Amazon may still be king of the cloud, but in recent months it’s had to watch its two closest rivals, Microsoft and Google, steal the spotlight with brilliant but error-prone chatbots that use cutting-edge “generative” artificial intelligence models.
Today, Amazon announced it’s joining the generative AI race. Not by launching its own chatbot, but by making two new AI language models available through its cloud platform, Amazon Web Services, which customers will be able to use to build their own bots.
The past few months have seen the tech industry foaming at the mouth over the potential of generative AI—algorithms that learn to produce text, code, imagery, and more. The boom has been inspired by the remarkable success of OpenAI’s text-generating bot, ChatGPT, as well as the success of AI-image generators.
The proliferation of these services also threatens to shake up the cloud market, as companies, from construction to the law, race to build their own generative AI tech or fold it into existing products. As investors and entrepreneurs dream up ways to use generative AI across all kinds of industries, they’re creating demand for the basic tools that companies can use to build their own applications.
“Many customers are intently interested in generative AI solutions,” says Adam Selipksy, CEO of AWS. “I would say in most of my customer conversations now questions about generative AI are coming up, with a leading question being, ‘When are we going to be able to get the solutions that we want?’”
Amazon remains the global market leader in cloud computing, followed by Microsoft and then Google. Microsoft has partnered with OpenAI, and is making the technology behind the ChatGPT available via its cloud platform Azure. Google, which has been developing the underlying language model technology for years, recently launched its own model PaLM, which is available through Google Cloud.
Microsoft, which invested $10 billion in OpenAI, offers access to GPT-3.5, one of the language models that powers ChatGPT, through an application program interface that lets developers make access calls to the model directly from their code. Google’s PaLM is similarly capable and accessible through its cloud API.
Today, Amazon announced it’s launching a platform called Bedrock, which will provide access to cutting-edge language models from Anthropic and AI21, two startups developing language models that compete with those of OpenAI and Google.
“It’s highly unlikely that there is any one model that is the right answer for all customers and for all use cases,” says Selipksy. “We want to provide choice and flexibility.”
Amazon has two generative language models. Titan Text can generate text from a prompt, and Text Embeddings generates a mathematical representation of text that can be used for tasks such as translation and search.
AWS will also offer access to Stable Diffusion, an AI model for generating imagery, from Stability AI, a startup that is developing a range of open source generative AI models.
“Bedrock removes significant friction from companies deploying generative models,” says Emad Mostaque, CEO of Stability AI, the company behind the image model. “[It] is ideal for our open models that can readily go to the customer data.”
Amazon says Bedrock will allow customers to use its models without the information they feed in being leaked back into the wider pool of data used to train those models, answering one key concern for organizations that want to build their own generative AI products. Companies, particularly those working with sensitive client information, such as financial services companies or health care providers, are concerned about the security of their data. As users interact with a model, it is sometimes possible for sensitive company information such as financial reports or business plans to leak into the model’s training data, and be shared with other customers.
“Data leaking into these incredibly powerful and useful AI tools is a major issue,” says Wayne Chang, an entrepreneur working at a stealth generative AI startup. “The ramifications are severe.”
Amazon’s generative AI foray will also see it offer new AWS instances optimized for running language models and other generative AI using custom silicon. Amazon is making its AI tool for coding, Code Whisperer, free for all individual users through AWS, something that may also help get users hooked on its cloud.
“Amazon is following its playbook of creating managed service versions of open source tools but this time also taking closed source models as managed services,” says Nathan Beniach, founder and general partner at Air Street Capital, a venture firm focused on AI. He adds that while Amazon remains dominant in the cloud, the machine learning tools it offers through AWS do not appear to have gained significant traction with developers.
Selipsky declined to say how the models measure up to the likes of OpenAI’s GPT-4 and Google’s PALM, insisting that it is “early days” for the technology generally.
