If OpenAI led the way with ChatGpt and Microsoft jumped on it by backing the project with $10 billion, Google launched Bard and Chinese competitors its own chatbots (despite the government’s giving it a hard time), Jeff Bezos’s Amazon could hardly sit idly by.
Its answer, with due differences, to ChatGpt is called Bedrock. And Amazon presented it as “a new service for building and scaling generative artificial intelligence (AI) applications, meaning it can generate text, images, audio and synthetic data in response to requests.”
Table of Contents
Amazon’s relationship with AI
“If it seems lately that artificial intelligence is everywhere, it is. But it has been powering our everyday experiences for some time”. Amazon was keen to point out as it unveiled its latest innovations.
“When you ask Alexa to play a song, when you walk out of an Amazon Just Walk Out store with a sandwich in hand, or when you press play on an Amazon Prime recommended movie, you are tapping into AI. More precisely,” the company says, “you are interacting with machine learning (ML) models.
In fact, Amazon, pointed out one of the vice presidents of Amazon Web Services (AWS, the e-commerce giant’s cloud), has been working on AI for more than two decades. And has amassed more than 100,000 AI customers.
How Bedrock is different
“No one model fits all,” Amazon said, explaining that Bedrock provides a range of language models from leading vendors. So that AWS customers have the flexibility and choice to use the best ones for their specific needs.
“Most companies want to use these large language models. But the really good ones require billions of dollars in training and many years, and most companies don’t want to go through that.
They want to work from a basic model that is already large and complete and then have the ability to customize it for their purposes. And Bedrock is just that,” said Amazon CEO Andy Jassy.
In fact, customers will be able to customize Titan models with their own data. Which – according to the company – will neither be used to train the models. Nor end up in the hands of other customers or competitors. According to Swami Sivasubramanian, one of AWS’s vice presidents and at Amazon since the mid-2000s, in a post, “C3.ai, Pegasystems, Accenture and Deloitte are among the companies looking forward to using Bedrock.”
Finally, Bratin Saha, another AWS vice president, wanted to add that “Amazon is very concerned about accuracy and ensuring that its Titan models produce high-quality responses.” Could this be a dig at ChatGpt’s so-called hallucinations? That is, to those answers provided by the AI that are on the surface convincing but in essence incorrect.
What Amazon did not say
However, the company did not reveal the size of the Titan models, details about the data Amazon used to train them. Or even how the problematic parts of the training were removed.
As Cnbc recalls, the cost of the Bedrock service has also not been revealed. This is because for now it is a limited preview and customers can sign up for a waiting list.
Microsoft and OpenAI have announced prices for using GPT-4. Which start at a few cents per 1,000 tokens, with one token equal to about four characters of English text.
Google, on the other hand, has similarly not announced prices for the PaLM language model.
Read also: Sam Altman, CEO of OpenAI, admits: GPT-4 now scares us, too
