三个级别的人工智能与失控

三个级别的人工智能与失控

人工智能可能对人类造成某种危害，甚至造成人类灭绝，这是不是危言耸听？如果有人恶意使用人工智能，当然可能会有十分恶劣的结果，就像曾有过的战争狂人或者恐怖分子所做的那样。

但就人工智能自身发展，或者自我演化而言，是否存在失控可能，而失控是否会毁灭人类？

有一种分类，把人工智能分为三类：人工窄智能（Artificial narrow intelligence或者ANI），这是一种单一功能的智能，现在的人工智能还处于这一阶段。第二阶段出现人工通用智能（Artificial general intelligence或者AGI）。最后是人工超级智能（Artificial super intelligence或者ASI）。

我觉得，到了通用智能，人工智能就有了自我演化能力，它们自己会发展出超级智能，所以不存在人工超级智能，或者不需要人造的超级智能，机器自己就造出自己的超级智能。而在机器自己的超级智能出现后，还会有人造超级智能的出现机会吗？

如果出现失控，不用到超级智能出现，很可能在通用智能阶段就会失控。

但失控是不是就很可怕？这要看是什么样的失控？

Three levels of AI and out of control

Artificial intelligence may cause some harmto humans, or even lead to human extinction. Is this alarmist? If someonemaliciously uses artificial intelligence, of course, there may be very badresults, just like what war enthusiasts or terrorists have done before. But interms of the development or evolution of artificial intelligence itself, isthere a possibility of losing control, and will losing control destroyhumanity?

There is a classification that dividesartificial intelligence into three categories: Artificial Narrow Intelligence(ANI), which is a single functional intelligence that is still in this stage.In the second stage, artificial general intelligence (AGI) emerged. Finally,there is Artificial Superintelligence (ASI). I think that with generalintelligence, artificial intelligence has the ability to self evolve, and theywill develop super intelligence themselves, so there is no artificial superintelligence, or there is no need for artificial super intelligence. Machinescreate their own super intelligence.

And after the emergence of the machine'sown super intelligence, will there still be opportunities for the emergence ofartificial super intelligence?

If there is a loss of control, it is likelyto occur in the general intelligence stage without the need for superintelligence to appear. But is losing control scary? It depends on what kind ofloss of control it is?