Skip to Content
Artificial intelligence

Your Best Teammate Might Someday Be an Algorithm

A new program from Google seeks ways for AI systems to work more effectively with humans.
July 10, 2017
PAIR

Google is launching an effort that it hopes may turn humans and intelligent machines into productive work buddies.

Through a project called People + AI Research, or PAIR, Google will release tools designed to help make the inner workings of AI systems more transparent. The company is also launching several research initiatives aimed at finding new ways for humans and AI systems to collaborate effectively.

A great deal has been made recently of the potential for artificial intelligence to eliminate jobs and replace humans, but in truth the technology may often serve as a powerful new kind of tool capable of automating only part of a person’s work.

Figuring out how to get humans and AI algorithms to collaborate effectively could have big economic significance, and it could shape the way the workforce is educated. It might also play a role in tempering negative reactions to the growing role of automation in many settings (see “Is Technology Destroying Jobs?” and “Who Will Own the Robots?”).

Barbara Grosz, a professor at Harvard University who has long argued that computer scientists should design AI systems to complement rather than replace people, says this approach is needed because AI is still so limited in what it can do. Grosz adds that human and machine capabilities can add up to more than the sum of their parts. “AI systems, like all computers, need to be developed for the people who are going to use them,” she says.

Some sense of how AI algorithms may collaborate with humans can be seen in games. Chess or Go players can team up with computer programs to perform at an elevated level. This requires a new set of skills and a new approach to each game.

“The past few years have seen rapid advances in machine learning, with dramatic improvements in technical performance,” the researchers wrote in a blog post announcing the initiative. “But we believe AI can go much further—and be more useful to all of us—if we build systems with people in mind at the start of the process.”

PAIR is being led by Fernanda Viégas and Martin Wattenberg, researchers who specialize in developing visualizations that make complex information more comprehensible. Viégas and Wattenberg previously developed a series of tools to elucidate the behavior of complex and abstract machine-learning models. The opacity of such models is a growing concern for those hoping to deploy them in a wider range of settings (see “The Dark Secret at the Heart of AI”).

The PAIR project is today releasing two tools for visualizing the kind of large data sets used to train a machine-learning model to make useful predictions. These visualizations can help a data scientist identify anomalies in training data.

Grosz says it is especially challenging to build AI systems that work well with humans, partly because those systems may be complex and opaque, but also because interacting with a human in an intelligent way is a major challenge. “In every situation you need to be able to model a person’s mental state,” she says.

Deep Dive

Artificial intelligence

conceptual illustration showing various women's faces being scanned
conceptual illustration showing various women's faces being scanned

A horrifying new AI app swaps women into porn videos with a click

Deepfake researchers have long feared the day this would arrive.

Conceptual illustration of a therapy session
Conceptual illustration of a therapy session

The therapists using AI to make therapy better

Researchers are learning more about how therapy works by examining the language therapists use with clients. It could lead to more people getting better, and staying better.

a Chichuahua standing on a Great Dane
a Chichuahua standing on a Great Dane

DeepMind says its new language model can beat others 25 times its size

RETRO uses an external memory to look up passages of text on the fly, avoiding some of the costs of training a vast neural network

ai learning to multitask concept
ai learning to multitask concept

Meta’s new learning algorithm can teach AI to multi-task

The single technique for teaching neural networks multiple skills is a step towards general-purpose AI.

Stay connected

Illustration by Rose WongIllustration 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.