[LUM#22] No AI Without Humans
Computers capable of learning are no longer the stuff of science fiction—they’re all about machine learning. And to learn, machines need humans to provide them with colossal amounts of data. To ensure the quality of that data, statistician Joseph Salmon is banking on cooperation.
How do self-driving cars know how to read road signs? How can a smartphone recognize a nightingale’s song or an oak leaf? If all of this is now possible—and even commonplace—it’s thanks to machine learning. “It’s the very foundation of artificial intelligence, ” explains Joseph Salmon, a statistics researcher at the Alexander Grothendieck Institute in Montpellier1.
To be able to tell the difference between a nightingale and a chickadee, or an oak tree and a poplar, the computer had to “look at” a lot of tree leaves or “listen to” a lot of bird songs. These images and sounds are the data—the lifeblood of AI. “Machine learning requires large amounts of data to function properly, especially for classification tasks.”
And the quality of this information—known as input data—is crucial to the success of the tasks at hand. “Garbage in, garbage out, ” the expert sums up. Garbage in, garbage out. In other words, feeding the computer flawed or nonsensical data will result in equally nonsensical responses from it. To illustrate this computer science adage, Joseph Salmon cites a well-known example: Barack Obama was reportedly identified by an AI as a monkey. The reason: the database used by the recognition algorithms consists mainly of faces of white people.
No AI Without Cooperation
So, to ensure he has a large amount of high-quality data, Joseph Salmon is relying on collaboration. In 2019, he was awarded a research and teaching chair in artificial intelligence funded by the ANR. Through Camelot—as the project is called—the researcher and his colleagues aim, in particular, to tackle the challenges of identifying biodiversity through citizen science and crowdsourcing—in other words, by engaging the general public to gather data.
This strategy has already been used to develop the Pl@ntNet plant identification app, a citizen science project designed to automatically identify plants using photos, in which Joseph Salmon participated. “Pl@ntNet has 25 million users who provide photo data and submit plant names. Crowdsourcing allows us to synthesize and utilize all this knowledge,” emphasizes the statistician, for whom “there is no AI without cooperation.” In fact, for the researcher, artificial intelligence is ultimately nothing artificial at all. “We take all the energy of human beings that we’ve synchronized; AI simply serves as the link in a collective effort.”
Neural network
In fact, learning how to learn is nothing other than what the human brain does. To endow machines with this ability, mathematicians and statisticians working on these issues draw inspiration from how the brain functions in order to mimic its learning processes. “The mathematical tool used is called a neural network. We can combine dozens or even hundreds of layers of neurons, each receiving and interpreting information from the previous layer. This is known as deep learning,” explains Joseph Salmon. Here, mathematics enables the creation of an algorithm designed to minimize error. “The neural network is a mathematical function that takes an image as input and must produce a name as output, which requires programming enormous functions, ” explains Joseph Salmon.
Character Recognition
In the late 1990s, the first major application of deep learning was the automated recognition of bank checks. The principle is simple: the computer must automatically recognize the amount—written by hand in both numerals and words—on each check. The task itself is complex, however, because handwritten numbers are never exactly identical. “To enable a computer to recognize a handwritten digit, we had to provide it with a large set of annotated images. In this learning process, the computer must provide the correct answer while minimizing errors, ” explains the statistician.
During the learning process, the algorithm aims to reduce the gap between the results obtained and the expected results in order to refine its recognition. This process is similar to how the brain works: it is capable of analyzing a multitude of pieces of information that are imprecise on their own and interpreting their combination to ultimately and reliably recognize a 2 as a 2—or an oak leaf as an oak leaf.
Check out theUM podcasts, now available on your favorite platform (Spotify, Deezer, Apple Podcasts, Amazon Music, etc.).
- Imag (UM, CNRS, Inria)
↩︎