It’s the new blueprint for Big Tech companies scrambling for dominance in artificial intelligence: Hire the cofounders of a high-profile AI startup along with some of its staff.
The latest example came on Friday when Google said the cofounders of AI chatbot startup Character.AI and some of its research team would join its already substantial AI efforts. The announcement, which included Google licensing Character.AI’s technology, sounds a lot like what Microsoft did in March when it hired a big part of the workforce at AI startup Inflection, including CEO Mustafa Suleyman, and what Amazon did in June with AI company Adept.
If three is a trend, there is clearly something trendy happening in the world of AI startups—and it may not just be the increasing number of Big Tech companies agreeing to absorb AI upstarts without actually buying them (a full-on acquisition would raise regulatory concerns, after all). Instead, it may be that the AI startup era itself, which has soared wildly for over two years, is beginning to implode.
Character.AI was part of the very nascent AI chatbot craze when it debuted in 2021, cofounded by former Google researcher Noam Shazeer, who became the company’s CEO, Daniel De Freitas, who was its president. The two co-authored the seminal research paper “Attention is All You Need,” which helped launch the Transformers architecture underpinning OpenAI’s chatbot ChatGPT and other large language models.
Character.AI’s technology let users chat and role play with real-life or fictional characters, from Queen Elizabeth to Draco Malfoy, or create customized AI companions. But Shazeer and De Freitas’ goal was never just providing AI entertainment. In 2022, the company showed its ambition in a blog post, asking, “What if you could create your own AI, and it was always available to help you with anything?” Then, in an interview with Forbes last year, Shazeer said Character.AI was betting that its new models would bring its technology closer to artificial general intelligence (AGI), or when AI can perform an important task at least as well as a human.
In March 2023, Character.AI was among a number of LLM companies receiving eye-popping investments. It received a fresh $150 million in funding, led by Andreessen Horowitz, and a valuation of $1 billion, even though Character.AI had no revenue. In September 2023, Character.AI was reportedly in talks for yet more funding—rumored to be from Google—at a valuation exceeding $5 billion, but that investment never happened.
Around that time, things started getting more complicated for Character.AI. Facebook-parent Meta debuted a family of AI characters (it discontinued them last week while introducing a feature that let users create their own). Meanwhile, the list of AI companion startups continued to grow—the New York Times’ Kevin Roose, for example, tested six different ones a few months ago, including Nomi, Kindroid, Replika, Candy.ai, EVA, and, yes, Character.AI. Apparently, there’s not much of a business moat, or barrier to entry, when it comes to AI chatbot companions.
Character.AI also faced criticism early on for its lack of policing, including letting users create chatbots based on Adolf Hitler and Saddam Hussein. The company responded by tightening its filters, but still faced questions about the wisdom of letting teens make friends with chatbots or rely on them as AI therapists.
Funding is a constant worry for AI startups because they require eye-popping amounts to survive, due to the massive cost of computing power required to train AI models. And because most AI startups make little to no money, finding someone willing to write a check can be challenging.
For example, AI startup Cohere, whose CEO Aidan Gomez was co-author the Transformers paper with Character.AI’s founders, recently landed a $500 million investment amid questions about its sales. In June, Paris-based Mistral AI raised $645 million at a $6 billion valuation, despite just starting to rake in modest revenue.
OpenAI and Anthropic, the two biggest LLM startup heavyweights, are considered to have the strongest chance of achieving profitability. But even they face questions about ever making money, and last week, tech news site The Information increased those doubts by reporting that OpenAI is losing billions.
And that’s where all roads lead back to Big Tech, which is providing a safety net to some top startup founders and employees. Microsoft provided one to Inflection’s Suleyman, who now leads Microsoft’s AI efforts. Amazon, meanwhile, adopted the team at Adept. And now Shazeer and De Freitas are returning to their old haunt at the Googleplex. The era of AI model startups may be in decline, but count on Big Tech to continue racing to create more powerful AI models, increasingly with the help of AI startup talent brought in as part of broader deals.
CEO Daily provides key context for the news leaders need to know from across the world of business. Every weekday morning, more than 125,000 readers trust CEO Daily for insights about–and from inside–the C-suite. Subscribe Now.