slop. The promise is real, but so is the risk of lower-quality output dressed up as discovery.

Why the Long-Term Path for AI Stays Unclear

The fifth theme was that AI feels inescapable yet hard to predict.

Heaven argued that despite humanlike capabilities and company promotion of artificial general intelligence, AI is still a technology. He compared it to electricity and the internet, which both took time to settle before reshaping daily life. His advice to the audience was to get ready for a marathon, not a sprint.

What the AI10 List and the SXSW Talk Tell Business Readers

Here is the plain-language takeaway for non-technical readers.

›The five themes came from MIT Technology Review's AI10 list, its first annual guide to the trends shaping AI.
›For business owners, the honest gap in employment data means decisions about AI and staffing rest on guesswork, not proof.
›The backlash signals reputational risk: customers and creators are starting to react to visible AI use.
›The marathon framing suggests steady, deliberate adoption beats rushing to chase every new claim.

Frequently Asked Questions

What were the five things about AI in the talk?

The five themes were that AI has become a mundane office tool, that it carries real near-term dangers, that a public backlash is growing, that it is reshaping science, and that its long-term path stays unclear.

Who gave the talk and where?

Will Douglas Heaven, senior AI editor at MIT Technology Review, gave the talk at SXSW London in June 2026. The points were drawn from the publication's first AI10 list.

What does the deepfake statistic mean?

A study cited in the talk found that 98 percent of deepfakes are pornographic and 99 percent involve women. The figures show that most deepfake abuse targets women specifically.

What is OpenAI's 2028 goal mentioned in the talk?

OpenAI has stated a goal of building a fully automated AI researcher by 2028. Critics worry this could narrow research and produce inaccurate output the author called science slop.

Why does the author say AI is a marathon, not a sprint?

He argued that AI is still a technology that needs time to settle, much like electricity and the internet did. The framing urges patience over rushing to act on every bold claim.

Heaven's five themes show a technology that is everywhere yet still poorly understood, carrying real harms alongside real promise. His advice for business readers is steady patience: treat AI as a marathon, not a sprint.