Hacker News Reader

Comment by thomashop | Hacker News Reader

A lot of decisions are not based on intelligence alone. A lot is about personal beliefs and tastes.

I've never totally understood this binary moment when AGI does "everything" better. How can one even define everything?

Our AI partner could be the most intelligent mathematician or researcher. That's great then we can bounce ideas off of them and they can help us realize our professional / creative ambitions.

Sure if our goal is to maximize profit then maybe we can outsource the decisions to an AI agent.

You can get a computer to create infinite remixes of songs. I haven't seen that replacing music producers doing the same.

View on HN

The I in AGI doesn't really mean intelligence as in "a mathematician has to be more intelligent than a janitor to do his job" *, it means the productivity equivalent of whatever human conciousness is; that is, to be able to have beliefs and tastes as well. And since it will have infinite patience, arguments such as "I prefer to have an actual human musician playing" is also up for persuasion.

When everyone can just press a button and have better music automatically generated, based on their exact preferences inferred from their DNA or an fMRI brain scan, what are your creative ambitions?

I'm obviously not talking about today's limited (public) AI, but far into the future, like in 5 years.

* whether or not that is actually the case is irrelevant

The music analogy doesn't work for me at all.

So much about music appreciation is about knowing the artist and for example knowing that what they sing about is shaped by their personal history allowing you to identify with it.

A lot about the appreciation of art is the process and the intention behind the artwork. The final image or sond is just one part.

I find it very difficult to define "better" when we're talking about art.

> A lot of decisions are not based on intelligence alone. A lot is about personal beliefs and tastes.

That's the easy stuff! That's the stuff ML has been successful doing for a decade or more.

Reply on news.ycombinator.com