3 ms·
Too little too late. Anything pre-ChatGPT is already scrapped, packaged and mirrored around the Internet; anything post ChatGPT launch is increasingly mixed up
by TeMPOraL 3y ago
Too little too late. Anything pre-ChatGPT is already scrapped, packaged and mirrored around the Internet; anything post ChatGPT launch is increasingly mixed up with LLM-generated output. And it's not that the most recent data has any extra value. You don't need most recent knowledge to train LLMs. They're not good for reproducing facts anyway. Training up their "cognitive abilities" doesn't need fresh data, it needs just human-generated data.
- KnobbleMcKnees 3y agoPrecisely, which brings us back around to the question: why are social media companies really doing this? I think "AI is takin' ooor contents!" is a convenient excuse to tighten the screws further. Having a Boogeyman in the form of technology that's already under worried discussion by press and politicians is a great way to convince users how super-super-serious the problem must be, and to blow a dog whistle at other companies to indicate they should so the same. It's no coincidence that the first two companies to do this so actively and recently are both overvalued, not profitable, and don't actually directly produce any of the content on their platforms.