Long-reasoning AI as an investment pick, alongside doubts about long-horizon tasks

A post attributes to Eric Schmidt the view that, if he could invest in just one AI theme, it would be long reasoning: models that stay on one problem for hours or days. The same poster quotes Chamath Palihapitiya saying current AI systems do not handle long-horizon tasks or complex problems well. Palihapitiya argues that making AI work in highly complicated environments will require a “symbolic space” to guide an “embedded space.”