Nov 20 (Reuters) - Numenta, a company with backing from
the Gates Foundation, on Wednesday released an open-source AI
model it hopes will require less energy and data to create
intelligent machines than existing technologies.
Numenta was co-founded by Jeff Hawkins, the founder of Palm
Computing in the 1990s, which in turn invented the Palm Pilot.
Hawkins' deepest academic interest is neuroscience, which showed
up in the Palm Pilot's handwriting recognition software that
leveraged an earlier era of artificial intelligence.
While most of today's popular AI systems are roughly modeled
on human brains with a technology called neural networks,
Numenta has a different idea of how the brain works and hopes to
persuade AI researchers to try its approach.
Most current AI systems ingest huge amounts of static data -
such as massive libraries of text or images - and run that
through thousands or even tens of thousands of computing chips.
After weeks or months of computing using vast amounts of
electricity, the AI system is trained to generate responses to
human queries.
Numenta believes that such training can be a more constant
process, where brain cells gather small bits of information from
senses such as sight and touch and constantly develop models of
the world that can be modified on the fly as conditions change,
like a young child learning a new toy.
The company says the approach can be applied to robots that
learn to navigate the world and other types of knowledge like
writing. On Wednesday, it will release open source computer code
to let other companies and researchers try its approach.
Other firms such as Meta Platforms ( META ) have also made
their AI technology freely available. That has helped Meta catch
up to rivals such as OpenAI, but has also led to the use of
Meta's technology by the Chinese military.
Subutai Ahmad, Numenta's chief executive, said the company
will closely monitor how its technology is used once it becomes
public.
"The technology is still early, and the way we deal with
this will depend on the exact way the technology matures," Ahmad
told Reuters. "These are important issues that we need to think
about."