Market Data 
For some of us, cooking is hard, and coming up with nutritious meals with what’s already in the fridge is harder.
I have always struggled with meal prep, often forgoing a meal altogether when I just don’t have the mental energy to decide what to make. So, despite having a healthy dose of skepticism toward AI, I decided to let Meta’s new Muse Spark AI judge the nutritional value of my lunch and help me decide what to make for dinner based on the few items I have in my fridge.
Muse Spark was freshly launched on April 8 after Meta waged a talent war to fuel its AI ambitions. In Meta’s press release, Muse Spark was partly promoted as a personal AI tool that could help you track your health or plan a trip.
“Health is one of the top reasons people turn to AI, so we worked with a team of physicians to develop the model’s ability to provide helpful information on common health questions and concerns,” according to the press release.
With those marketed functions in mind, instead of asking the AI tool to write me a paragraph, I skipped straight to lunch.
Rating my salmon lunch
I got my go-to take-out from my favorite Japanese bento place for lunch near my home. It consists of seared salmon on rice, with an egg, mixed greens, and a side of raw salmon and fish roe.
I uploaded a photo of the meal to Muse Spark and tasked it with creating a detailed breakdown of every type of food and sauce, along with the number of calories for each item. I also asked it to make a labeled image of my meal, give it a score out of 10, and explain how it factors into my nutritional needs of the day based on my biometrics.
Muse Spark was mostly accurate about what ingredients were in my meal.
It made it clear it didn’t know the weight of the ingredients or the exact type of oil it was cooked in, but it estimated that my salmon bento was around a total of 760 calories. Muse Spark also estimated, based on the photo, that the dressing and sauce I love are extremely high in calories and sodium, leaving me little room for additional sodium intake in my day.
Muse Spark also said that the meal appeared dense in micronutrients like Omega-3, based on the photo. However, the AI thought it was lacking in fiber, vitamin C, and calcium. Overall, it rated my meal a 7.5 out of 10 and reminded me that it is not a licensed nutritionist.
Where Muse Spark struggled was generating a labeled image.
The labeled words made no sense. After one more failed attempt to ask it for a labeled image with readable text, I gave up and moved on to the next task.
The dinner challenge
Now that Muse Spark knew what I had for lunch, I collected leftover and eclectic ingredients in my fridge for a visible photo and asked it if it could suggest some dinner recipes for me.
I explained that I have a wide array of condiments available that aren’t shown in the photo, and I would prefer recipes that are easy to cook and clean. Here are Muse Spark’s suggestions that came with a brief step-by-step recipe:
- Tomato-braised chicken + roasted potatoes + papaya side
- Light spaghetti alle vongole + blueberry-oat smoothie
- Japanese-style oyakodon + papaya-blueberry salad
According to Muse Spark, the meals are designed to make up for its analysis of my lunch, which it said lacked fiber and vitamin C, with the option to moderately increase carb intake. It also suggested that I rinse my canned tomatoes to lower their sodium level and skip any salt and soy sauce.
“Sodium is the main thing to watch — if the rest of your day is also soy-sauce heavy, you’ll likely overshoot 2300mg,” said Muse Spark.
Muse Spark was not able to tell that the package that says freeze-dried strawberries is actually covered in sweet yogurt and far too sugary for a smoothie, but its options gave me sufficient inspiration to cook and taught me clever ways to finish my leftover items, such as the papaya half.
At the end of the day, I decided on the Japanese-style oyakodon meal, and saved the other two recipes for another day.
Will I stop putting soy sauce in everything and skip my favorite condiment options in my bento? Probably not, even if I can hear Muse Spark telling me to stop in my head.
Read the full article here
