While leading Google through the aughts, Eric Schmidt made a miscalculation.
“I was proven completely wrong” about artificial intelligence, Alphabet’s executive chairman said at the RSA security conference in San Francisco on Wednesday.
Schmidt has initially skeptical about the technology, and he’s since acknowledged how vital it is to both the company’s mission and to the global economy.
Indeed, Google (GOOG) CEO Sundar Pichai has described the world as having entered an “AI-first” era. The preceding phase was a focus on all things mobile- and smartphone-first (see: Android), according to Pichai, who succeeded Schmidt after a second CEO stint by Google co-founder Larry Page.
Get Data Sheet, Coins2Day ’s technology newsletter.
Schmidt’s assessment back then was that artificial intelligence research faced tremendous obstacles that inhibited its progress. He “didn’t think it would scale,” he said of the machine learning tech.
And he said he also didn’t think it would “generalize,” meaning becoming more flexible and elastic, like the human mind, rather than remaining a specialized tool suited only to specific tasks.
Schmidt had underestimated the power of simple algorithms to “emulate very complex things,” he said, while qualifying that “we’re still in the baby stages of doing conceptual learning.”
Read more: Forget Artificial Intelligence. Why ‘Artificial Stupidity’ Is the Real Threat
In other words, computer scientists are still teaching machines to heuristically categorize basic elements of the world: building representations of “things” and “actions” by parsing the components of images (like colors, shapes, and lines), as well as sounds (like tones, pitches, and phonemes).
“General AI,” or mimicking the elasticity of human thought, is still decades away from reality, by Schmidt’s estimation. But he has become more bullish about the prospect in recent years.
For more on artificial intelligence, watch Coins2Day’s video:
The moment that changed everything, he said, was the success of a particular Google experiment involving neural networks in 2012. Ironically, Schmidt said, the team’s creation didn’t uncover some major mathematical breakthrough. Rather, it found something far more mundane.
“You’d think it would have been the discovery of basic set theory,” Schmidt said, referring to an esoteric realm of mathematics. “Instead, it was the discovery of cats on YouTube.”
Indeed, the Google Brain team had tasked thousands of computer processors with recognizing objects in YouTube video thumbnail images. The result—an ability to distinguish cats—helped launch a wave of renewed interest in the field of deep learning. (For more on that tech revolution, read this recent Coins2Day feature.)
As far as questions about apocalyptic scenarios, like a robot uprising, Schmidt echoed statements he has made in the past by throwing water on the alarmists. “These are important philosophical questions, but ones that we’re not facing right now,” he said.
In Schmidt’s view, the positives far outweigh the negatives for AI. “Things that bedevil us, like traffic accidents and medical diagnoses will get better,” he said.
“I will stake my reputation that that will be the real narrative over the next five years.”