PHOTO CREDIT: STANFORD UNIVERSITY
The Hot Chips conference is the hands-down favorite of the chip design crowd. The atmosphere is more relaxed, the media presence is less intense, and in general the mood is more congenial than at the heavy hitters like IEDM and ISSCC.
They had good coffee, so everyone congregated on patios between sessions and lingered in the California sunshine. One engineer bore a stunning resemblance to Late Elvis, but turned out to be much more reticent than his lookalike. To encourage the masses back into the auditorium, an ethereal elderly man with flowing white hair, tucked into a slightly-too-large brown suit and formal white shirt, stalked about ringing a shrill bell. I found out later that the wrangler was A Very Big Deal (he was the founder of the Hot Chips conference).
The best parts of the conference, and the parts that best reflected the comparative intimacy of this event, were the keynotes and the Monday night panel that looked back on 20 years of â''missed predictions.â''
The first keynote was about Stanley, Stanford's autonomous VW Touareg that won DARPA's first Grand Challenge to the tune of US $2 million. Head researcher Sebastian Thrun presented an overview of what it took to get from concept to completion, along with clips that often looked like outtakes from America's Funniest Home Videos.
The vehicles â''learnedâ'' to drive themselves based on human tutelage. Thrun described the process: a graduate student would drive the car, the car would record his speed and movements, and later reference them when left to its own autonomous devices. But Thrun said that after several such tutorials, the car had become overly timid.
â''Then I drove it, â'' said Thrun, â''and, as a German, the car became 20 percent faster.â''
PHOTO CREDIT: ANDREW BECRAFT
The second keynote, delivered by SunPower founder Richard Swanson, was when things really got interesting.
SunPower has been developing and evolving their photovoltaics for solar power systems since the late 1980s. Swanson recounted the tough times they faced back when solar was laughed out of the room, and had a bit of a gloat at the idea that in 2008, fossil fuels are expensive enough to make solar cells worth it. Now the company has funding, government subsidies, and lots of investors.
Swansonâ''s first mistake was a back-of-his-palm ad hoc calculation of how much electricity a solar farm could generate over ten years. â''If you installed one megawatt a day in photovoltaics,â'' he said, â''youâ''d have 360 MW at the end of the year, which is 3.6 gigawatts in 10 yearsâ''thatâ''s a nuclear power plant right there! Itâ''s the amount of power a nuclear power plant would generateâ''except if you ordered it today, youâ''d have it in 15 years.â''
Swanson had worked himself into a froth. â''We could supply the entire country by 2040 at reasonable growth rates.â'' A side benefit, he said, would be an 80 percent CO2 reduction by 2050.
The question and answer session that followed was run of the mill (â''What are your dominant failure modes,â'' etc.) until one engineer walked up to the mic, and promptly and thoroughly Hulked out.
His shoulders bulged, his shirt ripped, and his toes tore through his shoes.* â''Sunpower able to do this purely because of government subsidies!â'' he roared. â''Absurd to say you can do something as good as nuke plant! Self-righteousness of keynote speech and play fast and loose with facts make HULK SMAAASH AAARGH!!â''
Swanson disagreed, putting the annual capacity of solar at 30 percent, where nuclear plants are at 80 percent.
â''Nuke plants at 92 percent!!!â'' The Hulk bellowed and tossed several surprised engineers at the stage. â''To say 3 GW absurd!â''
Swanson protested meekly but agreed to downgrade his projections from one power plant to one half of a power plant.
*Disclaimer: I exaggerate somewhat. Having returned to his normal size, the engineer later told me that he thought the projections were off-base, and that his main argument with the talk was that it sounded like a marketing presentation rather than a technical overview.