IDF: Intel predicts artificial intelligence in 40 years

Intel's Justin Rattner has outlined his vision of artificial intelligence

Intel's chief technology officer has predicted that humans will create computers smarter than themselves within the next 40 years.

Speaking at the final keynote of the Intel Developer Forum Justin Rattner said that the company, currently celebrating its 40th birthday, would create true artificial intelligence by 2048.

"In that time, machines from Intel will surpass human intelligence," he said.

"And progress in the next 100 years will be more like the progress of the past 20,000 years because of technology."

However, there are significant problems ahead in the short and long term and Intel outlined some of the solutions it is researching.

Dr Mike Garner, programme manager for Intel's emerging materials group, explained that CMOS technology still has a way to go, certainly to 32nm.

"We will need new materials to improve transistor performance and new technologies like tri-gate transistors which give more power, lower leakage and better density," he said.

"CMOS will be the platform [of the processor] and then we will have other things on top of that."

These would involve a possible switch from binary logic systems when quantum computers finally come online.

Intel is also working on signalling. Rattner demonstrated a photonic system that has a laser built onto a silicon chip sending data at 3.2Gbps down optical cable.

Intel scientists are working on a chip that has 20 such lasers, capable of data rates of up to 1Tbps.

Power will also be an issue. Rattner demonstrated how power could be beamed to a device wirelessly, albeit with a 25 per cent loss from just a few feet. He envisioned offices without cables, and possibly the end of the battery.

Jan Rabaey, the Donald O. Pederson Distinguished Professor in the Department of Electrical Engineering and Computer Sciences at the University of California at Berkeley, then discussed how wireless devices could be improved.

Two techniques were suggested: cognitive radio and collaboration. The first would analyse the spectrum and switch to the best available option. This would lead to the freeing up of large amounts of spectrum for mobile connectivity.

Meanwhile, wireless devices would also have to learn to work together in the future. "Wireless radios do not work together, in fact they fight each other," said Rabaey.

"If they could work together the system is more efficient and you would save energy. The Federal Communications Commission is looking actively at experiments in this space."