Dive Brief:
- Appropriately working out of what was once a Nabisco factory, a Google research team paired up with a Pittsburg bakery to use AI to create "the best possible chocolate chips cookies," or "smart cookies," according to a company post.
- The team was tasked with designing a cookie given a set list of ingredients. They used the "Bayesian Optimization" technique. An AI system was given ingredient "knobs" that it could dictate portions to. The resulting cookie was fed to and rated by other Googlers, and the ratings were fed back into the system so it could "learn" from them. The process was repeated dozens of times.
- Eventually, the recipe behind Google's smart cookies combined with recipes from a local gluten free bakery. The chefs were guided by unusual ingredient combinations provided by the AI trials. After about two months and 59 batches, the team finalized the chocolate chip and cardamom cookie recipe.
Dive Insight:
Google's "AI-generated cookie" recipe was served up just in time for the holidays. It goes to show the limitless capabilities AI offers. The experiment, learn and repeat process in Google's AI cookie endeavor is the basis on which AI is built.
The intelligence behind AI is initially provided by the programmers. While this sometimes elicits unintentional biases, AI is trained to make decisions for its human counterparts.
Automation through AI stands to eliminate 1.8 million jobs but also to produce 2.3 million jobs by 2021. The inevitably of the technology is putting the direction of its implementation in the hands of CIOs. AI solutions are expected to be as popular as the cloud in the near future.
But less than one-third of companies think they have enough AI talent to expand on the technology in-house. Perhaps if potential AI experts knew eating cookies was part of the job description, the candidate pool would be a bit larger.
