3 ms·

Go back a hundred and fifty years and you didn't have audio or video of scenes anyway. You had a bunch of hearsay and unreliable information still, but at a dif

Go back a hundred and fifty years and you didn't have audio or video of scenes anyway. You had a bunch of hearsay and unreliable information still, but at a different scale. I think we'll get more "reputation-based" choosiness (not some sort of algorithmic reputation-score, more individually measured, and no, that won't be perfect for anyone) but I don't believe that authoritarianism is also a natural response. Yes, the scale of spam will be far higher than ever before. But I expect people to largely normalize/filter a lot of that out. The elderly will probably be most at risk, still, or even just those of us with habits that formed out of "Wikipedia is mostly reliable" or "random people on Reddit are generally trustworthy" etc, which will be exploited for a while before most of us move on. Most people have been mostly wrong about most things for most of history.

visarga 3y ago
A more practical way to deal with adversarial content would be a "filter" AI we can use like AdBlock to hide the nuisances. You can have your own preferences, not the ones prevailing on any platform you want to visit. It is possible to do this with current day technology. Pinbot (was on HN 3 days ago) has a transformer model for embeddings. It could also filter any topic or type of content we want, and the kicker - it runs all in the browser, private and no internet connection required We can just tick a box to avoid Elon, Bitcoin, or various kinds of hype and activism, it works by semantic matching, so it understands phrase variations. Or maybe we just want to skip the occasional cat videos. The solution to the current situation is end user empowerment, we can make our own AI filters.
- majormajor 3y ago
  I've seen this in some sci-fi stories, it makes a lot of sense, but I have a personal hunch that setting up a filtering solution on various feeds of mostly-crap is going to be too much effort for most people, and they'll just slowly abandon the crap-feeds for anything beyond the most trivial sorts of entertainment. And then you could see something like a move from commenting on Twitter about news articles linked out to going back to going directly to the source and discussing things with other "verified subscribers" (whether paid or otherwise verified) there. Even in a "free, but ad supported" world it wouldn't surprise me to see outlets push for more "I'm a real reader and commenter" verification to keep their ad rates up.
dredmorbius 3y ago
Photography dates back closer to 200 years, with "View form the Window at Le Gras" being created in 1826 or 1827. Widespread use is probably better dated to 1839, with the invention of the daguerrotype process, 184 years ago. Quibbling over specific dates notwithstanding, with the creation of photography, phonography (1887), cinema (also roughly 1887, though one could point to Thomas Muybridge's horse-in-motion series, 1878), audiotape (1928), etc., high definition renderings of images, sound, or video came to be equated with high fidelity. There's of course been a co-emergent tradition of editing or fabulation in such media, though computerised methods of media editing or generation take this even further. It bothers me when I see people describe CGI or GPT creations as "realistic" when what they actually are is closer to "plausible" or "high-definition". The period of human-generated evidence --- spoken testimony, first-hand or second-hand visual representations (Albrecht Dürer's Rhinoceros remains an exceptional example, so close in so many details, so off in others: <https://www.metmuseum.org/art/collection/search/388409 https://www.metmuseum.org/art/collection/search/388409>), were often backed by some form of physical evidence or documentation, often with provenance. This was of course far from perfect, but there are probably lessons for us and future us / our descendants in how to deal with potential fabulation of records or accounts.

visarga 3y ago

A more practical way to deal with adversarial content would be a "filter" AI we can use like AdBlock to hide the nuisances. You can have your own preferences, not the ones prevailing on any platform you want to visit. It is possible to do this with current day technology. Pinbot (was on HN 3 days ago) has a transformer model for embeddings. It could also filter any topic or type of content we want, and the kicker - it runs all in the browser, private and no internet connection required We can just tick a box to avoid Elon, Bitcoin, or various kinds of hype and activism, it works by semantic matching, so it understands phrase variations. Or maybe we just want to skip the occasional cat videos. The solution to the current situation is end user empowerment, we can make our own AI filters.

majormajor 3y ago
I've seen this in some sci-fi stories, it makes a lot of sense, but I have a personal hunch that setting up a filtering solution on various feeds of mostly-crap is going to be too much effort for most people, and they'll just slowly abandon the crap-feeds for anything beyond the most trivial sorts of entertainment. And then you could see something like a move from commenting on Twitter about news articles linked out to going back to going directly to the source and discussing things with other "verified subscribers" (whether paid or otherwise verified) there. Even in a "free, but ad supported" world it wouldn't surprise me to see outlets push for more "I'm a real reader and commenter" verification to keep their ad rates up.

dredmorbius 3y ago

Photography dates back closer to 200 years, with "View form the Window at Le Gras" being created in 1826 or 1827. Widespread use is probably better dated to 1839, with the invention of the daguerrotype process, 184 years ago. Quibbling over specific dates notwithstanding, with the creation of photography, phonography (1887), cinema (also roughly 1887, though one could point to Thomas Muybridge's horse-in-motion series, 1878), audiotape (1928), etc., high definition renderings of images, sound, or video came to be equated with high fidelity. There's of course been a co-emergent tradition of editing or fabulation in such media, though computerised methods of media editing or generation take this even further. It bothers me when I see people describe CGI or GPT creations as "realistic" when what they actually are is closer to "plausible" or "high-definition". The period of human-generated evidence --- spoken testimony, first-hand or second-hand visual representations (Albrecht Dürer's Rhinoceros remains an exceptional example, so close in so many details, so off in others: <https://www.metmuseum.org/art/collection/search/388409 https://www.metmuseum.org/art/collection/search/388409>), were often backed by some form of physical evidence or documentation, often with provenance. This was of course far from perfect, but there are probably lessons for us and future us / our descendants in how to deal with potential fabulation of records or accounts.