Feedback | SoME

On Generative AI

Audience:

Tags: generative-aiai-safety

Generative AI tools like OpenAI’s ChatGPT and Google’s Gemini have captured — and held on to — public imagination for a few years now. But how do these systems work, and why do experts prescribe abundant caution? Read on to gain technical intuition of the maths that underlies generative AI, and hence appreciate the risks of this technology, in particular the hazard of it being used to expedite propaganda machines. This essay has been written with a senior high school or undergraduate audience in mind. Less advanced readers are welcome to skip over the formulae and refer to the supporting text for the essential insights. More advanced readers are requested to forgive the (over-)simplifications :)

https://people.mpi-sws.org/~mvahanwa/publications/On_Generative_AI.pdf

Analytics

3.85 Overall score*

65 Rank

4 Votes

3 Comments

Comments

2.4

This was an interesting article. It is certainly clear that you’ve put in a lot of work into it, but I don’t think this is exposition.

It is a nice short read that explains the topic and concept well and puts it into context of usage, e.g. by discussing possible threats of false information trained into the AI. Some images, or even memes would have helped maybe.

Motivation and Clarity are good, since the intro is engaging and you were careful to explain the jargon. However I do not think that there was much novelty in the explanation, and the issues with GPT that the second half of the paper describes felt more like observations than insights (albeit this was admitted).

I think some sort of visualisation would have helped greatly to provide an a-ha moment. It is however well-written and engaging.