We have intrinsic negativity bias and intrinsic tribal nature. Any social interaction system we design should account for those biases so it can either avoid creating feedback loops or exploit creating feedback loops - depending on the incentives of the developer. We are also social and where incentives are right we want to work together and even be altruistic at times. Systems could also leverage those instincts and incentives. But I'm guessing the tribal/negative ones are stronger and maybe easier to poke with a stick to stimulate...?

Alternate take... https://www.youtube.com/watch?...

Maybe through a Bluetooth connected smartphone app?

My guess is that infection rate is underreported relative to deaths so the mortality rate is NOT nearly as high as 8%. Deaths are more likely to be reported for obvious reasons. Someone might get sick but never even go to the hospital because for many people the virus is uncomfortable but doesn't require hospitalization. Then again, death rates could be intentionally underreported. So who knows, really....

I don't know if more characters would help. Facebook allows longer responses and I think the lack of meaningful conversation is still there. I don't know if people *want* meaningful conversation on these platforms. It seems like mainly people want the instant gratification of quick likes/retweets. Anyway, that's my hot take. :)

I think the problem is not when people attack the message but attack the person delivering the message (i.e. death threats, doxxing, etc.). Whether it's a child or not. Happens on all sides and it's a cheap tactic.

What about methane? My understanding is that it's even better at trapping heat and there are vast amounts in permafrost and in the ocean floor that could be released with heating? Not really an expert - but I do think it's not JUST CO2 that's a problem, right?

As I understand things - I think you're mixing up censorship with violation of the first amendment. First amendment protection is only related to government. Censorship is a more generic concept that can be applied anywhere. In this case YouTube is definitely censoring. But it's not a first amendment violation because YouTube is not a governmental body.

I don't know if I agree that you can't identify a bias after the fact. But anyway, it's certainly better to try as hard as possible to have an unbiased dataset for learning to begin with. But it's probably also really hard because you don't know what you don't know.

it doesn't have to know the demographics to have a bias. if the learning dataset the AI uses is biased, then the AI will be biased. here's an overly-simplified example. let's say the learning dataset is all from kids from New York. And "good" essays by kids from New York (based on human evaluation) happen to often have the words "foo" and "bar" in them. So the AI thinks essays with the words "foo" and "bar" in them are "good". Now let's say "good" essays by kids from New Jersey (based on human evaluation) happen to often have the words "goo" and "jar" in them. But those essays were not put in the learning dataset! Now the learning dataset is biased towards kids from New York. Even though the AI has NO CLUE what state the kids are from. Make sense?

iOS's "Do Not Disturb" already does almost this. With DND you can set it to only allow calls from contacts. The only new thing here is that apparently it will let calls from people it THINKS you know based on emails etc. but you haven't created a contact yet. AFAICT.

Something is happening that we don't understand! Aliens. Yup. Aliens.

Why should they care? They're creating a market. For example, eBay doesn't review listings for quality. They just facilitate transactions. Same with KS. They're not a reseller. They are a broker. Getting mad at KS is like getting mad at eTrade that your stocks are tanking.