An AI startup that aims to complement workers rather than replace them has raised seed funding of $480 million from some of tech’s biggest names, already achieving a $4.48 billion valuation.
San Francisco-based Humans& launched just three months ago and has attracted backing from Nvidia, Amazon founder and former CEO Jeff Bezos, and Google Ventures, despite only having around 20 employees.
The seed round was led by venture capital firm SV Angel and Humans&’s co-founder Georges Harik, with other investors including Emerson Collective, Forerunner and S32.
At this stage, what Humans& will offer is short on detail, although its newly launched website described the initiative as a “human-centric frontier lab.”
“At its best, AI should serve as a deeper connective tissue that strengthens organizations and communities,” according to the website. However, this will require “rethinking everything” about how models are trained and how people interact with AI, the company acknowledged.
Among the Humans& team tasked with doing the “rethinking” are experts who hail from Anthropic, xAI and Google and work alongside the company’s chief executive Eric Zelikman. It would seem that the caliber of the hires plus its noble aspirations have stirred up plenty of interest.
Among clues provided by Humans& as to how it will achieve its goal is a pledge to focus on “innovations in long-horizon and multi-agent reinforcement learning, memory and user understanding.”
Long-horizon tasks are typically complex activities that require an AI agent to maintain focus over a lengthy series of interactions with sustained reasoning and planning, going well beyond the capabilities of a simple question-and-answer bot.
By using multiagent reinforcement learning, Humans&’s AI models will be able to work with other neural networks, with agents receiving feedback on how well they perform tasks. That feedback is then used to improve output quality.
In essence, Humans&’s tech will be trained to be more inquisitive or interactive than we are accustomed to at present, and as a result, will improve collaboration and communication. Zelikman told Reuters: “The model will coordinate with people, and other AIs where appropriate, in order to allow people to do more and to bring them together.”
Whether that is enough to back up Humans&’s “human-centric” credentials and allay fears that AI is going to claim millions of jobs remains to be seen. But enterprises will get an idea of what’s in store shortly, with a first product planned for release later this year.
