Design-Expert (Stat-Ease) vs ChatGTP
Roman Shustov
R&D manager/Development of paints and coatings/Design of Experiments
I wonder how to implement chatGPT for analyzing data in chemical industry. I spent a week “talking” with chat GPT and here is my brief review.
Firstly, I asked for to create a model for 6 components for the next analyses afterwards and Chat GPT did not do it typing ‘please use DOE for this purpose’, after several attempts I gave up. ??
Secondly, I created a table with 6 components in range 0-100 and a parameter depending on the components with particular logic (+, *, /, ^). To make it easier I put 100 runs in the table which was obviously enough to create a robust model. For instance, DE required around 30 runs for the model and working with chat GPT I realized the following nuances: inappropriate interface for analysis in general, can require knowledge in coding and advanced math, poor prediction.
Finally, I am still in the process however, DE looks faster now in any sense and more accurate. ?Maybe I did not manage to figure out how to use chatGPT in this manner so if you have some tips, it would be great to see in the comments. ????????
Head of the chemical laboratory (Development of passive fire protection) - DEKO (ООО "ДЕКО")
1 年Great job! Maybe chatGPT is still weak for modeling systems with 6 components. It would be interesting to see how it models with fewer components. P.S. Chemists, unlike journalists, can still exhale. СhatGPT does not threaten us. Bye.