Amazon has unveiled its chatbot called Amazon Q, which is focused on use in the workplace.
This puts Amazon in competition with other enterprise chatbots such as OpenAI's ChatGPT and Google's Bard chatbot. However, early reviews and a lack of benchmarks suggest that Amazon Q is not as advanced as OpenAI's GPT-4.
Amazon Q is a workplace assistant
Amazon Q was developed by Amazon's computing division to help employees with tasks such as summarizing strategy papers, filling out internal support tickets, and answering questions about company policies.
According to Amazon, the chatbot is designed exclusively for businesses and optimized for workflows. It is designed to support decision-making and problem-solving in the workplace. To this end, it can be linked directly to company data from numerous sources such as Google Drive, Github, Microsoft Teams, Slack and others.
Amazon Q can tailor its interactions to individual users based on their existing identities, roles, and permissions within the organization. Amazon guarantees the security and privacy of corporate data that is not used to train the underlying models.
According to Amazon, Q can also help with programming, especially developing applications for AWS with its "17 years of AWS knowledge and best practices."
GPT-4 remains the undisputed leader
According to an initial assessment by Ethan Mollick, professor at the Wharton School, Amazon Q is more of a GPT 3.5 class chatbot that cannot compete with OpenAI's GPT-4.
Wharton was involved in an extensive study of the practical use of GPT-4 in consulting. He is familiar with the capabilities of the different language models and notes that Amazon's Q is "very far short of GPT-4."
"With no benchmarks to go off, all I can say is it feels like a GPT-3.5 class model with lots of guardrails that is very narrowly focused on a knowledge base."
This makes Q another example of a phenomenon of the last few months: Google Bard, Claude 2, Musk's Grok, Meta Llama 2, and many others - no company, big tech, or startup, comes close to the quality of GPT-4.
Instead, the market is full of more and often less useful chatbots at the level of GPT-3.5. GPT-4, however, is the first LLM to reach a level that makes it usable for demanding text tasks.
The burning question for the AI industry now is: does OpenAI have some secret sauce that makes it just that much better - or has GPT-4 already hit the performance ceiling and OpenAI just bought it at a high price? After all, OpenAI currently seems more concerned with making its system cheaper than better.
But to keep things in perspective: GPT-4 has only been on the market since March 2023. It will take longer to develop and train better models.
In general, Amazon is playing catch-up in the AI race. The company plans to invest up to $4 billion in AI startup Anthropic, which competes with OpenAI, and is reportedly training a new exclusive LLM to GPT-4 "Olympus". Amazon recently disbanded a large part of its Alexa team to focus its resources on developing generative AI.