Machine learning models excel at static problems, like analyzing radiological reports, but they falter when predicting future events, which is crucial for investing.

When investing, you predict future outcomes and sometimes take on uncertainty, which can lead to negative results that are part of the investing process.

It's extremely challenging to predict the future in the rapidly changing AI market; one-year plans are feasible, but three- and five-year plans are nearly impossible.

Investments in AI are currently focused on infrastructure for inference rather than developing new paradigms, which may lead to a mismatch between expectations and actual advancements in AI capabilities.

Investors focus on the traits of individuals rather than the ideas themselves, as predicting success or failure is extremely difficult.

Most investments won't succeed, and it's easy to overthink and miss good opportunities; I try not to let past failures cloud my judgment.

Investing in areas I know well can be tricky; I often find that I need to unlearn my assumptions to avoid missing out on good opportunities.

Our job is to build smart AI tools, but we can't predict where this technology will go; it's essential to admit that uncertainty.

Investment is a challenging business, as it involves creating value that cannot be physically touched, unlike tangible products.