I basically don't see any legitimate use for chatGPT in science, and this likely applies to its future successors as well.
Don't use it for writing, and definitely don't use it for research
It is the exact opposite of what we want in scientific info sources: it is centralized, black-box, citation-less, for-profit, proprietary, and methodological unlinked to empirical thinking
Also, critical and careful reading and writing are skills that are maintained only by constantly doing them. If you've outsourced either of those two activities to an algorithm, you will get worse at them.
I don't really have much to say about the hundreds of techies claiming it is replacing Google for them, or they are using it daily, besides the fact that these tweets reflect hilariously poorly on the tweeter.
I haven't really used it much at all since a test or two at the beginning, but if you really want a poem or R snippet, I guess that's fine....
I am very uneasy about the simultaneous crescendo in misinformation on that platform concurrent with the rise of better algorithms for generating misinformation.
@alexcc I do think it has legitimate uses. The careful reading critique can disqualify anything.
For example I believe publication search engines to be a better way to search through literature rather than going to the library and reading all the publications, but you may argue that reading through the journals trains your critical and careful thinking.
Of course, chatGPT is not a search engine, the stuff it spits out cannot be trusted at all and definitely you cannot rely on it to provide you with information to write in your articles.
Despite this, I believe it has its uses in research as well.
I have written about one of these uses a little time ago: provide an overview of a topic you know little about in order to pick up a list of keywords to search for.
I have used this method the other week and within 5 minutes I was reading a technical article about the exact technique I was searching for. 5 minutes prior I knew nothing about this technique and very little about this field in general. My search was rather naive and going through the normal route I probably would have had to take a quick look at a couple papers, then find some kind of review about all the methodologies in the field and from that find the article I was interested in. Doing all this, would likely have required me at least one hour.
@alexcc I rather like not having the limit. I'll try to be shorter.
Do you know any alternative? I'd be interested. I have tried iris.ai some time ago and it didn't perform great. There's connected papers which is nice, but requires field knowledge.
I'm not really sure how you'd make the distinction between useful for society and only on a personal level.
This appears to me as a legitimate and good use for researchers.
@rastinza I am (unironically) a huge fan of no character limit as well
I don't have any alternatives that I know exist rn, besides the old fashioned way of searching and absorbing information and making connections piecemeal!
Distiction: I think twitter is (still) enormously valuable on a personal level. But I am very skeptical that is valuable on a societal level. Hope this analogy makes sense.
@rastinza I enjoy how your server has no character limit...
"provide an overview of a topic you know little about in order to pick up a list of keywords to search for."
Tools to do that would indeed be useful, but I think that there can be alternative solutions specific to this use case, that could do a better job anyways
I'm thinking about this more on the level of "useful for society" as opposed to "personally useful under given circumstances"