DARPA can't find a contractor who can build a self-driving car, so it stops looking for one and pays whoever finishes a desert race instead

问题

The technology needed doesn't exist yet, so there's no established contractor or team qualified to be awarded a development contract for it

背景

换别人会怎么做

Issue a request for proposals to established defense contractors and evaluate their competing plans on paper -- the standard procurement process, which assumes a pool of qualified vendors already exists to compare and select from and simply doesn't work when no one anywhere has built anything close to the capability being sought.

他们看到了什么

DARPA saw that the bottleneck wasn't picking the right contractor from a pool of qualified bidders, it was that the pool didn't exist yet, and no amount of paper evaluation could substitute for actually testing unproven approaches against reality. An open prize with a hard pass/fail test let anyone with a credible approach prove it directly, rather than requiring an agency to guess in advance who was worth funding.

那一手

The March 2004 race drew 15 finalist teams ranging from established robotics labs to self-funded hobbyists, and every one of them failed the course outright — the best-performing vehicle traveled just 7.5 miles before getting stuck. Rather than treat that as proof the approach didn't work, DARPA re-ran the race in October 2005 on a 132-mile Nevada course with a doubled $2 million prize, still open to anyone.

为什么管用

Because the prize was open to any team regardless of defense-contracting history, it pulled in academic labs and self-funded hobbyists who would never have won a traditional bid but who brought genuinely novel approaches, and because the test was a hard, objective finish-the-course benchmark rather than a subjective proposal review, DARPA didn't have to correctly predict which approach would work -- the desert simply proved it. Even the total failure of every team in 2004 wasn't wasted: it mapped which approaches were worth funding and refining for the next round, producing Stanford's winning vehicle within 18 months of a race where nobody finished.

值了多少

In 2005 five of 195 registered teams finished the course; Stanford University's vehicle 'Stanley,' built by a team led by Sebastian Thrun, won in 6 hours 53 minutes and took the $2 million prize — a result achieved not by an established defense contractor but by an academic team assembled specifically to chase the open prize.

什么时候会失灵

This approach depends on the prize being large enough, and the underlying capability being achievable with a plausible amount of independent effort, to actually attract serious teams -- a prize too small relative to the R&D cost required draws no credible entrants, and a challenge that's simply beyond reach regardless of approach (not just beyond current state of the art, but genuinely infeasible) produces total failure with no useful signal at all. It also only surfaces a solution if enough diverse teams enter with different underlying approaches; a challenge that draws only a handful of teams all pursuing the same known technique offers little more information than a single funded contract would have.

后来呢

Sebastian Thrun went on to found and lead Google's self-driving car project, the direct ancestor of Waymo, carrying the approaches proven in the Grand Challenge into commercial autonomous driving; DARPA's Urban Challenge follow-on in 2007 and the broader autonomous-vehicle industry that followed are widely traced back to the open-prize format that let outsiders with no defense-contracting history compete on equal footing with anyone else.

资料来源

1. [1]Robo-car enters Smithsonian collectionSmithsonian Insider, 2009insider.si.edu
2. [2]Stanford Racing Team — StanleyStanford University, 2006cs.stanford.edu

收下它

同一路聪明

• 2000 · AIGA Design for DemocracyAfter Florida's 'butterfly ballot' cost thousands of voters their intended vote, the people who fixed it weren't election lawyers — they were typographers.
• 2001 · Richard Ankrom / CaltransAn artist tired of missing his exit built a Caltrans-perfect sign and installed it himself
• 2017 · Behavioural Economics Team of the Australian Government (BETA) / Australian Taxation OfficeAustralia's tax office cut inflated deduction claims by writing to the accountant, not the taxpayer