Tyger
Paranormal Adept
Google CEO said that they don't know how AI is teaching itself skills it is not expected to have.
THE "BLACK BOX" EFFECT: AI is Learning Things Its Creators Can't Explain
Google CEO Sundar Pichai has issued a fascinating - and slightly eerie - warning about the current state of artificial intelligence. In a major interview, Pichai admitted that even experts in the field don't fully understand how their models are developing certain "superhuman" skills.
The Mystery of "Emergent Properties"
James Manyika, Google's SVP of Technology and Society, shared a mind-blowing example: a Google AI model was given a few prompts in Bengali - a language it wasn't explicitly trained to know - and it spontaneously learned how to translate the entire language.
Pichai confirmed that the industry refers to this phenomenon as a "black box". "You don't fully understand," he said. "And you can’t quite tell why it said this."
These "emergent" abilities include everything from advanced reasoning and multi-step logic to writing computer code and solving complex math, often without being specifically programmed for those tasks.
Is It Safe to "Turn It Loose"?
When pressed on why Google would release technology it doesn't fully understand, Pichai countered: "I don't think we fully understand how a human mind works either."
Pichai admitted that all current AI models still "hallucinate"- confidently stating facts that are completely fabricated - and that no one has yet solved this issue.
Despite the mysteries, Pichai believes AI will be "more profound than fire or electricity," potentially revolutionizing everything from healthcare diagnostics to professional architecture.
THE "BLACK BOX" EFFECT: AI is Learning Things Its Creators Can't Explain
Google CEO Sundar Pichai has issued a fascinating - and slightly eerie - warning about the current state of artificial intelligence. In a major interview, Pichai admitted that even experts in the field don't fully understand how their models are developing certain "superhuman" skills.
The Mystery of "Emergent Properties"
James Manyika, Google's SVP of Technology and Society, shared a mind-blowing example: a Google AI model was given a few prompts in Bengali - a language it wasn't explicitly trained to know - and it spontaneously learned how to translate the entire language.
Pichai confirmed that the industry refers to this phenomenon as a "black box". "You don't fully understand," he said. "And you can’t quite tell why it said this."
These "emergent" abilities include everything from advanced reasoning and multi-step logic to writing computer code and solving complex math, often without being specifically programmed for those tasks.
Is It Safe to "Turn It Loose"?
When pressed on why Google would release technology it doesn't fully understand, Pichai countered: "I don't think we fully understand how a human mind works either."
Pichai admitted that all current AI models still "hallucinate"- confidently stating facts that are completely fabricated - and that no one has yet solved this issue.
Despite the mysteries, Pichai believes AI will be "more profound than fire or electricity," potentially revolutionizing everything from healthcare diagnostics to professional architecture.