8 ms·
Very cool that these companies can scrape basically all extant human knowledge, utterly disregard IP/copyright/etc, and they cry foul when the tables turn.
by nyc_data_geek1 5mo ago
Very cool that these companies can scrape basically all extant human knowledge, utterly disregard IP/copyright/etc, and they cry foul when the tables turn.
- stavros 5mo agoYep, that is exactly what happens. It's a disgrace that their models aren't open, after training on everything humanity has preserved. They should at least release the weights of their old/deprecated models, but no, that would be losing money.
- copperx 5mo agoWe should treat LLM somewhat like patents or drugs. After 5 years or so, the models should become open source. Or at very least the weights. To compensate for the distilling of human knowledge.
- butlike 5mo agoAll extant human knowledge SO FAR. Remember, by the nature of the beast, the companies will always be operating in hindsight with outdated human knowledge.