Skip to Content
Artificial intelligence

How DeepMind plans to stop AI from behaving badly

September 28, 2018

Researchers at the Alphabet subsidiary DeepMind have spelled out how they will ensure that AI is developed safely.

The guidelines aim to make certain that powerful systems capable of learning and figuring out their own solutions to problems don’t start to behave in unexpected and unwanted ways.

The big issues: The researchers say the key challenges are specifying the intended behavior of a system in a way that avoids unwanted consequences; making it robust even in the face of unpredictability; and providing assurances, or ways to override behavior if necessary.

Erratic behavior: This is a growing area of academic research. There are plenty of often amusing examples of machine-learning systems that have started behaving oddly. Take, for example, the AI agent that taught itself a rather bizarre way to rack up points in the game CoastRunners. The AI learned it could accumulate more points not by finishing a race, as was intended, but by hitting certain obstacles around the course instead (as in the gif above). DeepMind’s AI Safety team has also shown ways to have an AI agent shut itself off if it starts behaving in ways that might prove risky.

Far out: We shouldn’t worry unduly about AI systems becoming dangerously autonomous. In any case, there are far greater issues to worry about right now, including the bias that may lurk in AI algorithms or the fact that many machine-learning systems are difficult to understand.

Deep Dive

Artificial intelligence

Large language models can do jaw-dropping things. But nobody knows exactly why.

And that's a problem. Figuring it out is one of the biggest scientific puzzles of our time and a crucial step towards controlling more powerful future models.

OpenAI teases an amazing new generative video model called Sora

The firm is sharing Sora with a small group of safety testers but the rest of us will have to wait to learn more.

Google’s Gemini is now in everything. Here’s how you can try it out.

Gmail, Docs, and more will now come with Gemini baked in. But Europeans will have to wait before they can download the app.

Google DeepMind’s new generative model makes Super Mario–like games from scratch

Genie learns how to control games by watching hours and hours of video. It could help train next-gen robots too.

Stay connected

Illustration by Rose Wong

Get the latest updates from
MIT Technology Review

Discover special offers, top stories, upcoming events, and more.

Thank you for submitting your email!

Explore more newsletters

It looks like something went wrong.

We’re having trouble saving your preferences. Try refreshing this page and updating them one more time. If you continue to get this message, reach out to us at customer-service@technologyreview.com with a list of newsletters you’d like to receive.