The terrible stories of robots that will take away your work in the future have become even closer to reality. The list of taxi drivers, journalists and accountants will soon be joined by professional chefs. Artificial intelligence is already making its way into cooking.
Shazam for food
In June, Facebook employees told about their own algorithm, which is able to recognize the recipe from one of its photographs. According to experts, their development differs from the rest in that they recognize all the ingredients and how they were processed. No wonder, as scientists have "fed" the neural network millions of pictures of food.
All previous algorithms worked as follows: they tried to extract a recipe from a fixed set of data based on an assessment of the similarity of the image. However, the new method implies an intermediate step - forecasting the list of ingredients. This helps to provide additional insight into how the latter have been processed to produce the final dish.
"Our method begins with pre-training the image encoder and ingredient decoder, which predicts a set of ingredients using the visual features of the dish extracted from the input image, as well as the sharing of these ingredients. We then train the ingredient encoder and decoder commands that generate the name and instructions for cooking," the developers described how their neural network works.
Scientists have noted that their new system is only the first step to more serious algorithms, which in the future will be able to estimate the number of calories from the photo. At present, however, it is difficult to determine the exact recipe for some dishes. For example, AI still finds it difficult to understand that chocolate or jam may be inside a croissant. In addition, during the preparation of many products are severely deformed, making it difficult to identify them.
Ninja turtles won't stay hungry
One popular pizzeria (Domino's) has already adopted artificial intelligence technology. According to its advertising campaign, AI will check how good the pizza looks before it is sent to the client. However, so far only in Australia and New Zealand.One popular pizzeria (Domino's) has already adopted artificial intelligence technology. According to its advertising campaign, AI will check how good the pizza looks before it is sent to the client. However, so far only in Australia and New Zealand.
The new DOM Pizza Checker system, developed using Dragontail Systems technology, works as follows. In the workshop, a camera hanging from the ceiling is directed towards the cutting and packaging tables. Special software analyzes the camera image, recognizes the type of pizza, analyses the distribution of the filling and assesses the appearance of the food.
Why this algorithm is needed? According to the CEO of the company, one of the main complaints of customers is that in reality the pizza looks worse than on promotional photos. If the system is working properly, the residents of Australia and New Zealand will no longer have to worry about the problem.
Pinch of machine learning
Artificial intelligence can be used not only to improve old dishes, but also to create new flavors. This was the conclusion reached by IBM Research staff in cooperation with McCormick & Company, one of the world's largest spice companies.
According to experts, AI technologies are ideal for developing new spices. Algorithms can process many of the available data accumulated over decades of company operations. They analyze all information about raw materials, seasoning formulas and consumer test results.
Based on the above information, the software selects new combinations of ingredients for spices. In addition, the algorithms find possible alternative raw materials and change formulas if necessary. This significantly reduces the development time for new products.
The first products created with the help of machine learning technologies will be available in the second half of 2019.
Taste forecast
Gastrograph AI is a machine learning platform for predicting and simulating consumer taste preferences for products and beverages. In other words, this program can tell manufacturers in advance whether their new products will please customers or not.
According to the developers, the algorithm helps to optimize the taste, aroma and texture of the product for the target audience. The data can even be segmented by demographic group.
In order to learn about people's preferences, the developers analyzed feedback on food and beverages. Tasters from different countries entered their impressions after each meal, thus creating a database. Based on this, the AI predicts how consumers will perceive new products.
Interestingly, the technology has already been successfully used by American brewers. They used Gastrograph AI to keep their beverages tasteful over time. For this purpose, representatives of Yards Brewing began to plant tasters in front of the bartender, who introduced their impressions of the drink into a special application. After that, the brewers learned what beer should taste like at each stage.