Home › Cosmetic Science Talk › Formulating › General › Off Topic › ChatGPT not too shabby for home lab assistant
-
ChatGPT not too shabby for home lab assistant
Posted by Unknown Member on July 10, 2023 at 10:59 amThat’s right, I started using chatgpt to assist me in formulations. It’s helping my hobby expand greatly! Not 100% grade a cosmetic chemist, but still getting accurate information, troubleshooting, usage guidelines, comparability. I am surprised at how much chat gpt understands about skin care for not having skin or being programmed to be a dermatologist or chemist.
Anyone else been trying AI out for formulating assistance? If only it could weigh stuff out and mix things. Dear Elon Musk, where is my tesla bot????gordof replied 1 year, 5 months ago 8 Members · 20 Replies -
20 Replies
-
Yes, I tried to play with it. It gives totally unusable formulas. For example it suggested me to dissolve 2% salicylic acid in water. Obviously, it will not dissolve. And as a software developer I understand that ChatGPT is just a compilation of sentences from internet without actual understanding of the material and not always meaningful.
I am more into image generation networks. For example my avatar is not real photo, but generated with a neural network which was trained with my fotos.
-
Unknown Member
Deleted UserJuly 10, 2023 at 12:06 pmfor sure, not following the example formulas to the T but to just be able to ask for an example has been a huge help. I just started searching for ai geared more towards actual cosmetic formulation. It’s been a nice help.
How much does the generated picture look like you? AI art is also fun to play around with.-
How much does the generated picture look like you
It looks 80% like me in real life. Some details of my face are omitted and people say I look kinda “robotic” on this image, not a natural look.
-
-
A few weeks ago, we experimented with the AI model, and we found that it
still has a considerable amount of progress to make. We encountered
similar problems when seeking formulation advice. What was even more
concerning (and we thoroughly tested this multiple times) is that when
we encountered regulatory issues, the model often referred to
do-it-yourself (DIY) sources from the first page of Google search
results, without citing any technical references or official
regulations.-
Yeah, the problem with using it in formulating is that it has been trained on the Internet info which is heavily filled with DIY & ingredient marketing BS.
For example, I asked it for a suggestion for the best cosmetic preservative. It said phenoxyethanol & “natural” acceptable options. It didn’t suggest parabens or formaldehyde donors.
These are the most commonly used ones in industry but not the most common on the internet.
When asked further about parabens, ChatGPT propagated the same chemical fearmongering about them as any clean beauty inspired blogger.
If looking for starting formulas, places like ULProspector & Happi are still superior
-
You mentioned Happi….so I registered…but for the life of me cannot figure out how to find formulas.
They have a section called the ‘Formulary’ ….but can’t seem to find a ‘search bar’ …. so call only scroll down the myriads of uninteresting formulas? Any guidance?
-
No specific tips except to use it you have to just use the search function to find what you want. They don’t have a great website layout.
-
-
-
Unknown Member
Deleted UserJuly 10, 2023 at 2:41 pmI’m having a complete opposite experience with it here getting correct usage guidelines and no marketing bs. Said silicones are safe and my sample preservative is often germall plus. Even honest about which ingredients absorb into skin and which ones do not, even though marketing claims that they do like with certain hydrolysed proteins.
-
And…what ingredients absorb in the skin? I’m always a bit skeptical of that claim. My first undergrad Internship was doing QA on Transderm Nitro patches. I spent an entire summer in a windowless room doing calculus-based Pharmacokinetics. Transdermal absorption is rarely achieved. Were a product actually absorbed it would arguably be a drug.
-
It’s also possible that I used ChatGPT 3.5. When I ran the same query with ChatGPT 4.0, I got a better response. Although, they still didn’t mention formaldehyde donors as good options, but they did mention Parabens.
-
-
-
I doubt if Chat GPT will stop hobbyist, DIY, and get rich quick folks from coming here to ask for professional suggestions from the professionals…. ????
-
I think the biggest hurdle for something like Chat GPT when it comes to formulating is that much of the knowledge of the industry has not been written down in books or published online. We have traditionally been a secretive industry and most knowledge is locked up in companies or in the brains of veteran formulators who don’t write about it online. Thus GPT will mostly get trained on DIY / Hobbiest info which isn’t always exactly accurate.
