![](https://ainews.one/wp-content/uploads/2023/03/1678800971_0x0-802x799.jpg)
Traditionally, more complex software is harder to use. Nabeel Hyatt, general partner at Spark Capital, says that if Adept’s team (pictured) achieves its full potential, this may no longer be the case.
Adept
Chatbots rule the day in AI for now, but soon, Adept cofounder David Luan predicts, AI won’t just display unsettlingly human responses to typed queries, it will execute them. It will do what you would do with your computer for you. Granted such technology is still years away, but the speed of innovation in the AI space means we’re talking about two to three years, according to Luan — not decades.
A self-professed auto geek, Luan imagines a world where an engineer can ask an AI assistant to make a blueprint for a new car part and watch as it does exactly that, step-by-step, selecting the right software programs and inputting the necessary commands or code, with its human as copilot. Want to modify a portion of the design, test it in a car simulator software or send the blueprint to a manufacturer? In Luan’s vision, the AI would take care of all of that too.
Adept, barely a one-year-old startup with just 25 employees, has raised $350 million of venture capital after demoing a rudimentary version of such a digital assistant. Instead of generating text, like OpenAI’s ChatGPT, or images, like DALL-E, Adept intensely studying how humans use computers—from browsing the internet to navigating a complex enterprise software tool—to build an AI model that can turn a text command into sets of actions.
“A synthesizer lets a musician play sounds of every instrument without having to learn how to play every instrument. We want to build the same thing for computing,” Luan told Forbes.
General Catalyst and Spark Capital bankrolled the bulk of the Series B funding round, which was completed at a post-money valuation of at least $1 billion, per two sources involved with the deal. The core part of the financing was completed last fall, before ChatGPT kicked off a consumer AI frenzy, according to Luan. General Catalyst, the largest stakeholder in the new round, beat out seven competing term sheets to win its lead investment position, managing director Deep Nishar said.
A synthesizer lets a musician play sounds of every instrument without having to learn how to play every instrument. We want to build the same thing for computing.
Part of the investor froth comes from the cofounders’ pedigree — rare among the flurry of founders who have flocked to start AI startups in recent months. “A lot of people talk the game, but it’s very difficult for them to play the game,” Nishar said. “Have they actually built something like this before? What are their capabilities?” Luan, the CEO, was vice president of engineering at OpenAI before jumping to Google to lead its large model efforts. His cofounders Ashish Vaswani and Niki Parmar coauthored the Google research paper which invented the transformer, the AI breakthrough which represents the “T” in GPT. (Vaswani and Parmar recently departed to launch their own startup, according to a report from The Information; Luan declined to comment on the reason for the split.)
Their deep machine learning expertise enabled Adept to craft a working demo called ACT-1 less than a year after it raised $65 million from venture firms Greylock and Addition. At the time, it performed tasks similar to ChatGPT, answering simple questions. In the months since then, it’s become capable of performing complex functions like importing LinkedIn URLs into recruiting software. Advancements like these have helped Adept line up strategic investors like Microsoft, Nvidia, Atlassian and Workday all of whom market software that might someday benefit from its AI assistant. Adept is raising additional money through these business alliances at a not-yet-finalized valuation which is poised to be higher than $1 billion, two sources said.
Practically speaking, Adept’s ACT-1 displays as an overlay window on top of existing software like Google Chrome or Salesforce. A prototype is ready for desktop, but Luan said it will also be available on mobile in the future. The company has “commitments and targets on the revenue side” from a handful of partners, he added, but would not say when the public would be able to play with the AI assistant. “The degree of interest from the [corporate investors] I think shows some sense of the maturity [of the product],” he offered.
We don’t think about [AGI] from the perspective of how other companies think about it, which is replacing humans at valuable tasks.
Still, models for controlling computer actions are significantly less mature than their language model counterparts. That’s why a company that’s barely one year old can possibly justify raising hundreds of millions of dollars without plans for a massive hiring spree or big acquisition — training such models isn’t cheap. “We haven’t yet hit the optimization phase,” Luan said. “What we really want to do is train really powerful models that can do a lot, and then over time we’ll figure out how to make them cheaper and make them smaller.”
Adept’s rapid capital accumulation echoes the strategies of other model-building companies like Anthropic, Cohere and especially OpenAI, where Luan’s former boss Sam Altman raised a reported $10 billion from Microsoft earlier this year in a bid to outgun AI competitors. Like Altman, Luan too aspires to help the technology achieve artificial general intelligence, or “AGI,” a hypothetical AI system which is smart enough to make its own decisions without human input. One key difference: Luan’s AGI would be more boring and business-focused, with humans remaining in the driver’s seat.
“We don’t think about it from the perspective of how other companies think about it, which is replacing humans at valuable tasks,” he said. “We’re just trying to build the best AI teammate possible for everybody.”