HomeArtificial IntelligenceArtificial Intelligence NewsAWS is ready to power AI agents with busywork

AWS is ready to power AI agents with busywork

On Wednesday at the AWS Summit in New York, Amazon unveiled Agents for Bedrock, a platform that enables businesses to create AI applications that can independently perform things like booking a flight for consumers rather of just informing them about it. Instead of only making restaurant recommendations, AI agents are the personal assistants that actually make a reservation.

Swami Sivasubramanian, vice president of data and machine learning at AWS, says he thinks this would help developers who sought a simpler approach to create agents and at the same time customize the data the models read. Even with today’s state-of-the-art generative AI, creating agents still required a lot of effort, but these advances are allowing programmers to access only the models they require.

The ultimate goal of AI developers is to create AI agents that can perform tasks that humans don’t want to do.

Although sophisticated, generative AI models like GPT-4 or Llama 2 do not genuinely automate some tasks for users without additional assistance, such as plug-ins. With agents, a travel agency could utilize generative AI to recommend itineraries, then create agents that would consider a user’s travel preferences and history, identify flight schedules, and finally book the selected flight.

Last week, Amazon and other AI firms like Meta, Google, and OpenAI signed a commitment letter with the White House promising to create responsible AI.

In comparison to Meta, Google, OpenAI, and Microsoft, Amazon hasn’t played a significant role in the generative AI arms race. Nevertheless, it has established itself as a major infrastructure supplier in the area through AWS. New versions of Stable Diffusion XL 1.0 and the chatbot Claude 2 that were released by StabilityAI and Anthropic are now accessible on Bedrock.

When building extra apps to do tasks, businesses can utilize their own data to train foundational models, such as image-to-text models or huge language models, using Agents for Bedrock, which is geared towards startups and corporations. The underlying model to be used, the commands to be given, and the data to be read by the model can all be chosen by the developer.

Other tech firms are also developing agents, and Mark Zuckerberg, CEO of Meta, told investors that there is a chance to “bring AI agents to billions of people in ways that will be useful and meaningful.” In a detailed interview with The Atlantic, OpenAI CEO Sam Altman pondered the ideal method for creating AI agents.

In April of this year, Bedrock, a platform where businesses can access basic generative AI models like Stable Diffusion, Claude, Jurassic, and Amazon’s own Titan language, was introduced.

Other AWS announcements included a new service called AWS HealthScribe that will apply generative AI to the healthcare industry and a partnership with Nvidia that will enable AWS to use Nvidia H100 Tensor Core GPUs to handle bigger quantities of memory and data.

AWS HealthScribe automates clinical documentation, or notes that summarize a patient’s complaint, and creates diagnoses-related medical notations. To make the tool available, AWS collaborated with 3M Health Information System. According to Sivasubramanian, utilizing generative AI to generate clinical documentation frees up healthcare workers’ time so they may spend it interacting with more patients.

It will be challenging to apply AI to healthcare, though. A highly regulated industry in the US is healthcare. Security and privacy are issues that AI firms must address. The Federal Trade Commission’s inquiry into OpenAI has data privacy as one of its key areas of attention. Reiterating that HealthScribe complies with HIPAA regulations, AWS stated that the tool will not be trained using HealthScribe’s data.

According to Sivasubramanian, AI has the potential to significantly ease the load on people in a variety of fields, including healthcare.

Entity Resolution, an analytics tool powered by machine learning from AWS, has been made broadly accessible and aids in document analysis and matching across apps. For data engineers and analysts, AWS has also made its CodeWhisperer code helper accessible in Glue Studio, a tool that makes coding simpler. In real time, it offers coding ideas.

Source link

Most Popular