Mustafa Suleyman, one of the original founders of Google DeepMind, has launched a new AI company called Inflection AI, which aims to provide a personalized conversational AI companion for consumers.

The assistant is called Pi, and it’s powered by the same kind of large language model (LLM) that serves as the brain behind ChatGPT. With a major caveat. It’s a somewhat smaller model, focused on becoming an expert on the needs and preferences of its human user. Pi will run on phones, laptops, in SMS apps and elsewhere; it talks to the user in a personal and emotionally intelligent way, and gradually builds up a memory of every conversation between machine and human.

Mustafa Suleyman [Photo: courtesy of Inflection AI]

“You’re going to store lots of your creative moments, your brainstorming moments, a lot of your personal information, and you’re going to want that information with you, just like, back in the day, we might’ve had a thumb drive with us,” Suleyman tells Fast Company. “This is the arrival of a new era in computing. This is going to be like bringing your digital life with you wherever you are.”

Using Pi is different from using ChatGPT. Pi talks to the user in a warmer, more personalized way. And the types of questions users might ask Pi are different than what they might bring to ChatGPT or the Perplexity search assistant.

During a demo, Suleyman told Pi that he was torn between going out to a NASCAR race with his friends or staying home and building a small pond in his backyard. Pi told him it was natural to “feel pulled in two directions” between social and personal duties. The assistant spoke to him in a supportive, understanding way that’s unique to AI chatbots. Suleyman entered a back-and-forth with the AI, in which Pi seemed to try to gently guide Suleyman to a decision. Suleyman eventually decided to stay home on the weekend and work on his pond project.

Once that decision was made, Pi accessed information from the internet and briefed Suleyman on the tools and supplies he would need for the pond, then presented a step-by-step overview of how to do the job.

And, like a human friend would, Pi is likely to bring the subject back up later on, perhaps asking Suleyman, “Say, how did that pond project go last weekend?”

[Photo: courtesy of Inflection AI]

Suleyman’s cofounders are LinkedIn founder Reid Hoffman, and Karén Simonyan, who led DeepMind’s research on large-scale machine learning models.

Suleyman says Inflection has one of the strongest teams of AI talent in the industry. He says the company employs the development co-leads of DeepMind’s Gopher and Chinchilla models, the project leads/co-creators of Google’s LaMDA and PaLM LLMs, and the co-leads/co-creators of OpenAI’s GPT-2 and GPT-3 models.

“We’ve also raised large amounts of capital, which allows us to train extremely large models on our own supercomputing clusters,” Suleyman says. Indeed, the company has raised $940 million so far, according to Pitchbook data, with a valuation of $1.23 billion as of April 28.

[Photo: courtesy of Inflection AI]

At present, the costs of building and training large language models are too much for most companies, many of whom have no choice but to access generative AI by tapping into LLMs that are now owned and hosted by just a small group of well-monied AI companies (such as OpenAI). It’s unclear if, in the future, Inflection AI will decide to make its language models available via an API, as OpenAI does. But for now, the company has the scale and talent to continue iterating on an LLM that’s purpose-built to befriend and support humans, adding a new twist to the generative AI mania of 2023.


Leave a Reply

Your email address will not be published. Required fields are marked *