Chat GPT

I typed a past LC HL history question. I would have given the Chat GPT answer a H3. It lacked the detail of a H1. Not bad for the early iteration of a new technology and undoubtedly it will get to a H1 in later versions but not good enough now so I'd be wary about the quality of answers I would get back on things I know nothing about.

I have found if you ask it to do certain tasks for you it will proclaim with complete confidence its answer is correct......The answer starts off with a the perky "sure" of a car salesman and then it proceeds to give a totally inaccurate answer

Anytime you challenge it, it will then say sorry and proceed to give you another inaccurate answer.....with complete confidence. You would absolutely need to know something about some topics to verify its answers otherwise you could look like a wally.....

Its still a very impressive bit of tech and it will no doubt get more refined with each iteration but the free version struggles with anything related to setting up shapes or spatially related but you wouldn't know it from the answers, I corrected it multiple times on a request and it still kept spitting back inaccurate answers, which it then described inaccurately...I found it interesting its unable to predict the exact length of its answers and it can give a wrong answer and then not be able to describe accurately that answer it just gave (in terms of numbers of items in it etc etc) - so its assembling on the fly and it doesn't appear to be storing info about its own answer so it can just look up the details and accurately give a description of what it has done - I assume that would take up too much memory and would mean it couldn't be offered to the public --[I'm wondering if there are rooms in places where these things have access to much more resources in terms of memory and what they can do]

I read that the more recent version was able to give quite a good answer to questions like how you would stack a nail, 3 eggs and a book etc in such a way that they would be stable

I found it much better on text but still prone to being generic, bland and inaccurate on occasion too.....as has been mentioned the better you get at prompting it the more the quality of the written answers improves.

I'm not sure what to really think about the tech yet but cant help feeling pandoras box has been opened, I think there is open source code out there so no doubt we will see multiple versions and applications of tech (with many more disruptive as well as probably very useful applications) like this pop up over the coming years

Did "AI" write this for you?

I think that's a fair point

In practice however I've noticed when its more about creativity etc...very little gets done and the student appears to know very little or doesn't tend to gain much understanding or be very skilled afterwards either.

I think the way education operates will have to change drastically to put a huge amount more responsibility/autonomy on the student if its going to be primarily team work, creativity etc etc ....the teacher would have to be more like a facilitator and its very much up to the student to engage/do or not do.....a lot of all the other functions such as the disciplinary stuff would need to be automated/or handled by another party

I also think there's still very much a place for written or oral assessment and actually being able to demonstrate your ideas/understanding on paper.....In fact I'd imagine AI means the only way to assess someone at the moment and be fairly sure that have gained some level of understanding/knowledge is via a written or oral test/examination or some form of live in the moment practical/demonstration