Applying ChatGPT in demosLike me you've probably used Tablet Production and White Polymer more times than you care to admit and usually it's a good idea to make demos relevant to the audience you're presenting to. Sometimes I've pulled up research papers and online articles (often hard to find) to learn about a particular industry and their processes, read through them and try to identify what are the critical to quality parameters (ie. dissolution) and process parameters which impact it (ie. screen size, mill time, spray rate, coating viscosity, blend time) and their interactions (ex. spray rate*coating viscosity, coating viscosity*mill time). Then with the magic of changing column names and scaling values, you have a more relevant dataset.
I was curious to see if this process could be simplified as it can often be rather time consuming, and to jump on the bandwagon, I find ChatGPT (ie. new Google) does a decent job at delivering high level information to be used for this task. Obviously it needs sanity checking, but I think it can definitely be a useful tool in sharpening demos.
In the example below, I show ChatGPT capturing the process parameters of a powder coat process and their interactions. Curious if other SE's have been finding ChatGPT useful in their work. Please feel free to share.
ChatGPT describing powder coat process parametersChatGPT describing powder coat process parametersChatGPT describing powder coat process parameters
Re: Applying ChatGPT in demosFascinating discussion, here are some issues I’ve often found when asked to show JMP using an example from their industry:
- It runs the risk of opening a debate about the relationships in the data, which further distracts from our goal.
- There is something deeper behind the request such as resistance to change (an education/enablement need) or they may not be ready for data analytics or DOE (in which case we might be better off putting them in our nurturing or field marketing bucket until they are ready to engage).
I do think this is a great opportunity to dig a little deeper with some open questions to determine if there is an underlying issue that goes beyond needing to see JMP on “their” data. If not then great, but if there is something else lurking behind the request, then it is better to discover it early.
Re: Applying ChatGPT in demos@nick_shelton , great suggestion, and while it does not construct a full dataset, it does give a small sample which is already useful. Some random formulas and regression formulas with noise can fill in the rest.
Re: Applying ChatGPT in demos@mzwald That's an interesting use case, For you me you've just opened my eyes to a million different applications! Let's start with this one first: Have you asked it to provide an example dataset? "Give me 250 rows of data based on the typical process parameters of a powder coating process?
Re: Applying ChatGPT in demosI found this to be particularly true with academics from particular domains. Oftentimes demos based on interesting or engaging yet largely irrelevant topics were more effective for exactly the reasons you state @mzwald .
Re: Applying ChatGPT in demosA lot of wisdom in this response, Mark.Re: Applying ChatGPT in demosGreat idea Mark. Thanks for sharing.
Re: Applying ChatGPT in demosAs a side note, I've found there is a potential danger in making demos too relevant to the audience. If the demo is overly aligned to the audience's domain knowledge, they will be able to put the numerical values into context which can result in sidetracked discussions about the veracity of the data, distracting from the demo itself. From my experience, good to make the demo in the general ballpark of the industry/application but not too specific to the audience's exact use case. Of course, use best judgement what makes sense to do when.
Re: Applying ChatGPT in demosWow. Here's another example from something I was thinking would be less known (the bot got very close)
