Home › Cosmetic Science Talk › Formulating › AI Friend or Foe to the Cosmetic chemist?
-
AI Friend or Foe to the Cosmetic chemist?
Posted by Perry44 on July 17, 2025 at 4:28 pmJoin us for this IFSCC hosted debate about the benefits and downsides of AI and it’s impact on the job of a cosmetic chemist. I’m honestly torn on the issue but will be arguing the “benefits” side.
https://event.webinarjam.com/register/159/oxw89spqWhat do you think? Are these new AI tools good for cosmetic chemists or bad?
event.webinarjam.com
AI in the Lab: Friend or Foe to Cosmetic Chemists?
AI in the Lab: Friend or Foe to Cosmetic Chemists?
Graillotion replied 26 minutes ago 5 Members · 13 Replies -
13 Replies
-
From the point of view of just using Goofle… Ai is as worthless as tits on a boar. It simply pulls material from Mommy blogger sites and presents it as accurate. Which of course it is not.
I realize that is not the aspect your question is directed at…but it is sending beginners and sub-beginners down a treacherous and slippery slope. If you thought it was bad before….you ain’t seen nut’n yet!
-
Again on the negative side - there are hundreds of publications - usually from academics - that would have some of the silliest preservative systems useful.
-
I only used a couple of tools for formulation: Mini tab, and GPTs.
With the former, I heard from different disciplines that younger professionals rely more and more on this tool and neglect formal training and knowledge, which I find more like a disadvantage since that tool needs to be feed, and its results interpretation will lack the perspective that knowledge and practical experience provides.
With GPTs, I find that you need to e really precise, to the point that you really need to know your stuff for it to be useful (else, it’ll lie right to your face). Also, when reviewing evidence, GPTs might often miss critical papers/reports.
-
I’m aware Mini tab is not an AI tool….it’s just my monthly rage against it 🤓
-
-
-
Thanks guys! Hopefully, you get a chance to tune in.
I’m taking the “it’s great” side but that is not exactly how I feel about the topic. I do see the real problems with it.
I will add however, in my testing it has produced some surprisingly good starting formulas. But you do have to steer it away from it’s terrible suggestions for preservation.
-
I enjoyed this one today:
Man who asked ChatGPT about cutting out salt from his diet was hospitalized with hallucinations
nbcnews.com
Man who asked ChatGPT about cutting out salt from his diet was hospitalized with hallucinations
A case study in a medical journal reported that the 60-year-old man replaced sodium chloride with sodium bromide after consulting the AI bot.
-
I am a software developer. I perceive AI as a great automation tool that can speedup your work. And like with any automation tool - you have to be smarter than the tool to use it efficiently.
There are a lot of code online, so AI had a huge database to train on. And yet the results even for coding are suboptimal, so you have to use with caution anyway. It is a google on steroids, nothing else.
What about formulation - there is no huge database to train on. Skincare formulation is mostly about experience rather than . The topic is too complex and the training data available is too sparse. Therefore AI can’t be trusted in skincare formulation from scratch.
What it can do then? It can pitch you ideas (marketing, labeling, product types, famous brands etc). It can do basic school chemistry math (e.g. how much NaOH I need to neutralize 10g of lactic acid 80% to pH 4.0). But nothing more.
Recently reddit launched AI to query their database. I suspect it will be a good tool for marketing research.
-
I agree that there isn’t a huge database of formulation to train on, however, there also isn’t a huge variety of formulas. Of the top 10 body washes, 8 of them use a blend of Sodium Laureth Sulfate and Cocamidopropyl Betaine. I think you’ll find similar results for most any category of personal care product. The ranges of use of all the ingredients is pretty easily obtained as every supplier posts a “use range” for every ingredient they use.
I agree AI isn’t going to give you a complete finished formula (yet) but it can get you a pretty good starting formula that you merely have to tweak.
-
-
In science AI risks profound BS..
A post on Researchgate raised my suspicions so I read a university educator’s most recent a published article. Checked a couple of references that seemed kind sketchy. Neither existed (not in Google Scholar or in the journal he claimed for them). Went on to find most of the references were fictional, and one that existed was misquoted. AI check of its abstract at three sites - all concluded it was AI generated.
Haven’t decided what I’ll do about it - it’s not in my field. BUT do not take AI as accurate.
-
The problem of research papers being AI generated is just starting. I’m not sure how to solve the problem.
-
Compounding factor - the journal is a Chinese predatory POS.
-
-
I think the final verdict is in!
I had long suspected this…but just needed someone to confirm it. 😅 I often end conversations on Facebook on various and sundry topics with: “I am old school and still have a functioning brain…. ” And I guess I was more right than I ever imagined. Enjoy the read, and confirmation of what I am sure you already knew!
Summary:
“MIT has completed the first brain-scan study on ChatGPT users, and the results are striking but perhaps not that surprising. Instead of strengthening the mind, long-term use may actually weaken it.
Over four months of testing, researchers found that 83.3% of participants using ChatGPT could not recall a single sentence they had written only minutes earlier, while non-users had no such memory issues. Brain connectivity scores dropped from 79 to 42, a 47% decrease, marking the lowest cognitive performance across all groups. Even after stopping ChatGPT use, these participants continued to show reduced engagement compared to those who never used AI, suggesting a lasting weakening effect rather than simple dependency.
Educators also noted that essays written with ChatGPT, while technically correct, were often robotic, soulless, and lacking depth (surprise, surprise.) While ChatGPT makes users 60% faster, it reduces learning effort by 32%. Interestingly, why are students wanting to use AI so much for these tasks? Are they not interested in education? Would they rather be doing something else?
The best outcomes came from those who began writing on their own and only later added AI. They retained strong memory, brain activity, and overall performance. (Likely because AI was used as a supplemental helper?)
This isn’t about avoiding AI altogether, but to use it wisely. Let it assist your thinking without replacing it. The MIT study shows that how we use these tools today will shape the strength of our minds moving forward.”
media.mit.edu
This study explores the neural and behavioral consequences of LLM-assisted essay writing. Participants were divided into three groups: LLM, Search Engine, and …
Log in to reply.