This week in AI and Machine Learning: A debate over AI-generated art, Google’s efforts to measure smells, and more. 

Author’s Note

Pro football is back and AI is poised to play a bigger role than ever in the NFL’s 101st season. Last year, the league teamed up with Amazon Web Services (AWS) to host a competition that  challenged data scientists to create AI models capable of detecting head injuries based on game footage. Kippei Mastsuda’s winning solution led to the development of the new Digital Athlete, which uses an AI model informed by game footage and specialized sensors to create digital avatars of every NFL player in a digital environment. Whoever your favorite player or team, learn how the digital athlete and other innovations promise to change football for players, coaches, and fans in our latest blog.

A graphic showing how computer vision can help analyze NFL gameplay to promote player safety

Artificial Intelligence News

How Do You Measure a Smell? 

Can you measure smells? Google researchers have been trying to find out for years. They introduced a graph neural network model in 2019 to pair distinct molecules with smell labels like “minty” or “floral.” This week, Google unveiled the Principal Odor Map (POM), which creates a sensory map not unlike a color map and enables researchers to discover and predict new odors and odor-producing molecules. 

To test POM’s capabilities, Google bombarded it with a range of unfamiliar smells tested by a panel. Since panelists were likely to differ in how they characterized specific scents, it was important to see how POM’s assessment compared to the consensus. POM was closer to the consensus than the average panelist, evidencing an impressive ability to discern smells based on molecular composition. Researchers have also shown that POM can predict the odor perception of non-human species like mice that have been subject to neuroscientific studies. Learn more about how POM works and how Google’s AI researchers test and refine it. 

AI Helps Win Art Prize and Sparks Debate

Théâtre D’opéra Spatial, a painting by Jason Allen, took 80 hours to complete and won First Prize for its creator at last month’s Colorado State Fair. News of Allen’s victory in the digital arts competition attracted controversy, however, when the Pueblo native posted about the win online. 

Allen created his prize-winning piece with the help of Midjourney, an AI-powered program that uses text prompts to create detailed images. Midjourney turned Allen’s prompts into a selection of 900 images. Allen selected a favorite trio and made manual updates to all three in Photoshop before entering them in the competition. He has been clear that the Fair’s judges were aware of his process and several have confirmed.

Allen has also said that he hoped to spark a conversation by collaborating with AI and submitting the result. He has certainly succeeded, the debate over what constitutes creativity, whether artists should fear replacement, and how the Colorado State Fair might update its rules has gone on for weeks. 

Translating Brain Activity into Speech 

At the end of August, Meta shared research related to its efforts to turn non-invasive recordings of brain waves into speech. Past efforts to decode thoughts have relied on surgical intervention. Without the need for brain surgery, a new model from Meta has been shown to turn seconds-long snippets EEG and MEG recordings into speech with greater than 70% accuracy. 

Meta’s work not only suggests the potential for dramatic improvements to the lives of individuals who have lost the facility for traditional communication, but contributes to the scientific community’s ongoing efforts to better understand the human brain. Check out a summary of the research as well as a complete research paper detailing the model’s development. 

Join our Community

See you next week! Until then, keep the conversation going on Plainsight’s AI Slack Channel

About the Author & Plainsight

Bennett Glace is a B2B technology content writer and cinephile from Philadelphia. He helps Plainsight in its mission to make vision AI accessible to entire enterprise teams.

Plainsight’s vision AI platform streamlines and optimizes the full computer vision lifecycle. From project strategy, through model deployment, and ongoing monitoring, Plainsight helps customers successfully create and operationalize vision AI applications to solve highly diverse business challenges.

 View All Blogs