RSA USA Alphabet exec chairman Eric Schmidt is worried that the future of the internet is going to be under threat once the worlds militaries get good at artificial intelligence.

Speaking at the RSA security conference in San Francisco, Google's ultimate supremo said he is worried the internet will be balkanized if countries lock down their borders to prevent citizens' personal information flowing into other nations. That would obviously be bad news for a global cloud giant like Google.

Schmidt also fears states are developing their own AI-powered cyber-weapons for online warfare. He said machine-learning research needs to be out in the open under public scrutiny, not locked away in some secret military lab.

For one thing, that would help everyone prepare their network defenses for AI-driven attacks, as opposed to being blindsided by highly classified technology. It would also get folks talking about whether or not it's a good thing to put powerful AI into the hands of untouchable exploit-wielding government intelligence agencies.

The technology industry needs to ask if we can come up with a way for countries not to use machine learning to militarise the internet, Schmidt said during a keynote address. If they did the internet would start to get shut down. Id like to see discussions on stopping that.

This is one of the reasons Google will open-source as much of its AI research as possible, he said. While some companies he mentioned no names want to keep their AI research private, Google thinks the benefits of being open and scrutinized by the crowd far outweighs any loss of competitive advantage.

We have to say: that's kinda funny, Eric, because Google and its AI wing Deepmind are close to being the most secretive closed-source organizations on the planet.

Schmidt said his Chocolate Factory has plowed millions of dollars into building machine-learning software, and thus had something of an advantage. But the next big breakthrough could be achieved by someone working out of their garage and thats healthy competition.

He had been surprised by the power of AI systems, as research in the sector hit a brick wall in the 1980s. But increasing computing power, and better machine-learning programming and algorithms, will make artificial intelligence commonplace soon.

The first area we are going to see it widely deployed is in computer vision, he predicted. Computers are already showing themselves to be superior to humans in this regard, he said, pointing to cases where computers are better at analyzing medical images than human doctors in part because they have been trained on millions of images instead of just thousands that a medic might see in their career.

Self-driving cars would also be early adopters, he said, for similar reasons. Then again, given the problems some Google cars have with balloons and bright sunlight this may not come as fast as Schmidt thinks.

While a self-aware AI system is a the stuff of popular fiction, Schmidt told the audience not to worry too much about it. While getting AI systems that share human values and which can be controlled is an important philosophical question, he said, theres no sign that the Singularity is on the horizon.

We are nowhere near that in real life, were still in baby stages of conceptual learning, he said. In order to worry about that you have to think 10, 20, 30, or 40 years into the future.

See the rest here:

THE SCHMIDT HITS THE BAN: Keep your gloves off AI, military top brass - The Register