What if a superintelligent AI turns rogue? OpenAI doesn’t want to find out

What if a super-intelligent AI, more intelligent than even the smartest humans, one day turns evil? A new team from OpenAI is here to make sure we never find out.

In an announcement this week, OpenAI said it is building a team “to drive and control AI systems smarter than us.”

Too: Beyond AI, This Other Technology Wave Is Booming

Superintendence, which could arrive within the next decade, will be the most influential technology ever created and could help solve the world’s most important problems, the artificial intelligence lab said. But it was followed by a dire warning that it “could also be very dangerous, and could lead to the empowerment of humanity or even human extinction.”

As the company says, humans can keep artificial intelligence under control because they are smarter. But what happens when AI passes through humans?

That’s where the new team comes in, led by the current Chief Scientist of the Research Laboratory and the current Head of Alignment. In addition to that leadership, the team will include OpenAI’s top researchers and engineers, as well as 20% of its current compute power.

Too: ChatGPT browsing feature defunct just a week after it was rolled out – here’s why

The ultimate goal would be to create a “roughly human-level automated alignment researcher,” or an AI system that achieves a set goal and does not stray outside set parameters. The team plans to accomplish this in three phases:

• Develop an understanding of how AI analyzes other AI without human interaction
• Use AI to find any problem areas or exploits
• Intentionally training some AI wrong to see if it figures it out

In short, a team of humans at OpenAI is using AI to help train the super smart AI to take control. And they think they will complete it in four years. They acknowledge it’s an ambitious goal and there’s no guarantee of success, but they’re confident in their work.