Matt Bevan

But at Palantir, they were quickly coming to a realisation.

They had been building software that was to be used by humans, but the way they built it was also perfect for large language models.

Palantir's Chief Technology Officer Shyam Sankar said that LLMs, large language models, weren't just great at reading their data maps.

but needed their data maps in order to be reliable.

It's like, wow, you actually cannot unleash the value of an LLM without these things.

They had, purely by luck, spent 20 years building a system that organized data in a way that could be intuitively read by the new wave of AI systems.

And it was laid out in an ideal way for the AI system to produce reliable information.

They developed something called the Palantir AIP, which stands for Artificial Intelligence Platform.

AIP is your AI operating system.

Now, individual users don't need to look at the marbles on the table at all.

They can just use AIP, which sits on top of the marbles and looks at them for you, making connections far faster and more nuanced than any human could.

Instead of handling a few marbles, it can handle millions all at once.

To use it, all you have to do is ask it a question.

The human operator can ask AIP in plain English to deploy surveillance drones.

Then the operator can ask it to look at what options are available in the area to destroy that tank.

Until pretty recently, executing this kind of operation would have required big maps being rolled out on tables and little wooden figures pushed around on top.

Phone calls right up and down the chain of command.

AIP allows it all to be done as easily as asking ChatGPT to plan your holiday to Queensland.

Not only can AI be used to operate the Palantir software, it can also be used to design it.

What once took software engineers years of observing operations and writing code now takes weeks.