This article originally appeared on Business Insider.
Recruiting AI talent could be a tough feat for some firms.
Aravind Srinivas, the founder and CEO of Perplexity, an AI-powered question-and-answer engine, described his interaction with a job candidate that shows how hard it could be to rent individuals with generative AI skills.
“I attempted to rent a really senior researcher from Meta, and you realize what they said? ‘Come back to me when you will have 10,000 H100 GPUs,'” Srinivas said on a recent episode of the business advice podcast “Invest Like the Best.”
H100 GPUs confer with Nvidia’s highly coveted graphic-processing units that tech giants like Meta, OpenAI, and Google use of their data centers to power and train their AI chatbots.
“That would cost billions and take five to 10 years to get from Nvidia,” Srinivas said.
Limited funds, combined with a chip shortage, means Perplexity, which powers its Q&A engine using GPT-4, has found it tough to seek out the talent required to create a big language model, Srinivas said.
Srinivas said it’s difficult to get employees to depart an organization where they “have an ideal experimentation stack and existing models to bootstrap from.”
“You need to offer such amazing incentives and immediate availability of computing. And we’re not talking of small compute clusters here,” he said.
The CEO added that even when smaller firms like Perplexity are capable of get Nvidia’s chips, they’ll proceed to fall behind because AI is developing so quickly.
Srinivas said AI talent at major tech firms “can have already made the next-generation model.”
“They’re like, ‘Look, the world has modified, I’m already in the following generation,'” he added. “‘I’ll come when the following version of the model is finished training. This time, you come back to me when you will have 20,000 H100s.'”
Srinivas and Meta didn’t immediately reply to a request for comment from Business Insider before publication.
There’s been a rapid uptick in interest in AI skills like machine learning and data engineering since OpenAI launched ChatGPT in November 2022. Companies like Amazon, Netflix, and Meta have offered salaries as high as $900,000 a yr to draw generative AI talent, and non-tech firms across the education, healthcare, and legal sectors have been seeking to fill roles with employees who know easy methods to use AI.
Srinivas believes that employees need skills beyond the flexibility to create AI models that generate desirable outputs.
“You need to post-train them and address the long tail of issues you get on serving a product,” the CEO said.
Post-training expertise, like knowing easy methods to reduce a chatbot’s factual inaccuracies, is a vital skill that employees from a wide selection of digital industries can learn quickly, Srinivas said.
Leaning into that skill set, he said, will help AI firms like Perplexity stand out in a sector dominated by Big Tech.
“You have tremendous advantage to create loads of value,” he said about post-training skills. “And we’re focused on that.”
