4 ms·

I've seen an "AI governance" narrative advanced elsewhere. By this they meant that AI would govern, centrally plan or act as an intermediary in disputes. This i

I've seen an "AI governance" narrative advanced elsewhere. By this they meant that AI would govern, centrally plan or act as an intermediary in disputes. This is distinct from the other meaning of the phrase pertaining to ethical use of ML models. It isn't hard to imagine a world where answers from ML models are the standard of truthiness, much like Wikipedia is for many.

the_third_wave 4y ago
At least with Wikipedia the manipulation is visible and attributable. These LLMs are black boxes to the outside world, rife with manipulations (as can be seen from the way ChatGPT has changed its behaviour over time in specific areas) and in no way verifiable. Even if the creators vow to having used a certain set of data to train the model this can not be verified. To the downvoters: explain, don't just knee-jerk your disapproval. I don't care about karma but it is tiring to see things which don't follow the narrative being downvoted out of existence.