Wolfram’s Image Recognition is a tremendous change in artificial intelligence, which is happening right now, but, in the past people never believed.
The website is a fun time, as does Wolfram’s Research. A British computer scientist, physicist, entrepreneur, and all-around free thinker, the namesake software firm.
It’s going to do things properly as frequently as it doesn’t.
Wolfram’s Image Recognition project
Besides, Wolfram’s normally lengthy article about this project, the future of artificial intelligence will make you think.
In this case, though, the Wolfram example is a huge change in AI, which is happening right now.
His technique is based on the “convolutionary neural networks.” The huge computer networks try to replicate neuronal networks within the human brain.
Neural grid an old idea
The neural grid is an ancient notion that dates back six decades, Wolfram points out. But, years later, this notion has led Photo Recognition on Facebook, to Google Voice Recognition through Skype language translation, as many claims, will never work.
Now, that idea has changed!
David Luan, the creator of a neural networking firm called Dextros, says: “More and more firms are taking this type of work extremely seriously.
The new Wolfram website indicates that software developers outside the major internet firms are easily accessible for such an AI — to some extent at least.
Image recognition with AI
The site is a show of the latest edition of the Wolfram Language. Thus, the general-purpose programming language offered by Wolfram and company. Using the language, Wolfram says, any developer can build image recognition into their own app. Further, tapping into a large cluster of machines operated by the company.
Similar work is being done by other firms. An outfit named Metamind provides tools to drive neural networks for your own apps. Dextro offers tools for identifying pictures in films from neural networks. And since many methods for deep learning are available as open-source software, even independent developers may run their own neural networks.
As Wolfram’s example shows, these techniques continue to evolve. It is now obvious, however, that neural networks are quite successful in performing better in humans in specific conditions. Language speech and translation and more may get identified and recognized in a consistent manner. This also demonstrates the demo of Wolfram.
Wolfram states that this is especially noteworthy. Because for so many years the neutral net notion was believed to be dead. “When somebody tried to do something long ago, I don’t know of any other technology and it ultimately succeeded,” he says.
Anyone can do it!
The amount of computing power that we have available has changed. Now we have tens, hundreds, and even thousands of high-performance CPUs to operate these systems.
So, just like Facebook and Google, Wolfram and the business developed their picture recognition model. Thus, on a cluster of workstations they are equipped with graphics processing units or GPUs. “It’s not a big breakthrough that finally worked,” he says. “It’s because we can now create huge enough systems.”
Even tiny systems are large enough in some instances nowadays. So, “Any clever child with a GPU-fitted PC can do so in the basement of his parents with open source tools,” says Yann LeCun, Facebook’s new artificial intelligence laboratory leader.
