#ai discource

[feathershy]

"this is as expensive and bad as AI will ever be" is likely correct but optimistic take about the potential enshitification of AI services. Compare like google search or netflix which have got worse or more expensive.

[feathershy]

Given how context engineering and llm agents work, "inside your brain is a little guy looking at screens and issuing commands" is more true of AI than of humans. (where the raw LLM is the little guy)

[feathershy]

even in the area of AI videos, videos still supply significant evidence the same way a detailed description of events that reference particular people supply more evidence than a vague description. It supplies more claims that can be verified (and thus more opportunities for the fraudster to make a mistake)

like "are the trees in the background plausible for the location?", or "can we find the location on google street view?" or "are the electrical outlets correct for the country it claims to be from?"

it does mean that you should expect basic meta claims like "who took this video and where is it from?" before trusting that it is a literal description of reality.

In this way, the existence of AI video encourages people to give artists credit.

[feathershy]

AU discourse: "is it still okay to force actors to kiss for real when we could just have them stand close and use AI to do the kissing scenes?"

[feathershy]

if a superintelligence has so smart it can figure out how to credibly to commit to not killing all humans or destroying earth, then making peace with humans is likely worth it. (since it removes the small risk that the humans will succeed in destroying it or otherwise foiling its plans; and space is very big and its orthogonal goals can just as easily be achieved in the rest of the solar system and beyond).

[feathershy]

opinion: a significant portion of people's dislike of AI could be avoided if AI companies spent less time trying to make the AI seem human and let it be obviously robotic. It would make the AI seem less threatening, and cause less problems for institutions and norms that assumed that various stuff worked as proof of human effort.

plus, for AI art, making things that a human never could make is more interesting than making things a human can but cheaper.

Though I wouldn't support a regulation requiring such. Technology should progress even when it causes problems for institutions and norms. I just think the corporations are doing a marketing failure not making the AI sound like GLADOS.