3 ms·

> Couldn't it be the case [...] need different advise than people who know less about the topic? > for people who are not deeply into finance the best advice i

> Couldn't it be the case [...] need different advise than people who know less about the topic? > for people who are not deeply into finance the best advice is likely rather "don't do it!". Oh boy, more nanny software. This future blows.

aleph_minus_one 1y ago
> Oh boy, more nanny software. This future blows. I think this topic is a little bit more complicated: this is rather a balancing of the model between 1. "giving the best possible advice to the respective person given their circumstances" vs 2. "giving the most precise answer to the query to the user" (if you ask me: the best decision would in my opinion be to give the user a choice for this, but this would be overtaxing to many users) - Freedom-loving people will hate it if they don't get 2 - On the other hand, many people would like to actually get the advice that is most helpful to them (i.e. 1), and not the one that may answer their question exactly, but is likely a bad idea for them
- dogleash 1y ago
  Everything can always be more complicated, of course. For example: 1. The AI will never know the user well enough to predict what will be best for them. It will resort to treating everybody like children. In fact, many of the crude ways LLMs currently steer and censor are already infantilizing. 2. The users' benefit vs. "for your own good" as defined by a product vendor's financial interest is a scam that vendors have perpetrated for ages. Even the unsubtle version of it has a bunch of stooges and chumps that defend it. Things will not improve in the users' favor when it's harder to notice, and easier to pretend they're not malicious. 3. A bunch of Californians using the next wave of tech to spread cultural imperialism is better than China doing it, I guess. But why are those my options?

3 ms·

> Couldn't it be the case [...] need different advise than people who know less about the topic? > for people who are not deeply into finance the best advice i

by dogleash 1y ago

aleph_minus_one 1y ago
> Oh boy, more nanny software. This future blows. I think this topic is a little bit more complicated: this is rather a balancing of the model between 1. "giving the best possible advice to the respective person given their circumstances" vs 2. "giving the most precise answer to the query to the user" (if you ask me: the best decision would in my opinion be to give the user a choice for this, but this would be overtaxing to many users) - Freedom-loving people will hate it if they don't get 2 - On the other hand, many people would like to actually get the advice that is most helpful to them (i.e. 1), and not the one that may answer their question exactly, but is likely a bad idea for them
- dogleash 1y ago
  Everything can always be more complicated, of course. For example: 1. The AI will never know the user well enough to predict what will be best for them. It will resort to treating everybody like children. In fact, many of the crude ways LLMs currently steer and censor are already infantilizing. 2. The users' benefit vs. "for your own good" as defined by a product vendor's financial interest is a scam that vendors have perpetrated for ages. Even the unsubtle version of it has a bunch of stooges and chumps that defend it. Things will not improve in the users' favor when it's harder to notice, and easier to pretend they're not malicious. 3. A bunch of Californians using the next wave of tech to spread cultural imperialism is better than China doing it, I guess. But why are those my options?