Food interactions in the future

Artificial intelligence and Machine learning are a powerful combination when used together. Machine learning generates algorithms that when in touch with large quantities of data can self-teach and develop. With that in mind, we should expect to see smart robots doing all types of jobs and improving themselves really soon.

2016 was a good year for Machine Learning and AI. Almost every CEO from technology companies stated that they are starting something on AI or even try to become a company for that. Although, only a few had significant growth in this field, such as: Amazon, Facebook, Baidu, Google, IBM, Microsoft, Tesla and NVIDIA.

Besides that, in 2016 we got the chance to see a lot of cool and new products being launched in the Machine Learning field, such as: autonomous vehicles and house assistants (Google Home, Amazon Echo). Yet, what should we expect for 2017?

Looking for some new technology in AI, I found a really interesting product that can suggests different recipes based on food photos.

Researchers from MIT computer science and artificial intelligence laboratory understand that analyzing this data could help us know more about people eating habits. Additionally, it can also help us learn new recipes. They got together with Qatar Computing Research Institute and came up with a program called Pic2Recipe that can analyze a food photo and distinguish which ingredients were used to cook it and even suggest similar recipes

How does it work?

The project team put together data from websites like allrecipes.com and food.com to create a database with over 1 million recipes. This database has enough information about the ingredients used to cook a huge range of dishes. By using this data they could train the program to make connections between food images and the ingredients used in each one. By doing that, the software was able to identify similar recipes with the exact same amount of ingredients used to cook a corresponding similar dish.

Pic2Recipe, just by analyzing a photo could tell the user the ingredients used to make that dish. For example, it could recognize flour, eggs and butter and suggest similar recipes just by going through the database and selecting similar images. This way people could track their nutritional habits and even know what type of ingredient were used to cook a specific dish in a restaurant.

The system is still being developed. But for now, it can distinguish desert options such as cookies and muffins, because that is a common theme in the database. Although, it is facing some problems when it has to identify uncertain food ingredients, like sushi or smoothies.

In the future, they believe the program will be able to identify and know food in the deepest way. Meaning that it could even recognize how the food was made and understand the difference between different variations of food, such as onions and mushrooms.

This idea could benefit people that have some type of allergy or food restriction when the nutritional information is not available. Additionally, people could take a photo of a dish that they had and liked at a restaurant and try to recreate it at home.

This product is an amazing example about how AI and Machine Learning could develop in the near future. Just by getting the right information from a large database, a program can create, identify and also be programmed to self-learn on a similar level as a human being.

Resources:

https://www.forbes.com/sites/moorinsights/2017/01/06/five-things-to-watch-in-ai-and-machine-learning-in-2017/#27acb279455a

http://www.dataversity.net/2017-machine-learning-trends/

http://www.neuroksoftware.com/machine_learning.htm

http://news.mit.edu/topic/machine-learning

http://news.mit.edu/2017/artificial-intelligence-suggests-recipes-based-on-food-photos-0720