I thought this bit from the replacement device marketing was funny:
> And thanks to Gemini technology on Google TV, you can now get full summaries, reviews and season-by-season breakdowns of content, so finding your next marathon-watch just got easier.
They know it needs to be "AI powered" but they can't figure out anything that actually needs modern AI, so it's relegated to doing ordinary internet searches.
It's interesting, though, because there are existing sources for this content, like IMDB and wikipedia.
I wonder if the real point is more about selling ads while avoiding having to cut a deal with those existing sources. An LLM can essentially "launder" content so that, in general, it's hard to determine the sources for any given response. (There are plenty of individual examples where you can tell exactly what the primary source was, but those are the exception.)
I suspect the quality of the LLM-generated content will be worse than existing sources, but since the real point is to avoid sharing ad revenue, not providing good content to the user, that will probably be fine. It content doesn't need to be good, just good enough.
Welcome to the AI future! It's a lot grubbier than I expected.
> And thanks to Gemini technology on Google TV, you can now get full summaries, reviews and season-by-season breakdowns of content, so finding your next marathon-watch just got easier.
They know it needs to be "AI powered" but they can't figure out anything that actually needs modern AI, so it's relegated to doing ordinary internet searches.
It's interesting, though, because there are existing sources for this content, like IMDB and wikipedia.
I wonder if the real point is more about selling ads while avoiding having to cut a deal with those existing sources. An LLM can essentially "launder" content so that, in general, it's hard to determine the sources for any given response. (There are plenty of individual examples where you can tell exactly what the primary source was, but those are the exception.)
I suspect the quality of the LLM-generated content will be worse than existing sources, but since the real point is to avoid sharing ad revenue, not providing good content to the user, that will probably be fine. It content doesn't need to be good, just good enough.
Welcome to the AI future! It's a lot grubbier than I expected.