Comradeship // Freechat

for me the arguments would be mostly negative because:

1. using it does not train your research skills
2. using it does not train your creative and academic writing skills
3. it is often just wrong when synthesizing text

so to me those are major cons in an educational context some positives would perhaps be:

1. it is useful as a phrase bank, as it can quickly give ideas on how to put words together.
2. it is alright at giving direction when starting research, sort of like wikipedia

thats all i can think of so far

My preferred way of thinking about these chatbots is that they're effectively just on-demand peers with quick Google skills to chat. Just like humans they can be confidently wrong a lot or have incomplete information or presentation, but also just like humans they can help you explore your ideas and give you quck insight.

Besides all the technical cons (blatant disregard for copyright law and it being randomly racist sometimes), I don't think they're particularly bad. You just have to keep in mind that they're about as trustworthy as your local arrogant lab intern. Usually you're already required to source your claims on higher education work anyways.

Main issue right now is that the current favourite implementation seems to be specifically trained to almost never admit to not knowing something.

For more general information on llms I recommend this blog. https://simonwillison.net/ The post about prompt injection is especially good.

It's been awhile since I skimmed it, but this article is a good source on the topic.