No. of Recommendations: 11
> For example, I hadn't considered the headache for established search engines of AI chatbots populating the web with "truthy" explanations and alternative facts.
AI PhD here.
An interesting problem going forward is that large scale datasets of images and text are will be increasingly polluted by AI output.
e.g. all the 'look at this thing chatgpt wrote!' posts on linkedin, all the shared images from MidJourney, Dall-E etc on Facebook.
As a general point, 'wrong and ugly answers from early AI' will be a problem for producing training sets for future AI.
lux
