Hot AI News

New AI learns to think more like humans

Scientists create a computer brain that makes decisions similar to people

New AI learns to think more like humans

Scientists at Georgia Tech have made a new type of artificial intelligence (AI) that thinks more like humans. This AI, called RTNet, can make decisions in a way that’s closer to how people do it. This is a big step forward in making AI that’s more reliable and useful.

An AI that weighs evidence and doubts like people

  • Scientists created an AI that makes decisions more like humans do.
  • The AI uses probability and gathers evidence before deciding.
  • It performed similarly to humans when identifying messy handwritten numbers.
  • The AI showed human-like confidence in its decisions.
  • This research could lead to more reliable and helpful AI in the future.

Most AIs today make the same decision every time they face a problem. But humans don’t work that way. We often make different choices even when we see the same thing twice. This is because we’re not always sure, and we gather information before deciding.

The team, led by Associate Professor Dobromir Rahnev, wanted to make an AI that could copy this human way of thinking. They used a special type of computer program called a neural network. This network was trained to look at handwritten numbers and figure out what they were.

To make the AI more human-like, the scientists added two important things. First, they used something called a Bayesian neural network. This lets the AI use probability, just like humans do when we’re not sure about something. Second, they made the AI gather evidence before making a choice. This is like when we look around and think before we decide.

The scientists tested their new AI by showing it messy, hard-to-read numbers. They also had real people look at the same numbers. They found that the AI made choices very similar to the humans. It was right about as often as people were, it took about the same time to decide, and it seemed sure of itself in the same way people did.

One interesting thing they found was that the AI naturally showed more confidence when it was right, just like people do. The scientists didn’t have to teach it this – it happened on its own.

This new AI did better than other types of AI, especially when it had to make quick decisions. This is because it works more like a human brain, which is very good at making fast choices.

The team hopes this research will lead to AI that can help us with everyday decisions. Humans make thousands of choices every day, from what to eat to whether it’s safe to cross the street. In the future, AI like this might be able to help us with some of these choices, making our lives easier.

The next step is to test this AI on different kinds of problems, not just numbers. The scientists also want to use what they’ve learned to make other AIs think more like humans. This could help make AI that we can trust more and that works better with people.

Sign Up for the Daily AI Pulse

One email a day. All the stories that matter.

Daniel ReedAI Research, Safety & Ethics Analyst

Daniel Reed currently works as an AI Research, Safety & Ethics Analyst at Aiholics, writing about how changes in artificial intelligence are affecting and will affect scholarship,…

You may also like