By Greg Bensinger
Dec 17 (Reuters) - Amazon is shaking up its artificial intelligence team with one long-time veteran heading for the exit and another taking over a more expansive group that includes AI models, custom silicon and quantum computing.
The move comes just two weeks after the Amazon Web Services annual cloud computing conference in Las Vegas, which the company has turned into a showcase for AI tools and services.
Amazon has been working to shake a reputation that it is trailing competitors
like Google, Microsoft and OpenAI in developing AI models. It has invested roughly $8 billion in startup Anthropic and integrated its technology into several of its internal and consumer products, and Amazon is considering as much as a $10 billion investment into OpenAI.
"We are at this inflection point with several of our new technologies that will power a significant amount of our future customer experiences," said Amazon CEO Andy Jassy in the post on Wednesday.
Cloud executive Peter DeSantis, a 27-year Amazon veteran, will lead the new unit overseeing its most advanced technologies, according to a company post. Rohit Prasad, who helped build the Alexa voice assistant and led its AI efforts recently, will step down at year's end.
Pieter Abbeel, who joined Amazon last year, is set to head up frontier model research with the AGI organization and continue to work on robotics. He came to Amazon when the company hired the founders of Covariant, a robotics startup.
DeSantis led AWS infrastructure and utility computing businesses, overseeing global data centers and core cloud services. He also spearheaded the 2015 acquisition of Annapurna Labs, which designs Amazon's custom chips.
The appointment comes as the company looks to unify development of its Nova AI models, chip programs such as Graviton and Trainium, and emerging quantum computing initiatives.
Jassy said Prasad was leaving of his own accord. It was not clear what he would do next.
(Reporting by Kritika Lamba in Bengaluru, Greg Bensinger in San Francisco; Editing by Arun Koyyur and Daniel Wallis)
