AWS has a new tool that aims to stop AI hallucinations for good
- AWS unveils automated reasoning checks
- A new tool will try to reduce the number of AI hallucinations by comparing responses with customer data
- Customers are shown AI responses to see how accurate they are
As companies increasingly use AI tools for their most important processes and tasks, hallucinations are proving to be an increasing challenge.
To address this, Amazon Web Services (AWS) has announced a new tool to tackle hallucinations.
The new Automated Reasoning control system, unveiled at the AWS re:Invent 2024 event, looks to reduce potentially harmful errors caused by hallucinations that could leave companies facing security risks or financial losses.
An end to AI hallucinations?
At the simplest level, hallucinations occur when an AI system or service behaves incorrectly or becomes unreliable, often due to issues with errors in the data it was trained on.
Described by the company as “the first and only generative AI security that helps prevent factual errors due to model hallucinations”, AWS’s Automated Reasoning checks attempt to solve this by comparing the responses generated by a model with those generated by the customer provided information. If it cannot determine if the answer matches exactly, the answer is sent back to the model for checking.
The new checks, available as part of Amazon Bedrock Guardrails, the company’s system for keeping AI models accurate and reliable, will also try to see how the model arrived at its answer, and if it considers it is incorrect, they will compare it with that of the customer. information.
It then presents its response alongside the model’s initial response, meaning clients can see the potential gap between the truth and the response and adjust their model accordingly.
AWS gave the example of a healthcare provider using the tool to ensure customer questions about specific policies receive accurate answers.
“As generative AI transforms more businesses and customer experiences over time, inference will become a core part of every application,” says Dr. Swami Sivasubramanian, Vice President of AI and Data at AWS.
“With the launch of these new capabilities, we are innovating on behalf of customers to solve some of the biggest challenges, such as hallucinations and costs, that the entire industry faces in bringing generative AI applications to production.”