An artificial intelligence startup from LinkedIn co-founder Reid Hoffman and DeepMind co-founder Mustafa Suleyman is rolling out a chatbot intended to serve as a supportive personal companion that gives friendly advice.
(Bloomberg) — An artificial intelligence startup from LinkedIn co-founder Reid Hoffman and DeepMind co-founder Mustafa Suleyman is rolling out a chatbot intended to serve as a supportive personal companion that gives friendly advice.
Palo Alto, California-based Inflection AI released Pi — which stands for “personal intelligence” — on Tuesday on multiple platforms, including on the web and via an iOS app. Pi is currently free to use and is intended for all kinds of interactions, including chatting about personal problems and getting advice.
“Many people feel like they just want to be heard and they just want a tool that reflects back what they said to demonstrate they have actually been heard,” Suleyman said in an interview Monday.
While chatbots such as OpenAI’s ChatGPT and Google’s Bard focus on productivity or search, Suleyman said Pi, which has been in the works for about a year, was built to be a supportive “neutral listener” that asks follow-up questions to keep a conversation flowing. The founders see the chatbot as offering a simpler way to get information that users care about.
Hoffman’s Greylock Partners is an investor in the company, and co-led Inflection’s $225 million seed funding round. The company declined to provide its valuation.
Pi is one of a number of chatbots that have come out in recent months that are based on large language models. Trained on huge swaths of internet content, they can then predict and generate text in response to a user’s prompt. The latest LLMs are increasingly capable when it comes to mimicking human writing, yet due to the nature of their training data, they can also surface many of the cultural, gender, and other types of biases that humans exhibit online.
Inflection’s plan for users to approach Pi as a personal helper is similar to how some people are already interacting with OpenAI’s ChatGPT by using it to supplement traditional mental-health services. Like ChatGPT, Pi is not meant to replace a human therapist and both chatbots will suggest users seek professional help in response to certain prompts related to mental health. Pi will also intermittently remind users that it’s a computer program.
“What we don’t want is for people to treat this as a romantic relationship,” Suleyman said. “This is really a companion, a safe, personal AI. You have to remember it’s an AI and not a human.”
New Pi users can conduct a conversation with 10 rounds of back-and-forth before they’ll be interrupted by a request to create an account in order to continue chatting. Inflection trains Pi on users’ prompts, Suleyman said, though personally identifiable details are filtered out.
Asked to help solve a personal problem – this reporter agreed to go camping this weekend but actually hates camping – it gave some simple, albeit useful advice. The chatbot helped clarify the problems with camping, such as sleeping in a tent, and came up with suggestions to make the situation more tolerable, such as using earplugs.
“It’s natural to be stressed when you’re doing something new,” the chatbot said.
Pi is meant “to help navigate your life and your relationship with others,” Hoffman said onstage at the Milken Institute Global Conference on Tuesday. During times when a relationship might be problematic, “you can debug it — you can talk to Pi,” he said. Hoffman compared AI to the steam engine in terms of its impact, and said he was optimistic society could solve the negative issues associated with the technology.
–With assistance from Sarah McBride.
(Updates with comment from Hoffman in the final paragraph.)
More stories like this are available on bloomberg.com
©2023 Bloomberg L.P.