Amazon is introducing a cloud service called Bedrock that developers can use to enhance their software with artificial intelligence systems that can generate text, similar to the engine behind the popular Microsoft-backed OpenAI-powered chatbot ChatGPT.
The ad indicates that the largest provider of cloud infrastructure will not leave a hot growth area to challenges like Google and Microsoft, which have started to offer developers great language models that they can take advantage of. Broadly speaking, large language models are artificial intelligence programs trained on large amounts of data that can compose human-like text in response to prompts that people type.
Through its Bedrock generative artificial intelligence service, Amazon Web Services will offer access to its own proprietary language models called Titan, as well as language models from Google-backed startups AI21 and Anthropic, and a model for turning text into images from startup Stability AI. . A Titan model can generate text for blog posts, emails, or other documents. The other can help with search and customization.
“Most companies want to use these great language models, but the really good ones take billions of dollars to train and many years, and most companies don’t want to go through that,” said Amazon CEO Andy Jassy, on CNBC.squawk boxThursday. “So what they want to do is build off of a foundational model that’s already great and great and then have the ability to customize it for their own purposes. And that’s what Bedrock is.”
The Bedrock initiative comes a month after OpenAI announced GPT-4, a great language model that powers ChatGPT, a chatbot that went viral after its launch in November. The most formidable competition for Amazon’s AWS business comes from Microsoft, which has billions invested on OpenAI and provides computing power to the startup through its Azure cloud.
People who use ChatGPT and Microsoft’s Bing chatbot based on OpenAI language models have occasionally encountered inaccurate informationdue to a behavior called hallucination, where the result may seem convincing but actually has nothing to do with the training data. Amazon is «really concerned about» accuracy and ensuring its Titan models produce high-quality responses, Bratin Saha, a vice president at AWS, told CNBC in an interview.
Customers will be able to personalize Titan models with their own data. But that data will never be used to train Titan models, ensuring that other customers, including competitors, don’t end up benefiting from that data, another vice president said.
Sivasubramanian and Saha declined to discuss the size of the Titan models or identify the data Amazon used to train them, and Saha did not describe the process Amazon followed to remove problematic parts of the model’s training data.
Amazon isn’t disclosing the cost of the Bedrock service, because it’s starting a limited preview for now. Clients can be added to a waiting list, a spokesperson said. Microsoft and OpenAI have announced pricing for using GPT-4, starting at a few cents for 1,000 «tokens,» with one token equaling about four characters of English text. Google has not released pricing for its PaLM language model.
Sivasubramanian, who has been with Amazon since the mid-2000s, said Amazon has been working on AI for more than two decades and AWS has amassed more than 100,000 AI customers. Amazon has been using an enhanced version of Titan to deliver search results through its home page, he added.
But Amazon is just one of the big companies that have rushed to bring generative AI capabilities to light after ChatGPT came along and became a hit. ExpediaHubSpot, Paylocity and Spotify are among the companies that have committed to integrating OpenAI technology.
Morgan Stanley analysts said in a note on Wednesday that, based on a February survey of CIOs, they expect AI to become a bigger part of cloud spending, with Google and Microsoft being the main beneficiaries, not Amazon.
“We always launch when things are ready, and all of these technologies are very early,” Sivasubramanian said. He said that Amazon wants to make sure Bedrock is easy to use and cost-effective, thanks to the use of custom AI processors.
C3.ai, Pegasystems, Accenture and Deloitte are among the companies hoping to use Bedrock, he wrote in a blog post.