Artificial Intelligence (AI)
Statement on Artificial Intelligence (AI)
With the widespread adoption of platforms such as ChatGPT, Grok, Suno, and many others, a concerning trend has begun to emerge. While friends and colleagues are finding these tools helpful for everyday, largely harmless tasks—everything from drafting reports and creating PowerPoint presentations in seconds to generating fully produced musical pieces from simple poems or lyrics—we am now seeing a more troubling development: the emergence of the “AI inventor.”
​
The AI inventor approaches Mix Solutions seeking formulation work, explaining that they have a promising formulation they would like to bring to market. However, upon further discussion, it often becomes clear that this “great formula” is the result of a wish list of desired properties entered into an AI platform. In response, the AI provides a complete formulation, a preparation procedure, a safety data sheet, a patent application, product specifications, “performance testing results” (based on presumed effectiveness), and more. On the surface, this appears impressive. Unfortunately, the underlying information is frequently flawed.
AI systems do not always understand the real-world properties of ingredients or how those ingredients interact with one another. For example, in a recent case, an AI-generated personal care formulation included three functional ingredients intended for a water-based system. However, all three ingredients were almost completely insoluble in water. While two could potentially be made soluble under specific conditions, one required acidic pH and another required alkaline pH. Neither the AI nor the customer recognized that the formulation, as recommended, was fundamentally incompatible. Despite this, the AI confidently reported numerous advantages and claimed high effectiveness. In short, the information was inaccurate.
These “AI inventors” often subscribe to multiple AI platforms and compare results, using one system (for example, Grok) to critique or expand upon the output of another (such as ChatGPT). After spending hours refining prompts and adjusting recommendations, the enthusiastic customer arrives at Mix Solutions with what they believe to be a “complete” formulation, requesting a sample for evaluation. What appears to be a straightforward request frequently turns into a significant formulation effort, requiring substantial corrections to address deficiencies in the AI-generated data—assuming the formulation can be salvaged at all.
Going forward, all Mix Solutions proposals will include the following disclaimer:
Mix Solutions will not be held responsible for the success of AI-generated formulations. We make no guarantee that such formulas will be safe, stable, effective, or possess the performance characteristics reported by AI platforms. All AI-generated formulations will be treated as experimental concepts requiring further development and validation.
