-
Predictive analytics
Hello friends,
A client of mine acquired a licence for a program that uses predictive analytics. He said he’s been able to cut down costs of few formulas in up to 40%, without the need of real formulators, and validating those changes (blindly) with consumer input.
Nevertheler, I know a completar of cases where the project chief wasn’t able to solve issues with his formulas even after 3 months of endless trials generated by the software, while a person who worked with me solved those issues in 3 attepmts.
My fear is that the software is used not only by my client, but by others, to replace knowledge with these programs that in practice, underperform in the presence of real knowlege.
I don’t know how to convince my client, and I’d like to also know your experience with these tools
-
7 Replies
-
I don’t think the programs are good enough to replace chemists quite yet but it’s not too far away. We’ll still need people to put formulas together though. AI can’t make a batch & robotics is much farther away. AI is not yet good enough to make a formula that you can go right to production
-
I understand what you are talking about. A non-technical person I work with beamed fizzily after he thought he “cracked the perfumery code” by asking ChatGPT a few questions, and then he was emboldened to make some big changes in his organization. But if he ran the GPT output by anyone with a trained eye they would have bat it down immediately.
Copy paste that dynamic over entire populations. Where a (currently) inaccurate GPT confidently answers questions and emboldens non-technical people or shallow thinkers to do things.
What program was it, if you don’t mind me asking? I’ve been pitched a lot of AI formulation helpers lately. Would be good to know what to sidestep.
-
@Perry44 I totally agree. One major flaw of AI is their dependence on data and how reliable that is, since it can’t discriminate for instance between good data and bad data. AI also doesn’t have access (yet) to many paid journals which is a major limitation.
@sagestudent What amazes me is how easy it is to convince someone in power (who very often don’t have technical training) about some new technology, and how eager they are to replace their technical people. Btw, the tool my client is using is Minitab (https://www.minitab.com/en-us/solutions/analytics/statistical-analysis-predictive-analytics/).
-
The software your client is using would appear to be totally useless for cosmetic product development. I’m involved in the development of AI tools for cosmetic chemists and AI is great for tasks for which you have a good dataset and routine/repetitive workflow.
As for AI replacing cosmetic chemists or being able to create effective formulas, that’s a long way off in my opinion. Where AI can be useful in product development is in competitive analysis of top selling products. For instance, you are tasked with developing a Moisturizer. You can use AI to evaluate the top selling, top rated moisturizers on the market to identify common features/ingredients that are most likely responsible for that superior performance and then use that as a guide in your own product development.
-
Thank you @MarkBroussard for your comments, I really appreciate them.
My client, for one project, actually built some data set from different trials where he modified the ammount of each ingredient ofnthe formula, to see where there is an optimal formulation, using Minitab. Nevertheless, the product was very simple and I don’t see that working in shampoos (where there are different and complex interactions) or even emulsions (since the products were made by a technician, manufacture is also another variable not taken into account). What’s your take on this?
-
You can do all of that in an Excel spreadsheet. Seems to me to be a waste of time.
-
Thank you @MarkBroussard
Log in to reply.