Artificial Intelligence Advances: Scientists Teach Machines
Source: Sumayah Aamir
The experts have brought artificial intelligence abilities closer to those of living breathing humans.
A bunch of researchers have built an algorithm that truly epitomizes our learning repertoire as human beings. This algorithm will allow computers to detect and draw simple visual stuff. Mostly, this visual stuff is capable of being detected and drawn by human beings only.
The results of the study were published in in the latest issue of the journal Science. It is a great advance in the annals of the field. The time span taken by computers to learn novel concepts will be considerably shortened and they will be put to good use in completing creative tasks.
"Our results show that by reverse engineering how people think about a problem, we can develop better algorithms," explains Brenden Lake, a Moore-Sloan Data Science Fellow at New York University and the paper's lead author. "Moreover, this work points to promising methods to narrow the gap for other machine learning tasks."
Other authors of the study were Ruslan Salakhutdinov, an assistant professor of Computer Science at the University of Toronto, and Joshua Tenenbaum, a professor at MIT in the Department of Brain and Cognitive Sciences and the Center for Brains, Minds and Machines.
This algorithm has been produced via a process of reverse engineering. It will go a long way in bridging the gap between machine learning and human learning. When human beings are introduced to a novel concept, they take in the data and make sense of it rather quickly.
Whether it is a dance move, a cooking maneuver or a new language, they easily pick up on the basics. And although machines are able to copy some of these moves (ATM machines are a prime example) they take a lot of input to be able to perform the tasks at hand.
Building such machines that rely on the same amount of data as humans has proven to be a difficult task. The replication of human abilities in machines is a leading edge field of endeavor. It involves statistics, computer vision and cognition.
The whole shebang has to do with what may be called deep neural networks. Researchers introduced a Bayesian Program Learning (BPL) framework. In this the letters of the alphabet are represented by computer code. No programmer is required too.
Standard pattern recognition involves pixels or features. The BPL way though involves generative models and model building. There is a learning of learning in this particular setup. Speed learning is a very real possibility too. The algorithm allows learning in new contexts.
Human beings were compared to the machines in a test of alphabet recognition. The results showed up on this visual Turing test. Kids in kindergarten could recognize concepts taught to them and build new concepts from these taught examples all by themselves.
This is how their audio-visual abilities and language recognition powers progress with the times. Yet even today the most complex of machines cannot match a three year old child’s vision and concept recognition capacities.
"Before they get to kindergarten, children learn to recognize new concepts from just a single example, and can even imagine new examples they haven't seen," notes Tenenbaum.
"I've wanted to build models of these remarkable abilities since my own doctoral work in the late nineties. We are still far from building machines as smart as a human child, but this is the first time we have had a machine able to learn and use a large class of real-world concepts -- even simple visual concepts such as handwritten characters -- in ways that are hard to tell apart from humans."
| }
|