What is Luna AI, and how can it help me?
Luna AI is your intelligent partner in formulation development. Designed to accelerate R&D processes in industries like food, beverages, and personal care, it simplifies creating and optimizing formulations. By reducing manual trial-and-error and saving time and resources, Luna AI helps you innovate and achieve your goals faster.
How is Luna AI different from traditional formulation tools?
Traditional tools rely heavily on manual input and static methods. Luna AI revolutionizes the process by leveraging machine learning to predict outcomes, optimize ingredient combinations, and dynamically adapt to your constraints. Think of it as an AI-powered lab assistant that evolves with your needs, providing smarter, faster, and more insightful recommendations.
What is a project in Luna AI?
A project in Luna AI represents the goals and constraints you aim to achieve during formulation. It’s tailored to specific needs, whether you’re improving an existing product, benchmarking against a competitor, or working from a marketing brief. Within a project, you can define key parameters, upload data, and explore AI-driven recommendations, all in one streamlined workspace designed to help you reach your formulation objectives efficiently.
Do I need a lot of past formulation data to get started with Luna AI?
Not at all! While having past formulation data can enhance Luna AI’s ability to generate insights and recommendations, it’s not a requirement to get started. You can begin by defining your goals, constraints, and inputs, and Luna AI will use its intelligent algorithms to provide recommendations based on the information you provide. As you upload more data over time, Luna AI adapts and becomes even more accurate, helping you optimize your formulations faster.
What is the objective of the thumbs up and down functionality?
The primary goal of user feedback is to fine-tune the optimization process within the system. Users are asked to provide feedback on two main areas:
Outcome Levels: Feedback on outcomes helps in adjusting the objective function by influencing the priority of certain outcomes. For example, outcomes that receive more thumbs-up will have higher priority in the optimization process. This feedback ensures the system better aligns with the user’s preferences, though it’s more of a subtle tuning rather than a major change.
Ingredients and Inputs: Feedback on ingredients, which includes all input parameters, is based on the user’s expert knowledge. This feedback is subjective, allowing users to guide the search process by indicating whether certain formulations are likely to produce good or bad results. While this does not override the data, it helps steer the optimization process more intelligently.
The system uses this feedback to guide the search for optimal formulations, balancing between the user’s expert input and the system’s data-driven insights. The goal is to enhance the efficiency of the optimization process by combining human expertise with machine learning.
What happens when the user provides feedback by giving thumbs up or down to a set of formulations?
When a user provides feedback (e.g., thumbs up or thumbs down), the system subtly adjusts its optimization process. Specifically, the feedback fine-tunes the priority of certain outcomes and guides the search process for better formulations. However, the changes may not be immediately visible to the user, as the system continues to search based on a combination of user input and existing data.
In the short term, users might not see noticeable changes in the outcomes or suggestions because the system is balancing their feedback with other factors. Over time, it is incorporated to accelerate the search process, but the immediate effects might not be apparent.
What happens when I provide feedback on both ingredients and outcomes? Could these conflict?
When a user provides feedback on both outcomes and ingredients, the system treats these inputs separately. Feedback on outcomes helps the system better understand which results are most important to the user and prioritize those in its search. Feedback on ingredients helps guide the system to explore areas that the user thinks are more likely to produce good formulations.
You could find a scenario where the ingredient values on a formulation look good but the outcome ranges are not what you would expect, so you provide different feedback on the same formulation. There’d be no conflict between these two types of feedback; the system would use both to refine its search and deliver better suggestions that align with the user’s preferences.
Why doesn’t cost have an option to provide feedback?
Currently, cost is not directly included in the feedback process or the system’s prioritization. While users can see cost information, it’s not factored into the thumbs-up or thumbs-down feedback on ingredients or outcomes.
However, there is a recognition that cost is an important factor for you when deciding on a formulation. Cost could be integrated into the feedback process in the future, allowing users to prioritize cheaper options if they choose. This would require adjustments to how cost is factored into the system’s decision-making process. For now, cost remains visible to the user but is not actively influencing the system’s optimization based on feedback.
What happens if I change my scenario settings once I have provided feedback on a different scenario?
