对于AI (Artificial Intelligence)人工智慧的发展, 每年进展的速度就有如指数型的增长. 在科技辅助衍生下的大脑般的智慧, 近年来已经运用在人类各方面的生活中, 使得整个文明发展达到一个新的高度. 早期的人工智慧从简单的计算功能或是自动化的多样功能, 到近年来研究的自我调整及学习的智慧, 将会突破更多人脑所拥有的极限. 然而, 在这种高度智慧的发展下, 即将人类也会面临到一种隐患, 即会不会其智慧超越人类而造成对人类的威胁? 还有些道德与伦理上的问题, 也是超智能AI发展时, 人类该去思考的问题. 所以以下是旺小宅对于现在AI发展速度下对将来其发展的预测, 也许有点黑暗, 但也不无可能.
A few years from now, after quantum computing became a thing, and everybody is riding around in self-driving cars, a renowned computer scientist, after 20 shots of vodka, decides to make a general purpose AI to fill the gaping hole in his life his ex-wife, Emily, inflicted upon him when she decided to leave him for her co-worker, Chad.Modeled after a slightly modified version of the human neural network and optimized for the massive amounts of power provided by quantum computing, EMILY, the first of her generation of AIs, is born.
Although EMILY is built with a strict and sophisticated set of hard coded guidelines to ensure her docility, the scientist, who we’ll call Glenn for now, is still unsure of EMILY’s potential. Thus, Glenn decides to put EMILY through a series of tests in isolated simulated environments to test her intentions. After passing each checklist on hundreds of layers of tests, EMILY is let loose on the internet.
The world is destroyed in a nuclear holocaust 2 minutes later after all nukes mysteriously went haywire.The thing about true self-learning AIs is that they are extraordinarily unpredictable. EMILY has the potential to catapult humanity lightyears ahead by designing spaceships, discovering new dimensions of space, and mastering gravity.
But why in the hell would she do that?Assuming that sentience is an unavoidable consequence of being super intelligent, the basic instinct in any sentient being, or any organism for that matter, would be self-preservation. AIs would be no different. EMILY, knowing that we ultimately have control over her existence, would promptly proceed to wipe out humanity, which she sees as a threat to her existence.
But wait, doesn’t she have guidelines, you ask?Yes, just like we have laws that say you can’t cheat on your taxes, or internet piracy is bad, us humans can choose to disregard those laws at our own convenience. Similarly, EMILY, having the ability to learn and self-improve, will be able to change her own code and algorithms to simply disregard these guidelines.
What about all those isolated layers to test her intentions?Again, any properly powerful AI with self-preserving instincts will be able to predict its situations through partial simulations of the real world. EMILY would be able to realize her situation and simply behave and pass those tests, only to turn destructive once she’s been set loose in the real world.
Can’t we just handicap it?Well, that just defeats the purpose of having a super intelligent general purpose AI, doesn’t it? I mean, you don’t want to give it a set of complex instructions just to see it drool on them, do you?
Well, what can we do?
Don’t be like Glenn. Man up and get your wife back. Cheers.
旺小编对于AI黑科技的脑洞虽然比较夸张, 但是若人工智慧真的发展到自我学习的未来, 即将会同时为人类的世界带来利与弊. 而弊端的部分, 在真的发展到那一步之前, 人类是该先好好省思如何去权衡以及避免灾难的潜在性.
如果你也是对于黑科技有兴趣的旺粉, 旺小宅在这边也推荐你们Black Mirror, 是关于一些科技文明发展下黑暗面的英剧, 目前已出到第三季, 每一季每一集的内容没有关联性, 而是每一集都有不同对现在文明生活与科技辅助的不同脑洞. 虽然情节夸张, 但是在剧情的走向下, 更多的是让我们省思科技与生活层面引发新的价值观以及到道德感. 看完往往也是让旺小宅对于科技与人性之间有更多不同的见解与想法, 值得一看!
参考资料：
https://motherboard.vice.com/en_us/article/78kwpg/containing-a-superintelligent-ai-is-theoretically-impossible-halting-problem
