I appreciated your post on Musk and I think it was important to emphasise the wider social and deeper personal context of his behaviour, especially given the reluctance/inability for other publishers to do so.
I think I could have easily dismissed it as ‘ah, more blowhard bs’, but placing it in the context of his history as an individual, his responsibilities as a business leader, and the communities he apes/values (e.g. Edgelord, Twitter) helped me appreciate the dangerous scale of his buffoonery.
So, thanks for writing it, even as I am also absolutely tired of the guy.
"So if it’s this good when we don’t really know how to use it yet, it’s reasonable to believe it will get there, probably quite soon."
I do not understand how people in tech have this level of apparently unfounded optimism after everything that's happened in the past 20 years.
OpenAI has said that ai could be very dangerous. Existing tech platforms have facilitated genocides and empowered dictators. But yeah I guess it's totally reasonable to assume this thing we don't understand will work out fine.
I can understand why this particular use of automation elicits a hopeful reaction though. I don't wish the job of sifting through the worst parts of the Internet on anyone. It would be good if we had effective content moderation tools that weren't cursed monkey paws so that humans didn't need to be exposed to that level of evil.
It's just that this ai stuff looks pretty paw-shaped at a distance.
Yoel Roth's questions are key, particularly around non-English and non-US cultural context. Those are aspects that are too often overlooked leading to okay-to-good moderation in English and huge gaps in other languages. That also feeds into his question about the DSA because the standard of moderation in European languages needs to be equally high and equally explicable in order to comply with it. And then we have all the *other* widely-spoken languages that don't fit into the DSA bucket.
I appreciated your post on Musk and I think it was important to emphasise the wider social and deeper personal context of his behaviour, especially given the reluctance/inability for other publishers to do so.
I think I could have easily dismissed it as ‘ah, more blowhard bs’, but placing it in the context of his history as an individual, his responsibilities as a business leader, and the communities he apes/values (e.g. Edgelord, Twitter) helped me appreciate the dangerous scale of his buffoonery.
So, thanks for writing it, even as I am also absolutely tired of the guy.
"So if it’s this good when we don’t really know how to use it yet, it’s reasonable to believe it will get there, probably quite soon."
I do not understand how people in tech have this level of apparently unfounded optimism after everything that's happened in the past 20 years.
OpenAI has said that ai could be very dangerous. Existing tech platforms have facilitated genocides and empowered dictators. But yeah I guess it's totally reasonable to assume this thing we don't understand will work out fine.
I can understand why this particular use of automation elicits a hopeful reaction though. I don't wish the job of sifting through the worst parts of the Internet on anyone. It would be good if we had effective content moderation tools that weren't cursed monkey paws so that humans didn't need to be exposed to that level of evil.
It's just that this ai stuff looks pretty paw-shaped at a distance.
I thought your recent post about the feckless Musk was fine.
Great stuff as usual!
Per Emily Bender: “Do not use synthetic machines if you care about the accuracy of the information.”
Yoel Roth's questions are key, particularly around non-English and non-US cultural context. Those are aspects that are too often overlooked leading to okay-to-good moderation in English and huge gaps in other languages. That also feeds into his question about the DSA because the standard of moderation in European languages needs to be equally high and equally explicable in order to comply with it. And then we have all the *other* widely-spoken languages that don't fit into the DSA bucket.