Home > News content

Gates and Karannik invest in AI chip factory Artificial intelligence hardware still has the opportunity to set new standards

via:博客园     time:2019/6/5 8:02:02     readed:128


Tencent Technology News, artificial intelligence has become the world's "red fried chicken" new technology, artificial intelligence chip companies naturally become the investment or mergers and acquisitions market, such as the Israeli computer vision identification chip company Mobileye was invested by Intel Corporation The tens of billions of dollars in acquisitions have become a springboard for Intel to enter the field of autonomous vehicles.

According to the latest news from foreign media, recently, Microsoft co-founder Bill · Gates, Uber co-founder Travis · Karanic's 10100 fund and current Uber CEO Dara · Cosro Shahi jointly invested Luminous, a small startup that develops artificial intelligence chips.

This investment shows that heavyweights in the technology industry believe that there is still an opportunity to develop new standards in the hardware of artificial intelligence that can be incorporated into a variety of software applications in the future.

In recent years, as artificial intelligence has become more popular, some start-ups have been developing the next generation of artificial intelligence-related hardware. Intel purchased a company called Nervana in 2016. The company's early investors Ali & middot; Ali Partovi also invested in the Bright Company based in Palo Alto, California.

Pan Tuowei told the media by email: “I always prefer to bet on a talented young team rather than a big, established company. Although there are huge risks, it is the reason why it is worth investing now: if the market competition is over, then the investment is too late. ”

Bright Co., founder and CEO Marcus G., said in an interview with foreign media on Monday that the company's chips can run artificial intelligence models, and they hope to make huge profits from these chips.

According to foreign media reports, the company has only seven employees, their goal is to create a super-performance artificial intelligence chip, each chip can replace 3000 boards equipped with Google's TPU processor.

“We know this is feasible because we already have prototypes. & rdquo; Gomez said.

Before, Google entered the artificial intelligence chip market, which is a bold move in itself. Instead of relying on NVIDIA's graphics processors, the company takes the development process in its own hands. It is reported that these graphics processors are very popular among artificial intelligence researchers to train artificial intelligence models with large amounts of data.

Google's public cloud computing service uses its own TPU processor, but so far, the company has not begun selling these specialized chips to other companies.

Bright companies really want to sell their chips to different types of companies, and cloud computing providers like Google or Amazon can buy them. In addition, Gomez said that companies that manufacture drones, robots or self-driving cars can incorporate these artificial intelligence chips into their systems.

According to reports, Gomez used to work for Google Inc. and social network Tinder.

Gomez said that at Tinder, he would start training an artificial intelligence model on the cloud computing platform in the morning and see if the training process was completed at night. Gomez said that in the future people should be able to train an artificial intelligence model in a matter of minutes instead of spending hours or days.

Gomez said that in terms of the design of artificial intelligence chips, the company has a differentiated feature that uses silicon photonic chip technology, which distinguishes them from other artificial intelligence chip companies, such as Graphcore.

It is reported that silicon photons can achieve fast data transmission in a small range (such as a server), similar to optical cable can quickly transfer data over long distances. This approach draws on the results of the company's co-founder Mitchell · Michelle Nahmias's neuromorphic photonics research at Princeton University.

Currently, other startups, such as LightMatter and Lightenence, are using silicon photonics to improve the processing power of artificial intelligence chips.

It is reported that the investment organization of Google's parent company Alphabet invested in LightMatter. The company's goal is to make Bright's chips compatible with popular open source AI software frameworks such as TensorFlow and PyTorch.

Gomez said that it is still not clear when the company's artificial intelligence chips will be available.

He said: "We use the same technology as other competing companies, so our schedule is similar to other companies. & rdquo; (Tencent Technology Review / Cheng Hao)

China IT News APP

Download China IT News APP

Please rate this news

The average score will be displayed after you score.

Post comment

Do not see clearly? Click for a new code.

User comments