"AIce” Cream

Dr. Bruce Tharp taught that much depends on factors not yet in evidence, i.e., “it depends.” Then, turn around with directive to “stay tuned.” Given the pace of development and utility of artificial intelligence (AI), when does “it depends” become “it’s certain?”

Ask an AI app, “What type (or kind or style) of ice cream should I make?”

The response: “Make what tastes good to you.” Pretty simplistic. So, for now, the answer remains, “it depends” with much more yet to consider (i.e., “stay tuned”). Then, more questions than answers.

Development and utility of AI itself is traveling faster than the speed of light with science, capabilities, value propositions, issues (for good/bad) changing daily. So, what is it?

AI is the development of computer systems (hardware and software) capable of performing complex cognitive tasks traditionally only a human could perform. AI should be able to respond/learn from results of such tasks, and change/manage its own “cognitive” functions versus classical repetitive execution of same tasks. There are “not-so-intelligent,” “intelligent” and “ultra/super-intelligent” AI systems all of which are still learning. Of which one(s), are we talking?

Steven Young, Ph.D., is principal, Steven Young Worldwide; Bill Sipple is principal, Wm Sipple Global Services.

Pros: Automatic repetitive tasks

• Quick analysis of large data sets
• Improve decision making
• Generate new material(s)/directions
• Reduce operational costs

Cons: Loss (change) of job functions?

• Bias, skewed and/or wrong assessments
• Loss of creativity (including wrong directionals, applying wrong/missing information, etc.)
• Varying ability to extrapolate accurately and precisely

How might AI influence the manufacturing of ice cream that includes a complex set of “what if” conditions and variables? Each of these influence the others in real time and space, where many such factors are qualitative in nature.

Can AI “fill-in-the-blanks” (learn, execute, direct reactions to same/new tasks)? AI could well make and apply assumptions (for good/bad) not yet known nor fully understood.

Does AI know ice cream is the only food designed, formulated, manufactured, marketed and sold with express intent of being consumed frozen? It probably does now.

Can Mother Nature’s rules and preferences be assumed, managed and changed? Think of climate and human influences on ingredients (sourcing, compositions, functionalities, influence on sensory appeal, economics, etc). Then, to ingredient selection, formulation considerations (what may or may not be made? what is to be declared about the finished ice cream?), variables at points of manufacturing, storage, distribution, sale and consumption.

Can AI define/redefine the undefinable? What about current/future decision making?

Complexities of making ice cream

Mother Nature. Rules of chemistry and physics do not change yet how we apply these does. Mother Nature does have preferences. If she had her way, ice cream would not exist. Can AI override all this as we do every day that we make ice cream?

Political/economic/supply chain dramas at home and abroad.

Formulation considerations as technologies advance and markets evolve.

Formulations where multiple options yield same objectives.

Dynamics of the market: what sells, where, what doesn’t, why.

Mix processing: hardware designs and operating conditions.

Hardening: multiple systems, designs, conditions of use (times, temps, etc.)

Differences of all above within a manufacturer and between manufacturers.

Sensory appeal (appearance, body, texture, flavor, elements of taste, acceptability.)

Quality/food safety.

Financial considerations: constant changes of supply versus demand.

Expect the following sooner versus later...

Flavor Development

• Flavor Profiling: Analyzing, in real time, customer preferences, sales data, and food trends to identify what to do (or not do) next.

Formula Development: Algorithms to access thousands of ingredient combinations to “optimize” taste, texture, and nutritional content without endless physical trials. Matching and managing, again, in real time, verses consumer expectations. AI to set finished product objectives, identify current and new opportunities, manage costs, declarables.

• Net: faster, more efficient, more accurate, more precise

Process Optimization

• Setting and Managing Processes: Sensors throughout the production process already monitor pasteurization temps, hold times, back pressures, viscosities etc. Process equipment “talk” to each other. It is not hard to envision “histories” from which AI can learn, predict and direct future action. AI can create parameters to understand and manage the overall “health” of operations ELF (end-to-end logistics.) AI can clearly, quickly identify deviations from norm and mitigate same.
• Process Optimization: Identifying deviations from norm; correcting same. AI allows for real time process optimization. Results: lower start-up/shut-down costs, reduced waste, improved efficiency, and, perhaps, increased production speeds.

Quality Management

• Define and Manage Quality and Process Efficiencies: Most ice cream manufacturers have continuous monitoring checks, e.g., weights and metal detection. Visual quality management tools exist. AI can identify material waste, set, identify and mitigate deviations from quality targets, saving time and improving efficiency. Same could be true relative to elements of food safety, i.e., identifying, measuring, and mitigating physical, chemical and microbiological hazards. Value-added elements to any HACCP program.
• Sources of Variation: Accurately identifying and mitigating variations due to manpower, methods, materials and machines, i.e., enhanced total quality management, TQM.
• Customer Feedback: Pre-emption of customer feedback (both good and bad), proactively identify and respond as needed.
• Sustainability: Identify and manage environmentally friendly approaches.
• Shelf-life: Quickly, accurately, precisely predict and manage shelf-life expectations. Already a large set of unknowns; can AI negotiate its way through?

What else?

Knowns vs Unknowns. Identifying what is known vs unknown; how, when, where, with what to bridge the gaps.

Barriers to utility to be identified, understood, mastered, perhaps, by AI itself!

Guidances as to what is correct, practical, prudent, appropriate and safe.

Might AI have edited this column? We didn’t dare ask! (Editor’s Note: No, it did not).

Stay tuned for the evolution of AI in ice cream manufacturing. DF