Intel Makes AI Understandable and Accessible at DevCon 2018

By Charles King, Pund-IT, Inc.  June 6, 2018

Though they populate an industry that prides itself on tackling and solving complex puzzles, many IT vendors prefer simplistic story-telling. That’s partly due to simplicity being easier to sell than complexity, even if it fails to address many or even most of the issues related to complicated engineering efforts. But simple tales also feed the industry’s love of self-promotional mythologies, including the triumph and massive remuneration of plucky entrepreneurs.

I raise this issue because storytelling shorthand also tends to infect areas where accuracy is a critical component, like still-emerging technologies. Keeping things easy may seem to be beneficial in terms of helping an audience initially understand difficult subjects. But relying on simplistic exposition can also mask over-inflated claims and promote questionable reports about a technology’s potential for commercial success.

We’ve seen this dynamic occur many times in the past—virtual reality headsets and associated technologies are just one good example. More than four years after Facebook paid an unprecedented $2B for VR start-up Oculus—a deal that was supposed to rapidly propel VR into the commercial mainstream—the industry and vendors continues to be hindered by many of the same core technological barriers that existed in 2014.

So, it’s a pleasure to find vendors that are willing and able to discuss complex work in both realistic and understandable terms. That was certainly the case at Intel DevCon 2018, the inaugural conference for artificial intelligence (AI) developers that Intel hosted recently in San Francisco.

What’s wrong with AI

Many outside the industry have likely been shocked by the massive bump in attention AI has sparked during the past year. That shouldn’t be a surprise given the decades of attention and content that purely fictional AI has sparked. Those range from the politely homicidal HAL in 2001: A Space Odyssey to the unapologetically murderous leads in the Terminator franchise to lighter and more nurturing presences, like the 1960s Lost in Space Robot.

In other words, during the decades when serious work in AI was being pursued by universities, government research facilities and commercial companies, it was also being used and manipulated as a popular trope by the entertainment industry—never a good means for promoting technical accuracy and understanding.

What incited the recent spate of interest in and focus on AI is the happy confluence of new, affordable technologies that reduced its time, power and investment requirements by orders of magnitude. Quite literally, what once required millions of dollars to support weeks or days of intensive processing can now be achieved in hours or even minutes with tens of thousands of dollars of investment.

In other words, AI processes and projects are increasingly commercially viable. That led to a Gold Rush mentality among some vendors, along with spotty, often unsubstantial claims about what they could do and help others achieve.

However, like many other complex computing methodologies, AI is an area where specialized expertise can reward both vendors and their customers. Problems also occur when vendors over-promote the roles and abilities of their AI solutions, customers buy into those claims and everyone ends up with egg on their face.

What Intel is doing right

For all practical purposes, Intel has doubled-down since AI technologies were central to the news it shared at CES 2018 in January. Consider this also: In the last half of 2017, Intel made twenty announcements in which AI played a primary or highly supportive role. In contrast, during the first five months of 2018, AI has been front and center in thirty-seven announcements.

Quantity doesn’t automatically indicate quality, but I’d argue that much of the company’s AI news has been substantive and forward looking. Moreover, Intel should be lauded on three fronts. First, it is doing an admirable job of realistically educating customers and partners about the possibilities and limitations of AI (see the recent Intel post defining six key AI terms, and Intel CEO Brian Krzanich’s call for a national strategy on AI).

Second, the company is doing an admirable job of evolving its AI-focused silicon. At the DevCon event, that included discussing updates to its family of Nervana Neural Network Processors and announcing that it is targeting 2019 for the availability of its first commercial Nervana NNP-L1000 (Spring Crest) solution.

Finally, Intel continually emphasizes the work and qualities that its partner OEMs are bringing to the AI table. That was clearly evident at the AI DevCon event in San Francisco, where partners, including Google, AWS, Microsoft, Novartis and C3 IoT were spotlighted and executives from those organizations joined Intel onstage.

These sessions weren’t simplistic exercises in mutual backslapping, but instead focused on how the relative strengths of each organization resulted in larger, more beneficial outcomes. That emphasis on innovative collaboration began with the keynote from Naveen Rao, VP and GM of Intel’s Artificial Intelligence Products Group.

Rao came to Intel in 2016 with the acquisition of Nervana Systems where he was co-founder and CEO. During his keynote, Rao laid out the argument he did in Beyond the CPU or GPU, a post that acknowledged that “AI isn’t something Intel can do alone” and emphasized the need to develop solutions “that offer the largest breadth of compute, with multiple architectures supporting milliwatts to kilowatts” of performance.

The sessions at Intel’s AI DevCon focused on a wide variety of projects, implementations and practice areas. There was a great deal of talk about the benefits AI can and will deliver in critical business processes and work in healthcare, scientific and environmental research. But other sessions focused on media and entertainment, including how AI is bringing jaw-dropping visuals to mass market films and the ways that AI-enabled technologies can contribute to popular music.

Final analysis

In other words, rather than following an easy yet inaccurate path of dumbed-down messaging and claims surrounding AI, Intel is instead taking the high road by accurately and compellingly discussing the complexity of artificial intelligence and its attendant technologies. In addition, the company’s willingness to spotlight partners’ contributions to its efforts offers a welcome alternative to vendors who claim, through one form of posturing or another, to own sole leadership in AI.

By following this approach, Intel is not merely acknowledging technical and market realities. It is also providing customers the information they require to learn about and discuss AI, to understand how these technologies can and probably will impact their organizations, and to make informed decisions about the best technologies and solutions to use for their own efforts. For its inaugural AI developer event, Intel set and successfully cleared a high bar. Whether other vendors will attempt or achieve as much remains to be seen.

© 2018 Pund-IT, Inc. All rights reserved.