|
|
By Opalesque: Howard Marks, co-chairman of Oaktree Capital, wrote his latest memo AI hurtles ahead, as a follow-up to his December 2025 piece asking whether AI represented a financial bubble. To prepare, he commissioned a detailed tutorial from Claude, Anthropic's AI model, and was struck by the quality of the output - personalised, logically argued, self-aware about its limitations, and even laced with humour. The experience shaped much of what follows.
Understanding how AI works
Marks begins by reframing how we should think about AI. It is not a search engine retrieving stored facts - it is a reasoning system. AI models go through two phases: training, in which they absorb vast amounts of text and learn how to think, reason, and synthesise ideas (analogous to how a human brain develops through experience); and inference, the operational phase in which the model responds to user prompts. Marks stresses that the quality of those prompts matters enormously - and that most people are currently underestimating AI's potential simply because they don't know how to prompt it well.
Can AI think?
This is a question he finds genuinely fascinating: can AI have an original idea, or is it merely very sophisticated pattern...................... To view our full article Click here
|
|
RSS
