Are AI Promises Too Good To Be True?

You've probably heard a lot of buzz about AI lately – how it's going to change everything, solve all your problems, and generally make life easier. It's exciting stuff, but one of the biggest names in the AI game is now saying hold your horses. Alex Karp, the CEO of Palantir, a major AI company that works with governments and large businesses, has suggested that some AI firms might be overselling what their technology can actually do.

Karp specifically called out 'token-based pricing models' as being 'completely, irresponsibly oversold'. What he's getting at here is how some AI tools, particularly those built by giants like OpenAI and Anthropic, charge users based on units of data processed (like words generated or analysed). While these models are common, Karp argues they can lead to unexpected costs and complexity for businesses trying to adopt AI.

Palantir, on the other hand, is pushing its own 'open-weight' alternative. This generally means their software gives businesses more control over the data and the AI models themselves, rather than relying solely on services hosted by another company. It's a different approach that aims for greater transparency and control for the client.

For Australian small business owners, this is an important conversation. When you're looking at investing in new technology, you want value for money and clear expectations. Understanding these different approaches to AI – and the potential pitfalls of over-promising – can help you make better decisions about what AI tools might genuinely help your business without breaking the bank or causing headaches down the track.