Gene drives and community input
Hi! This is Alyssa Vance, from the EA Global conference. I saw your talk on gene drives, and while I think the social and legal concerns are real, the approval processes you talked about seem likely to fail in the long run. I'd like to share some relevant experiences from living here, in the San Francisco Bay Area.
In the 1950s, like most US cities, San Francisco was run from the top down. Planners could do whatever they wanted, and they'd just show up to your house, announce there was a new freeway, and your neighborhood would get demolished a month later. After years of this, people started to object, and the city created more mechanisms for direct democracy and community input. At first, it did a lot of good: bad projects could be stopped, and the harmful effects of good projects could be reduced by working with neighborhoods.
However, once this was in place, planners and developers started to bargain with influential community members to get things done. Maybe you objected to this project, but in exchange for not campaigning against it, we'll donate to a wildlife protection fund. Or we'll build it, but in this other neighborhood, instead of yours. Or we'll spend a bunch of extra money on noise mitigation, or hire only union construction workers, or cancel another project that you don't like even more; whatever people in that area were interested in.
Once this process started, two things happened. First, it created a powerful incentive to object to everything, even projects that made sense. The harder you objected, the more you fought, the more people you could turn out to oppose something, the better a bargain you could get in exchange. Second, it created an incentive to add more and more layers of approval. If you ran a nonprofit, or a neighborhood group, or a government agency, you wanted to be involved in the process - if people needed your permission to do something, you could get concessions in return for not making a fuss. So every year, more and more meetings and discussions and approvals got added. It was a democratic process, after all, and who could object to democracy?
As a result, after fifty years, San Francisco is literally falling apart, because it's so impossible to get anything done. Just as one example, the city wants to put an express bus route down Geary Boulevard. Planning for Geary bus changes started in the 1990s, and the city got voter approval in 2003, when 75% of San Franciscans voted to fund it with a sales tax increase. But construction still hasn't started. It's now expected to begin in 2019, in the unlikely event there are no more delays. The streets are full of potholes; transportation is the slowest of any major US city; there aren't enough shelters for the homeless, who live in tent cities on the streets; there aren't enough cops, so thieves chop up bikes in broad daylight downtown; trains run over bridges built when Teddy Roosevelt was president; BART is still using its original subway cars, from when it opened in 1972; the average rent for a one-bedroom apartment is $3,500, and more in a good neighborhood, since it takes a decade of bureaucracy and litigation to build any housing. It's a godawful mess, and it just keeps getting worse, even though the city's budget is $10,700 per resident (for comparison, Boston's is $4,400). If gene drives follow this path, I worry that every year, they would get slower and costlier and more bureaucratic until scientists just gave up trying.
Thanks! In general, I agree about open science. In particular, I find it very annoying when scientists publish papers without their data or source code. However, I think there's at least one area where that might be too dangerous, and that's advanced artificial intelligence. Right now, AI is mostly harmless, and will probably stay so for a while.... but imagine if someone discovers how to create fully intelligent, conscious beings, and puts the source code on BitTorrent. If computers are fast enough by then to run lots and lots of AIs, then that means anyone who's grumpy and having a bad day at work can create their own Holocaust on a home laptop. Of course, a government could make that illegal, and most people would follow the law, but the numbers here might be so lopsided that it doesn't matter much. If 99.9% of people follow the law against murder, then any group of a thousand will (on average) have one murderer, one murder victim, and 998 people who just live their lives normally. But if 99.9% of people follow no-AI-abuse laws, then in a similar group, you'd have 999 normal people, one sadist, and 1,000,000 people whose whole lives are pain and suffering. I work in machine learning myself, and I'm not really sure what to do about this problem... what do you think?