Palantir CEO Alex Karp delivered a passionate critique of the current artificial intelligence market during a recent CNBC interview, asserting that "something has gone completely wrong" with how AI is being sold. Karp, whose company is a significant provider of AI and surveillance technology to government agencies, expressed strong dissatisfaction with the prevalent model of charging consumers on a per-token basis.
"The basic view among enterprises in this country is 'I'm going to chillax and waste my time with tokens. I'm going to get no value, and they're going to get my IP,'" Karp stated, suggesting that this approach undermines trust and obscures data provenance. He advocated for AI to be implemented within closed environments rather than through a per-token system where data security and ownership can become ambiguous. Karp emphasized the need for clarity regarding data ownership, caching locations, prompt security, and data transfer protocols to rebuild trust in AI systems.
Karp posited that the true value of AI lies in the combination of a robust model, an application layer, and sufficient compute power, with the model itself being only one component. He questioned the logic behind charging for tokens if the underlying AI models are as valuable as they are made out to be. "Why are they charging for tokens if it's so valuable?" he asked, suggesting that AI models have been "completely irresponsibly oversold."
Touching upon the implications of AI in warfare, Karp expressed concern about outsourcing battlefield decisions to the consensus views of Silicon Valley. He described the idea as "effing insane" and stated that many enterprises are hesitant to voice public criticism due to the perception of AI advocates as "neurodivergent, crazy person that apparently is on drugs." Karp, who made a point to deny any personal substance use, insisted that his agitated tone represented the "voice of American Business" and highlighted a significant problem for the country.
Karp also urged other panelists and viewers to contact CEOs directly to gauge their sentiment on the state of AI, claiming that "They're twice as livid as me." Palantir has recently announced a collaboration with Nvidia to develop open AI models for U.S. agencies, a partnership that places the company at the forefront of AI development for critical infrastructure.