Medical devices that monitor and respond to changes in our health. Robotic assistants that know what we want before we do. Kitchens that help us with our shopping and plan our meals.
Every day, we hear about how artificial intelligence is going to change the world. Amid all this focus on the future, it’s easy to ignore an unavoidable truth: AI is already changing the world in significant ways.
From predicting weather and detecting cancer to taking our first steps toward autonomous transportation, AI is powering applications that are having an impact today.
To spotlight the many inspiring stories of AI’s influence, NVIDIA is launching “I Am AI,” an all-new, original docuseries. The video vignettes are a window into the work of innovators achieving groundbreaking accomplishments with AI technologies. By celebrating those achievements, we hope to provide inspiration for others who are developing AI applications.
The first “I am AI” episode introduces Aiva, a French AI startup that’s developed an algorithm that composes original music of various styles. Future episodes will air every two weeks on the NVIDIA YouTube channel and NVIDIA website.
Subsequent episodes will feature PACCAR, which is developing autonomous commercial vehicles; Sigma Technologies, which is leveraging deep learning and AI to identify lung cancer nodules earlier; and Roborace, the world’s first motorsport championship for driverless cars. (Click here to receive alerts each time a new episode is posted.)
Aiva is an ideal subject for our premiere episode, in part because the company’s technology was front and center at this year’s GPU Tech Conference in San Jose. The score written by its music-composition algorithm served as the backdrop for the original “I Am AI” video that preceded NVIDIA CEO Jensen Huang’s keynote.
Aiva’s primary innovation is an algorithm that can create music very fast. And not just music, but music packed with emotion that’s on par with work written by top composers.
The company’s algorithm starts with a huge database of classical music stored in midi format. Using a recurrent neural network trained on NVIDIA GPUs, AIVA looks for patterns in selected tracks to understand the basic style. It practices and improves itself by predicting what will come next, then it uses those predictions as the basis for a set of rules for that style of music.
So what’s to stop AIVA from stealing from the masters it studies? Well, the company’s leaders thought of that.
“Using GPU computing, we’ve created what we call a plagiarism checker,” said Aiva CEO Pierre Barreau. “It’s able to understand whether a created track is partially or fully plagiarized from the database.”
Pushing the Limits
For those who doubt whether AIVA should be considered an actual composer, consider this: Not only has the software worked in French television and cinema, it’s also been recognized by SACEM, a French professional association that copyrights artistic works.
As powerful as Aiva’s algorithm is, Barreau said the company has its sights set higher, with three features planned to boost its developing capabilities. He wants AIVA to be able to assess whether a piece of music was good or missed the mark, compose for a full orchestra directly, and convert the emotions in a script into a matching score.
Eric Breton, a well-regarded French composer and conductor, said he believes the ability for AIVA to determine whether a piece of music is effective could be game-changing. Composers, he said, are too entwined in their compositions to make such judgments.
“To judge music implies that you can be outside and to look from outside,” Breton said. “When I’m conducting, I must be in the heart of the music. It’s a big chance for musicians to use technology — no, to ask it — to push us toward excellence.”
Watch more episodes of the “I am AI” docuseries on our AI Innovators page.