Now very few people will actually take a recipe into the kitchen, even if they really crave a dish, take out a mobile phone random search can find a lot of help. So, when Forbes reported that Facebook's artificial intelligence research group has developed an "AI recipe," many people were grin and grin and forget it.

Although Facebook has no plans to market the results of this research, the application value of this research may be really interesting to foreigners. Different from the mysterious "appropriate amount" of Chinese cooking, foreigners have food scales in their kitchens, and how much sugar and cinnamon powder are used, or even on the scale to enter the dish, and the use of AI to give accurate analysis of food ingredients may be a good selling point.

AI recipe research hot

Artificial intelligence is really a versatile technology, although it is still in the era of weak artificial intelligence at this stage, but it has laid more and more eggs along the way to completely change our lives.

In his lab in Montreal, Andreas Romero and his colleagues trained an AI with lots of food images and recipes. Given the images, it could list the ingredients and generate a recipe.

Before Facebook, MIT's Computer Science and Artificial Intelligence Lab was doing something similar, building a database of more than a million different recipes collected from various websites and uploading information about the ingredients to the database. Using the database, the AI system, called Pic2Recipe, can look at a photo of a food, predict its ingredients, and come up with a recipe to make similar foods.

When tested, Pic2Recipe can give 65% of the correct recipe, the academic community is satisfied with this ratio, but if it is to be put into the market, artificial intelligence recognition ability is obviously limited, especially when encountering extensive and profound Chinese cuisine.

Currently, the team is ramping up Pic2Recipe's precision and continuing to enrich its skills, says lead research author Nick Hynes: "If you know what ingredients are in a recipe, but you don't know the quantity, you can take a picture, enter the ingredients, and run the model to find a similar recipe in a known quantity, and then use that information to output a recipe close to your own meal."

Recently, a team of MIT students posted a video of themselves baking pizzas, provided by their AI-trained chef Strono, which automatically cooks the pizzas through hundreds of "handmade pizza recipes" from online food blogs. Strono has already started working with an artisanal pizza restaurant in Boston, and the AI-created recipe is still not perfect, with some mistakes that can only be corrected by humans.

But the interesting thing is that, with the exception of artificial intelligence, no one else will be able to create a strange-looking but tasty pizza, and the Boston pizzeria is said to have considered incorporating Strono's innovation into its menu.

Obviously, artificial intelligence's "understanding" of recipes is still very different from that of humans, and they are still confused in the face of cooking.

AI and diet health are fermenting

In the fourth season of the popular American TV series Silicon Valley, incubator engineer Jane Yang developed a Not Hotdog app, which can tell you whether it is a hot dog or not by pointing your phone's camera at food or by looking at a photo.

The app looks a bit funny, but it points to sophisticated artificial intelligence technology.

Luo Haoyuan, partner of Sino-Guangdong Golden Bridge Investment, said: "A lot of careless or seemingly useless research or development is the necessary means and process for scientists to understand the world. Artificial intelligence recipes, such as using AI to recognize images and interpret images, may not be necessary for some people, but many interesting applications of computer vision technology are definitely on the right path."

"Food is often overlooked in computer vision because we don't have big enough databases to delve into it," says Antonio Torralba, a professor at MIT. "But those seemingly useless images on social networks actually give us a lot of information."

Professor Christophe Ratner, from the Department of New Media Technologies at the University of Vienna, said: "Imagine that people could use this system to find out the nutritional content of their daily intake, or take a picture of a meal they enjoy in a restaurant and then go home and make it themselves."

AI and diet health are constantly fermenting, and even a very experienced dietitian would struggle to master 100,000 food information, but AI can.

According to Wang Man, a private customized nutritionist, at present, China's nutritionists are using each meal with a quantity of recipes, the so-called quantity of recipes, refers to, such as breakfast: 25 grams of egg soup, 25 grams of coarse grain cake, 250 grams of milk, such nutritional meals seem healthy, difficult to implement, and the same.

Wang Man said: "AI catering can start from the personal health state of the diner, conform to personal tastes and habits, and carry out reasonable matching." AI food preparation will help Chinese people have a healthier diet and will also be a good helper for nutritionists."

The Strono team also writes on their project website: "In the process of AI awakening, robots are becoming better and better at doing many jobs that were originally human, and there may be concerns that AI will eventually replace humans and lead to mass unemployment, but we believe that when humans and machines work together to enhance each other's complementary strengths and skills, we can achieve the most creative and productive results."