AWS’ AI clearinghouse faces competition beyond Microsoft and Google

Article by Jacob Bourne | Apr 17, 2023

The news: On Thursday, Amazon Web Services (AWS) released Bedrock, an AI toolkit for the enterprise.

Bedrock provides AWS customers a suite of generative AI tools, including tech for companies to build their own chatbots, per Insider.
The platform grants access to a variety of AI models from different companies including Anthropic’s Claude, Stability AI’s Stable Diffusion, AI21’s Jurassic-2, and Amazon’s Titan.
A preview of Bedrock is currently available to a few AWS customers like Uber, The New York Times, and Coda.

A strategic road less traveled: Amazon has been more gun shy about diving into commercial generative AI compared with rivals Microsoft and Google, but Bedrock could be a turning point.

The ecommerce giant has a partnership with startup Hugging Face, but Bedrock’s collection of models from different startups shows that it’s casting a wider collaborative net than big cloud competitors—giving clients a one-stop shop for trying out different tools.
Another key strategic difference is that Amazon is focusing on corporate customers instead of deploying generative technologies—like ChatGPT, Bing, and Bard—to individual consumers.
Focusing on the enterprise helps Amazon avoid some of generative AI’s social pitfalls, like effects on mental health, while still capitalizing on the tech’s hype.
More corporate access to AI tools could exacerbate rising concern about widespread AI-induced job losses, including among sales professionals.

Fresh competition: Cloud providers like AWS, Microsoft, and Google are betting on generative AI to help bolster revenues, but on-prem alternatives could weaken the outlook.

Software startup Databricks just released its Dolly 2.0, an open source, instruction-following large language model (LLM) as well as free training data to help companies cheaply develop AI chatbots in-house.
Meanwhile, Meta’s new family of small, efficient generative AI models called LLaMa could be a precursor to a class of advanced LLMs that can run on mass-market hardware.
Incidents like Samsung employees leaking corporate secrets to ChatGPT are a sign of a growing need for custom AI tools that keep IP data secure.
Although Amazon’s Titan model could be a safer choice because clients can adapt it with their own training data, some customers may eventually prefer to keep generative AI capabilities on-prem.