Google CEO Sundar Pichai has weighed in on AI development, saying all models have hallucination issues and no one has solved the problem.
Hallucination refers to the issue when AI models make up fake information while insisting they are correct. Every single generative AI model has experience hallucination issues to some degree or another, and the problem poses one of the biggest challenges to widespread AI adoption.
In an interview with CBS’ 60 Minutes via Fortune, Pichai admitted that AI hallucination remains a major issue:
“Yes,” Pichai admitted, saying hallucinations are “expected.” “No one in the field has yet solved the hallucination problems. All models do have this as an issue.”
When asked if the problem will eventually be solved, Pichai said “it’s a matter of intense debate” and that his team will continue to “make progress.”
At the same time, progress will be difficult since researchers still “don’t fully understand” so aspects of modern AI.
“There is an aspect of this which we call—all of us in the field—call it a ‘black box,’” he said. “And you can’t quite tell why it said this, or why it got it wrong.”