An example I often think about is whether to “take more risks“. If I had to take a generalized stance it would be “pro“ - there is even a cool study to support this! https://www.nber.org/papers/w22487
But I, and I suspect everyone, has an acquaintance to whom you would strongly suggest taking fewer risks.
An example I often think about is whether to “take more risks“. If I had to take a generalized stance it would be “pro“ - there is even a cool study to support this! https://www.nber.org/papers/w22487
But I, and I suspect everyone, has an acquaintance to whom you would strongly suggest taking fewer risks.
Moreover, if it could be conditional on the category of opportunity/risk I could give even better advice. I’d advise more friends to switch jobs than start companies, for example. That’s a function of both the person and opportunity.
If I could see their financial situation I’d be even more confident, etc. But again that would make for a bewildering TED talk, so it never happens.
Actually I just made a connection: the article “the unreasonable effectiveness of data“ argues that ML’s success in NLP comes less from finding elegant principles and more from being able to see every little corner case of language use. But each edge case is more parameters, and there is simply no elegant compression - which will never fit in a witty interview answer. I hate it when it venture capitalists say “pattern matching“, but maybe the truly best they can do is have 10,000 highly specific rules of thumb for success
An example I often think about is whether to “take more risks“. If I had to take a generalized stance it would be “pro“ - there is even a cool study to support this! https://www.nber.org/papers/w22487
But I, and I suspect everyone, has an acquaintance to whom you would strongly suggest taking fewer risks.
Moreover, if it could be conditional on the category of opportunity/risk I could give even better advice. I’d advise more friends to switch jobs than start companies, for example. That’s a function of both the person and opportunity.
If I could see their financial situation I’d be even more confident, etc. But again that would make for a bewildering TED talk, so it never happens.
Actually I just made a connection: the article “the unreasonable effectiveness of data“ argues that ML’s success in NLP comes less from finding elegant principles and more from being able to see every little corner case of language use. But each edge case is more parameters, and there is simply no elegant compression - which will never fit in a witty interview answer. I hate it when it venture capitalists say “pattern matching“, but maybe the truly best they can do is have 10,000 highly specific rules of thumb for success