Amazon Reportedly Working on a Multimodal AI Chatbot to Take on OpenAI’s ChatGPT

Amazon is reportedly working on developing an artificial intelligence (AI) chatbot that can rival OpenAI’s ChatGPT. The company’s internal project is said to be code-named Metis. As per the report, the under-development chatbot will be able to perform all the general tasks of generative AI such as creating text content, responding to queries, and more. It will reportedly also support image generation and have access to the internet. The report further claims that the new AI platform could be launched in September 2024, around the time when Amazon hosts its annual Devices and Services event.

Amazon could soon launch its own AI chatbot

According to a Business Insider report, the e-commerce giant aims to directly compete with ChatGPT with its in-house AI model. Citing people familiar with the project, the publication claims that the project is codenamed Metis, after the Greek goddess of wisdom, prudence, and deep thought. The chatbot is being designed to be accessed through a web browser, similar to most popular AI chatbots.

Based on an internal document obtained by the publication, it claimed that the Metis chatbot will be powered by the company’s in-house AI model called Olympus. It is said to be more advanced than the existing Titan large language model (LLM) that powers some of Amazon’s products.

In its functionality, the AI chatbot is said to be able to perform text-based tasks such as having conversations, answering queries, and generating content. Additionally, it can also generate images, as per the report. This suggests that Metis will use a multimodal AI model. Notably, ChatGPT cannot generate images, but users can get a subscription to Dall-E and use it via ChatGPT.

Metis to use retrieval-augmented generative AI framework

The Metis chatbot will reportedly work on the retrieval-augmented generation (RAG) AI framework. The mechanism improves the quality of responses by using a combination of text generation and information retrieval from large datasets. It is said that the chatbot will be able to access and retrieve information from the internet.

For instance, it will be able to show near real-time stock price updates which is something many AI chatbots struggle with. However, both GPT-4o-powered ChatGPT and Gemini can do this.

The report claims that the project is being developed by the company’s Artificial General Intelligence (AGI) division which is led by the Senior Vice President and Head Scientist Rohit Prasad. At the same time, Amazon CEO Andy Jassy is also said to be very involved with the project.

The report highlights that some employees are concerned that Amazon might be too late into the AI chatbot race, which is already getting a bit too crowded.