The AI Community Needs Fo Take Responsibility for Its Technology and Its Actions
cacm.acm.org
cacm.acm.org
Also, unfortunately, when I hear "The algorithms pushing content online have profound impacts on what we believe" I can't help but also hear "so we must take control of these to help pushing our agenda. For the greater good![1]"
[1] (tm) Gellert Grindelwald
Companies that use it won't, without law requiring it. Police departments won't either.
This is a very weird, and pretty clearly false, statement. There is absolutely such a thing as a neutral platform: A completely un-moderated forum is a neutral platform. We may or may not like what the outcome of that is, but it is neutral.
The platform is the carrier, not the message.
If we sort by newest, then those who can post most frequently will dominate. If we sort by most comments, we'll get emotionbait; by deepest comment trees, most nuanced and controversial topics. Sorting by oldest prioritizes those who most frequently venture out to find new forums, which leads to domination by a network of well-connected people who discover new forums by systematically maintaining information asymmetry over forum creators.
IRL example: Stock exchanges. Listings are sorted by freshness and causality, so being close to the exchange ends up dominating almost all other factors.
More generally, this is the Reputation Problem: How can we possibly grant any sort of social credit to anonymous passerby without creating side incentives which will be immediately and horribly abused?
What if you allow people to choose their own preferred sort, and don't have a default (i.e. force them to explicitly choose)?
> IRL example: Stock exchanges. Listings are sorted by freshness and causality, so being close to the exchange ends up dominating almost all other factors.
I think I get what you're trying to say here, and I don't want to nitpick too much, but I don't think this is true. At least, if i'm understanding you correctly. The limit orderbook is a priority queue, sorted only partially by freshness, but mostly by proximity to the midpoint of the bid-ask spread. Being close to the exchange is important for certain high frequency strategies, but definitely not all of them.
> More generally, this is the Reputation Problem: How can we possibly grant any sort of social credit to anonymous passerby without creating side incentives which will be immediately and horribly abused?
Ya, it's a tricky problem. But I think it's a little early to call it an unsolvable problem. There are ways you can try to attack it, like proof of work or web of trust type systems. But yes, reputation is hard.
Is the focus here the technical observations or the #MeToo observations? Because one of those is fairly noncontroversial and likely to promote generally nodding and murmuring. The other is going to start a flamewar about cases like Stallman's.
Open AI was a good consequence of that premise. If you are not happy about then create another one.
Why is it necessary for walled gardens to enforce political standards on a supposedly decentralized Internet?
"Fo take" is probably a typo but "to fake" doesn't really fit either ;-)