Amazon.com announced a series of incremental improvements to its artificial intelligence products on Wednesday, aiming to maintain a competitive edge amidst ongoing investor interest in AI technology. The retailer is working to counter the perception that rivals Google, Microsoft, and OpenAI have taken the lead in developing generative AI, which can quickly respond to complex prompts with full sentences or images.
At its conference in New York, Amazon revealed enhancements to its AI agents that automate business tasks. Vasi Philomin, Amazon’s vice president of generative AI, explained that the improvements include adding more memory to these agents, allowing each new request to build on previous ones. This update enables the agents to deliver more personalized and seamless experiences, particularly for complicated tasks. For example, the updated AI agents can now remember user preferences for aisle or window seats on flights for subsequent requests, a feature that was not previously possible.
Additionally, Amazon has updated the Q chatbot, initially announced last November, to provide better suggestions for writing software code, one of the popular applications of generative AI.
The company also announced enhancements to its Bedrock service, which helps businesses create applications using various AI models. These improvements aim to detect and filter out “hallucinations,” where AI generates incorrect or misleading answers. Hallucinations have been a persistent issue, causing user mistrust. For instance, Google faced criticism earlier this year for an AI-powered search feature that mistakenly suggested adding glue to pizza sauce to help cheese stick.
According to Matt Wood, vice president of AI products at Amazon Web Services, the new controls will reduce the occurrence of hallucinations by about 75% for certain applications.
Amazon Web Services, which oversees much of the company’s AI development, is on track to reach $100 billion in annual revenue, as reported in April.
