Biden signs executive order aimed at growing AI infrastructure in the US
ICC chief prosecutor wants Israeli objections over Netanyahu warrant to be rejected
What to watch for in Hegseth's confirmation hearing
Some Paris Olympic athletes ask for medals to be replaced after quickly deteriorating
Pope Francis tells in memoir how documents on abuse and corruption cases were handed to him
Los Angeles money managers grapple with wildfire impact on operations, staff
Here's what we know about Trump's planned executive orders after swearing-in
The big DC mystery: Where will Trump and his circle hang out?
Republican-led US House to vote on limits for transgender student athletes
By Greg Bensinger
(Reuters) - Amazon.com announced a series of largely incremental refinements on Wednesday to several of its artificial intelligence products as it seeks to keep rivals at bay amid a continued investor frenzy over the technology.
The retailer has sought to counter a perception that competitors Google, Microsoft and OpenAI have taken a lead in developing generative AI, which can respond almost instantaneously with full sentences or pictures to complicated prompts or queries.
The improvements Amazon announced at its conference in New York include adding additional memory to so-called agents that automate work for businesses, so that each new request can build on prior ones, said Vasi Philomin, Amazon's vice president of generative AI.
"This allows agents to provide more personalized and more seamless experiences, especially for complicated tasks," said Philomin in an interview Tuesday.
He said, for instance, that the updated AI agents could remember for each subsequent request that a user prefers aisle or window seats on a flight, which was not previously possible.
As well Amazon said it updated the Q chatbot it announced last November to make improved suggestions for writing software code, addressing one of the more popular uses for generative AI.
Amazon also said it made improvements to help customers of its Bedrock service, which lets businesses create applications with a range of AI models, to detect and filter out so-called hallucinations -- when AI creates answers to questions or requests that may be wrong or misleading.
Hallucinations have been a nagging problem in AI systems because they breed mistrust among users. Google, for instance, was criticized earlier this year for an AI-powered search feature that, among other things, recommended users add glue to pizza sauce to ensure cheese sticks to it.
The new controls will help reduce the occurrence of hallucinations by about 75% for certain uses, Matt Wood, vice president of AI products at Amazon Web Services, said in an interview.
AWS, which oversees much of Amazon's AI development, is on a pace to reach $100 billion in annual revenue, the company said in April.
(Reporting by Greg Bensinger; Editing by Michael Perry)
