Why Research Talent Still Costs So Much
One of the rarest resources in artificial intelligence—perhaps even more than Nvidia servers or training data—is the people who make it. Founders tell me that the shortage of talent that can develop and improve language and vision models is still extreme, two years into the conversational AI boom.
That’s demonstrated by recent developments at startup Voyage AI, which Databricks and Snowflake both recently discussed buying—mostly because of the value of its AI talent, said a person briefed on the deal talks.
Voyage develops embedding and reranking models that can reduce the mistakes (hallucinations) AI-powered applications make. There are lots of free, open-source embedding models, developers tell me, but not a lot of people with the talent that Voyage employs.
For that reason, in the talks with Snowflake and Databricks, the startup was negotiating to be acquired in the $250 million to $300 million range, according to the person briefed on deal talks. That’s high for a year-old startup, especially one that was generating just $500,000 in annualized revenue last fall. That means it was seeking a valuation multiple of more than 500 times forward revenue—expensive even by today’s frothy standards.
Related Knowledge Center: Artificial intelligence
On top of a failing business model, newsrooms must navigate the powerful — and not fully reliable — new capabilities of artificial intelligence and the deployment of large language models.
State of the business model | Copyright and intellectual property