-
Yes, I understand the our industry has had a veil of secrecy and still carries this veil to this day among us professionals. I have experienced it. However eventually you will hobbyist or DIY folks come here quoting CHAT GPT as the gospel without thinking about everything that went into answering their question.
-
-
-
I went for a company sponsored chatgpt training for marketing purposes. I find that it takes quite some refinements in my requirements to get the job done. I tested by asking for a skincare routine, it still took about more than 10 to 20+ times of refinement to get to what expect it to be.
It can probably help with the normal procedural or operational stuff but things that need specific requirements, its information would be limited, so we will need to know its limitations.
-
At Incosmetics EUrop This year they Presented an AI Tool wich is a labformulation and Documentation programm.
The results they showed in the Presentation wear very interesting the AI was able to Propose formulations that 80% of the time wear stable Formulations. If not Stable with small adjustments from a Traind Chemist got it stable. I my self think of myself as a good Cosmetic Chemist and i don’t have that kind of accuracy if i don’t use an already existing formulation out of our systems and adapt it for the purpose needed.
they just started to Collect data and even if I see some issues with who the formulation belongs to and so on, I think tools like this will change the Formulation further rather dramatically. wear you needed 3 Formulatore today you will maybe need 1 that does the fine Work and the adaptions for the customers based on the Formulations the AI will generate. and yes i know that the Formulations will probably all feel very similar and have a kind of generic formulation parameters than and i do not like this idea but i don’t think that many customers would recognize the difference between the two formulations that are decently developed from the skin feeling. The Biggest Decision making for the First buy is the advertisement and Scent of the product and if that fits the need of the customer if the formulation is ok the rebuy can be generated. And at the end that is what companies want to generate. If a Cream that is developed by a formulation Expert is always better in a way the Costoumer can Clearly destigueisch I am not sure at least if I look into the Formulations that some of my colleges are putting into the market (unstabel 2 Phased Fetty unpleasant etc.
I think this will in one or another way get to us in the next 5 to 10 years.
-
What software was it?
I personally think these AI formulators are a long ways from having a big impact on the job of a cosmetic chemist. For a few reasons
- There isn’t enough data for them to collect.
- Systems are too complicated for a tool like this to be really accurate.
- There is no simple way to determine if something is stable or not.
- You still need a chemist/technician to actually make & test the formula.
I don’t see how the AI really adds much to what a chemist with a book of starting formulas can already do.
-
The software was Coptis
I think you are right if you say it’s not better than a cosmetic chemist with a booklet of basic formulations but in my opinion, it’s faster and cheaper than a Cosmetic Chemist.
Yes, in the end, you need someone to Produce and test the Material but some of the Thinking will be taken over and will no longer need manpower in a company and the person that is Performing this work will maybe no longer need to be a Chemist at all because they are just producing and testing what a system is telling them to do. Although yes at the moment it does maybe not seem as impressive but it just started that is the first version they are willing to put into the market if a few years go on i think it will be more and more a tool to be aware of .
I see the issues with data and where they are getting the data and how reliable they are but Companys like Evonik already started to make things like the Kalaidoskop to make development easier and i would think there are some companies willing to share and anonymize their formulations and know-how out of there copits database to get access to the tool.
And yes I know that copis is not always the best tool that way i did not include the name because in my opinion, it’s just the first one to provide something others will do in the end as well in some way 🙂 Agree to Disagree on that point.
ps sorry if i come on to strong 😉
-
-
Unknown Member
Deleted UserJuly 11, 2023 at 11:13 amI actually got better advice from chatgpt on certain things than I did from anyone here. Like the time I couldn’t figure out why a body wash was thickening like how I expected to. All it needed was to sit a bit longer and it thickened just fine. People here were telling me to use all kinds of weird stuff that I would have to go spend more money for.
The context of what you ask seems to play a part in if you get a “diy” answer or a “science” answer. We are all here to learn and not shame others about how they like to make stuff and work.-
Are you referring to this discussion? https://chemistscorner.com/cosmeticsciencetalk/discussion/natrosol-hec-not-working-well/
I’m not sure anyone suggested weird, expensive stuff. But of course whenever you have a thickening issue, “let it sit for awhile” is pretty standard advice. We assume here that you already tried that.
chemistscorner.com
Natrosol/HEC not working well - Chemists Corner
I am trying to thicken a body wash with HEC and it is not forming a gel. It goes from white powder to clear grainy…
-
Log in to reply.