Intel announces new Stephen Hawking speech system will be open source

'It has the potential to greatly improve the lives of disabled people all over the world,' says renowned physicist

Intel and Professor Stephen Hawking today announced a new speech system designed to make it easier for the physicist to communicate, and declared the intention to open source the system for the wider benefit of humankind early next year.

"We are pushing the boundaries of what is possible through technology - without it I would not be able to speak to you today," said Hawking at today's London press conference.

"Intel's research and development is bringing about changes in the world, and in the way that disabled people can communicate.

"By making this technology freely available, it has the potential to greatly improve the lives of disabled people all over the world."

Intel and Hawking have been working on the technology for the past three years, with Intel Labs experience designer Peter Denham describing the celebrated scientist as "magical" at finding bugs in the software.

Lama Nachman, manager of Intel's anticipatory computing lab, explained how the ACAT (Assistive Contextually Aware Toolkit) software operates in the same way as Hawking's 20-year-old cheek muscle-operated system, but is able to double the speed at which he can choose characters, and thus speak.

"So far we've seen about double his speech rate using this new system as well as speeding up some of the comments he does with the keys by about 10 times," Nachman said.

"The Intel team and I have been working together for almost three years on upgrading my communication system. My old system is more than 20 years old and I was finding it very difficult to continue to communicate effectively and to do the things I love to do," said Hawking.

"With the improvement made, I am now able to write much faster, which means I can continue to give lectures, write papers and books and, of course, speak with my family and friends more easily."

The project also involved SwiftKey, who used a version of its predictive text engine commonly found in mobile devices to build Hawking a new language structure to accompany Intel's UI.

"Text input shouldn't be about input method - it should be about language," said SwiftKey technical lead Joe Osborne.

Osborne explained how, with "the cost of each character [input being] high in terms of time and effort" for Hawking, finding a way to predict what he'd want to say, in an ongoing way which the system could learn, was essential.

Hawking now only needs to type "15 to 20 per cent", said Osborne, of any word he needs in order to have the system guess it correctly.

"Stephen's language is slightly atypical in terms of all his scientific lexicon that he uses," Osborne added. "The system will adapt very quickly, and will continue to adapt as he uses it, but having good baselines is never a bad thing. But it certainly does adapt to each new user."

Hawking described ACAT as "life-changing", adding "I hope it will serve me well for the next 20 years".

Intel Labs' Denman said the development team studied hours and hours of screen-captured activity showing how Hawking used facilities liked timed mouse cursor movements and even Windows shortcuts.

"We watched everything he did when he was at work, when was at play, everything we could possibly learn from him we wrote down and acknowledged," said Denman.

Denman, who uses a wheelchair himself, added that his own experiences helped him more closely collaborate with Hawking.

"When I first started working with him, I didn't want to step over any boundaries," said Denman, who added that he had been able to pay particular attention to spatial aspects of Hawking's life, like "equity of wheelchairs" and even "the small things" which can assist disabled people.

"I have this small mirror I have on my desk [I use to look behind me] and I gave it to him, and I think that changed his perspective around how he could view and see the world.

"Even though Intel is really good at building the biggest, fastest things out there, sometimes it's the small things that can actually help people out," concluded Denman.