Bedrock is a new service from Amazon that allows users to build generative AI applications using pre-trained models from startups like AI21 Labs, Anthropic, and Stability AI. Bedrock also provides access to Titan FMs, a family of AI models developed in-house by Amazon. The service is currently in “limited preview,” and the company has made significant investments in training in-house models ahead of the launch. According to Grand View Research, the generative AI market, which includes systems that generate text, audio, and speech, could be worth more than $100 billion by 2030. Amazon’s motivation for entering the generative AI space is not only to capture a piece of this rapidly growing market, but also to re-energize its cloud business, which may be slowing. Amazon reported 27% year-over-year revenue growth for its cloud services in Q3 2022, but the increase slowed to a mid-20% rate by the end of the quarter. Meanwhile, Amazon’s cloud division’s operating margin was down 4 percentage points year over year in the same quarter, indicating that Amazon expanded too quickly. Amazon is serious about generative AI, so it has made CodeWhisperer, a system that generates code from text prompts, available for free to individual developers.
In other AI news, Elon Musk is reportedly planning to take on his former ally OpenAI and is attempting to gather the necessary funds and resources. Musk is preoccupied with several other businesses, so there is reason to be sceptical of this venture. Anthropic, an AI research startup, plans to raise up to $5 billion in the next two years to compete with rival OpenAI and enter over a dozen major industries. The company intends to build a “frontier model” known as “Claude-Next,” which will be ten times more capable than today’s most powerful AI, but this will cost a billion dollars over the next 18 months.
Poe, a new app, will now allow users to create their own chatbots by combining prompts with an existing bot, such as OpenAI’s ChatGPT, as the foundation. Stanford and Google researchers recently conducted an experiment with ChatGPT, creating a virtual town filled with AI-powered characters. Their attempt to create a “believable simulacra of human behaviour” appeared to be successful, as the 25 ChatGPT-powered AIs interacted convincingly human-like. Finally, Ron discusses how transformative technologies like ChatGPT could be if applied to enterprise applications that people use on a daily basis in a piece for TC+. He observes that designing new AI-powered interfaces in an elegant, non-bolted-on manner will necessitate creativity.