If you change your goals or constraints after providing initial feedback, the system will automatically regenerate suggestions or options based on the new settings. The feedback given before the changes remains useful, as it helps guide the system within the new constraints.
The system ensures that any previous feedback is still relevant unless it directly violates the new constraints. You do not need to restart their entire assessment process when making adjustments to goals or constraints; the system will continue to use the updated information to refine its search and provide better results. This approach allows the system to adapt to changes while still benefiting from the feedback already provided.
How many thumbs up or thumbs down do I need to provide to effectively guide the system? Should I assess all the cards?
The number of thumbs-up or thumbs-down feedback guides the system with its search for the best formulations. More feedback generally helps the system fine-tune its recommendations and align more closely with your preferences.
While you can skip the feedback exercise, it is highly encouraged. Doing so can help narrow the search space and potentially reduce the number of rounds required to achieve the desired goal. We recommend giving at least 8 thumbs up or down.
When providing feedback do I need to compare with other formulations or consider each formulation individually?
Currently, the system treats each piece of feedback—thumbs up or thumbs down—individually. You do not need to compare options against each other when providing feedback. Each ingredient or outcome is evaluated on its own merits, and the system uses this individual feedback to adjust its search process accordingly.
While comparing options might seem helpful, the system is designed to aggregate individual feedback and use it to refine its overall approach. Therefore, you can focus on evaluating each option on its own without worrying about direct comparisons to others.
What happens if I change my mind once I have provided feedback on a formulation?
The system will adjust accordingly. For example, if you change your thumbs-up to a thumbs-down, or vice versa, the system will update its records to reflect the most recent input. If you simply deselect your feedback without providing a new one, the previous input will be treated as neutral. This ensures that the system’s optimization process is always based on the most current feedback, allowing it to stay aligned with your evolving preferences.
What happens when I upload test results into the system? Do I have to go through the same feedback process again?
When you upload new data, they do not need to repeat the entire feedback process from the beginning. The system already takes the new data into account, integrating it with the feedback that has been previously provided. While the system may adjust its suggestions based on the new data, the previous feedback remains relevant and continues to guide the optimization process.
Is there a way to track the impact of thumbs-up or thumbs-down feedback?
The system can collect and analyze the thumbs-up or thumbs-down feedback provided by users, but tracking the exact impact of that feedback on the optimization process requires additional evaluation. To understand the full effect of the feedback, it would be necessary to compare the results with and without the feedback, which is something the system is capable of but has not yet been fully implemented.
While the feedback helps guide the search and can potentially speed up the optimization process, the specific impact—such as how much faster or more accurate the search becomes—depends on further experimentation and evaluation. This is an area where the system could evolve, offering more insights into how user feedback influences the outcomes over time.
How can I generate recommendations for my project?
To generate recommendations for your project, simply click the “Generate Recommendations” button. This powerful feature uses your current scenario—including goals and constraints—to create optimized formulations tailored to your needs. Here’s how it works:
Click “Generate Recommendations”: The system activates a machine learning algorithm powered by Bayesian Optimization.
Top Formulations: The algorithm evaluates thousands of possible formulations and ranks them by their Desirability Score, providing the best matches for your scenario.
View Results: The generated formulations are automatically saved in the database and can be found under the “Suggested Formulations” filter on the Execute Solutions page.
What is an initiative in Luna AI?
An initiative in Luna AI represents a series of rounds based on the same set of goals and constraints within a project. It allows you to explore different recommendations and formulations while maintaining consistency in your settings. Each initiative tracks progress and results over multiple iterations, making it easier to evaluate outcomes and refine your approach without starting from scratch. It’s a structured way to test and optimize within a specific formulation framework.
Do I need to test all formulations in an initiative?
While it’s not mandatory to test every formulation in an initiative, testing the entire batch as recommended by Luna AI provides the greatest chance of reaching your goal faster. The system is designed to learn from all the formulations it recommends, helping it identify the most effective paths and refine its suggestions for the next round. Selecting only a few from a larger batch can limit the information Luna AI gathers, potentially reducing its ability to optimize outcomes. Testing as recommended ensures you get the most accurate insights and the fastest route to success.