Adobe today announced it is developing its own chips for machine learning. The program, codenamed Asperger’s Syndrome, will be used by special education classrooms, or by teachers who want to train visually impaired people in a classroom. It can integrate many existing programs in a variety of systems, such as an MRI or Laptop. This will allow teachers to integrate the traditional text-based training with virtual reality or digital sound and video. The idea is that students who experience the training will gain a context in which to understand the information, and can transfer it in a live learning environment.
We know from the research on autism and language disorders that visual stimuli, combined with appropriate auditory inputs can have a profound impact on the learning process. Children with autism easily learn new tasks when shown pictures. A child with Asperger’s Syndrome has trouble understanding facial cues and sounds. They respond to spoken words but are hardwired to respond to pictures. Adobe’s new program will take these principles into account and will be able to interpose various images and sounds within the program.
Although Adobe has long been considered one of the most prestigious software companies in the computer industry, and its products are used by nearly everyone, the company has never developed its own chip-based computer program. Other companies like Apple and Motorola have taken similar approaches, however. Google is rumored to be working on a project using its Google Cardboard technology.
For now, though, these types of programs are still fairly simple. The software can recognize faces, animals, objects, and other items. For example, it could recognize a cat in a picture, or a cat sleeping on a bed. Even something as simple as recognizing a dog’s face in a picture could be helpful in teaching a young child how to recognize other animals.
What makes this interesting, though, is that these programs are designed to improve machine learning. Instead of a program relying solely on labels, or on hand-eye coordination, the new software program will rely more heavily on pictures of things. This is, in a way, similar to computerized vision. Most of the software used for vision systems is still in the hands of professionals, who work with people to teach them what they should see. With a machine learning program, though, it will be up to college students and even high school students to teach the software what they should see.
The new software will be especially helpful to companies, too. As students grow and mature, it makes less sense to purchase lots of different programs for different age groups. The programs should be designed so that the user doesn’t have to train for years on one type of program. Instead, the software should be designed to teach over a longer period of time.
Another part of the future of chip learning will come as programs become more machine-specific. Right now, there are some machines that are capable of learning from handwriting recognition. In the future, these may become commonplace. It would be great for Adobe to have control over the types of machines that are capable of learning. If an educational program needs to learn a certain type of handwriting in order to provide a good interface for a child, then the company could control the hardware that learning this.
Clearly, Adobe is moving toward creating its own chips for machine learning. It is also exploring other ways to create a computer-programmable chip. The company has already shown signs of having computer chips that can recognize human voices and this will only continue as it develops programmable chips for other uses.