Sure scientific application might be useful, if it was able to think straight, apply critical and deductive reasoning, prioritise data sources for reliability etc. I think there is a long way to go before that but the progress on the language front is an improvement on previous generations of chat bots I don't deny and I am sure it will be developed and get better.
Chat GPT is a cut down version of GPT-3 if I understand correctly. But constraints, data sources and training are not really the issue, I dont think it thinks very well at its core and that is why training is so necessary and how bad training can derail it so easily. It needs to be propped up, can't stand on its own hind legs as it were and train itself.
I tested it out on a medical topic and it failed to deliver but was very polite about it.
I asked it about comparing nutritional values of food ingredients and it was really helpful. Its limited in what it can do but can be a useful tool for home economics if we understand and work within its limits! That is about where its at.
I think what it really offers as a next step is the next level of computing language. Meaning, we started with Babbage and Turing, punch card and tape binary, later hexadecimal machine code, then third tier languages like Fortran and Basic then object oriented compilable languages like C++. I think this is the beginning of programming a computer with words, just by telling it what you want in everyday English, which could be useful.
I think it would be fair to call it an advanced interface, I dont see intelligence yet.