Artificial intelligence technologies are blooming in various areas today. Despite the debates on the ethics and efficacy of some specific AI-powered tools, we find changes in how we work, think, and live under the influence of AI.
Meal planning is one of the many daily tasks that AI can help us with. Evidence, both scientific research[1] and real-life cases[2], supports the capability of AI-powered meal planners in generating healthy meal plans. They seem smart enough to consider users' dietary restrictions, personal preferences, and health conditions, and to tailor the meal plans in a pretty personalized way.
Why Should We Do Meal Planning?
Meal planning brings much more than just convenience. It offers multiple benefits and is essential in lifestyle interventions for chronic conditions like obesity[3] and diabetes.
Planning meals allows you to pre-determine portion sizes and also increases the variety of food you eat. The health benefits are beyond weight control[4]: improved balance in nutrition intake, reduced processed food consumption, and more mindful dietary choices.
Specific meal-planning techniques are also incorporated with medical nutrition therapy and help mitigate chronic conditions with dietary restrictions. A highly personalized meal plan excludes allergens, supports blood sugar and cholesterol management, and meets the client's nutritional needs.
What AI Can Do in Meal Planning?
To test the capability of AIs in generating personalized meal plans, I asked ChatGPT 3.5 to do me this favor:
"Create a 7-day meal plan for a 30-year-old male who wants to lose 5 pounds in a month, vegan, balanced in nutrition, and prefers Asian cuisines. Include iced coffee in every breakfast."
The results exceeded my expectations. The seven-day meal plan generated by ChatGPT covers a wide variety of plant-based foods and signature Asian cuisines like curry, bibimbap, ramen, spring rolls, and sushi bowls.
In terms of nutrition, the meal plan provides several good protein sources for vegans: tofu, legumes, and edamame. It also emphasizes diverse whole grain choices like whole grain bread, whole grain tortilla, brown rice, soba, and quinoa.
Of course, this AI-generated meal plan has significant limitations. Since ChatGPT can not cook so far, it does not consider preparation time and difficulties, especially for breakfast cuisine. However, it can be fixed by additional instructions like "make the breakfast easier to prepare."
I also tested ChatGPT's ability to calculate caloric intake. Although it is not able to provide a detailed portion guide for every single meal, it gives a pretty accurate estimation of daily Calories for me to achieve my weight loss goal.
As mentioned in the prompt for the meal plan, I aimed for a slow weight loss of 5 pounds in a month. Based on my own calculation, I should eat, on average, about 2,100 Calories per day.
And here's what I got from ChatGPT:
"Adding up the rough estimates for each day, the total daily calorie intake for this meal plan falls roughly within the range of 1,900 to 2,400 calories per day. Remember, these are rough estimates, and actual calorie counts may vary. It's essential to listen to your body's hunger and fullness cues and adjust portion sizes as needed to support your weight loss goals."
The range it provides is rough but correct. However, it failed to match the specific caloric information for each meal in the seven-day plan. As I pointed out, ChatGPT quickly admitted its mistake and recalculated the caloric estimates.
AI+Human: The Future of Health
The overall results of the AI meal planning test are satisfying. The seven-day meal plan is good but several steps away from a mature, ready-to-use plan. It needs to be adjusted based on reality and requires some additional prompting to reach the ideal result.
Compared to traditional meal planning services from a dietician or a nutrition coach, AI-powered meal planning is more convenient, more accessible, and costs less. On the other hand, since the machine is not yet flawless, you need to have basic nutrition knowledge and know how to communicate with AIs effectively. In other words, AIs are not yet powerful enough[5] to replace human nutrition professionals.
Will AIs continue to learn and develop and finally outcompete humans in health and nutrition? Probably yes. However, a better way to think about this is to see the potential of combining AI and human touch. Studies have shown the effectiveness of the AI-plus-human plans in weight loss[6].
And it's fair enough to believe that incorporating AI-powered tools will benefit both professionals and clients in other aspects of health and nutrition, such as meal planning.
