AI Becomes An Ethical Chip And Project Catapult Chips In

3rd October, 2016 by

Tech Giants Partner on AI Research; Nvidia Unveils Xavier ‘AI’ Chip; Venture Capitalist Apologies for Women-in-Tech Op-Ed; Microsoft Looks to FPGAs for Data Center Acceleration; IEEE Ratifies New, Faster Ethernet Standard.

Tech Giants Partner on AI Research

A collection of some of the largest technology companies in the world has announced the formation of the Partnership on AI, a non-profit designed to develop best practices for artificial intelligence technology research, adoption, and exploitation.

The partnership is comprised of representatives from Amazon, Facebook, Google, and its DeepMind subsidiary, IBM, and Microsoft, the latter of which also announced the merging of multiple divisions to form an AI-focused business group containing over 5,000 computer scientists and engineers.

“AI technologies hold tremendous potential to improve many aspects of life, ranging from healthcare, education, and manufacturing to home automation and transportation,” the Partnership on AI explained in its launch announcement. “Through rigorous research, the development of best practices, and open and transparent dialogue, the founding members of the Partnership on AI hope to maximize this potential and ensure it benefits as many people as possible.”

Nvidia Unveils Xavier ‘AI’ Chip

Nvidia has announced the development of a system-on-chip (SoC) processor, codenamed Xavier, with claims that it is ideal as a host for the kind of artificial intelligence required for self-driving vehicles.

Described by Nvidia chief executive Jen-Hsun Huang as “the greatest SoC endeavour I have ever known,” the Xavier chip includes eight ARM-based central processing cores, a 512-core Nvidia Volta graphics processor, a visual processor capable of running at 8K resolution and a “Computer Vision Accelerator” module specifically designed to boost the performance of pattern recognition, motion tracking, depth sensing and other computer vision algorithms required for autonomous vehicle use.

Xavier is Nvidia’s latest automotive computer chip, but not its first since the formation of the company’s Michigan-based automotive technology center in 2013, the company has been working to take its low-power high-performance Tegra processor technology into the market. Xavier, it claims, can replace its last-generation Drive PX 2 system at a considerably lower power draw, and will be sampling to customers in late 2017.

Venture Capitalist Apologies for Women-in-Tech Op-Ed

Venture Capitalist John Greathouse has issued an apology for an opinion editorial in the Wall Street Journal in which he advised women working in technology industries to hide their identities and attempt to pass off as male – or, at least, remain genderless.

“I apologize for the dreadful article I wrote in the WSJ. I told women to endure the gender bias problem rather than acting to fix the problem,” Greathouse’s apology, posted via social networking site Twitter and linked from the still-live original article, read. “I hurt women and I utterly failed to help, which I wholly regret and I apologize for having done. Women have a tough enough time having their voices heard and my insensitive comments only made matters worse. I am truly sorry.”

Greathouse’s ill-thought-out advice included the use of initials rather than full names and the removal of any photographs from pitch decks and portfolio sites that would reveal a woman’s gender.

Microsoft Looks to FPGAs for Data Center Acceleration

Microsoft has released details of Project Catapult, a practical research project started in 2011 which has developed custom hardware based on field-programmable gate arrays (FPGAs) to accelerate data center performance.

Based on Altera Stratix V FPGAs, which can be reconfigured to act as custom chips, each Catapult accelerator is claimed to offer an order-of-magnitude performance boost over CPU-based solutions at a third of the cost and a tenth of the power draw.

Following the foundation of Project Catapult in 2011, Microsoft trialed its first FPGA-based accelerators in 2012, began ramp-up in 2015 and as of this year has implemented what it calls a “Configurable Cloud” architecture based around the Catapult accelerator boards in “nearly every new production server.”

“Today’s Project Catapult combines an FPGA integrated into nearly every new Microsoft data center server with a unique distributed architecture,” the company’s write-up explains. “The distributed architecture deploys FPGAs as an addition to each data center server, rather than a bolt-on isolated cluster, to create an ‘acceleration fabric’ throughout the data center.”

Microsoft has not announced any plans to offer the Catapult boards commercially.

IEEE Ratifies New, Faster Ethernet Standard

The Institute of Electrical and Electronic Engineers (IEEE) has formally ratified the 802.3bz Ethernet standard amendment, outlining methods to boost current 1Gb/s copper networks to 2.5Gb/s or 5Gb/s.

Existing copper Ethernet networks are capable of supporting 1Gb/s data throughput over full-length runs, increasing to 10Gb/s over short runs. While higher-speed standards exist, these require the replacement of existing copper wiring at considerable expense and logistical difficulty. 802.3bz, by contrast, promises to boost the network performance without replacing existing Cat5e and higher cabling.

“IEEE 802.3bz is a valuable addition to Ethernet’s expanding family of standards,” claimed John D’Ambrosia, chairman of the Ethernet Alliance and senior principal engineer at Huawei. “The Ethernet Alliance is excited by the promise and opportunities presented by IEEE 802.3bz. We are committed to validating the industry’s expectation of the multi-vendor interoperability of this standard as the follow-on to 1000BASE-T, considered the most successful Ethernet project ever.”

Thus far, no member companies have announced a roadmap for commercialization of 802.3bz.

(Visited 4 times, 1 visits today)