Monday, January 20, 2020

The Last Invention Man Need Ever Make

What exactly makes us smarter than chimpanzees? I would argue that it's our qualitative superiority in intelligence.

For example, a chimp can learn what humans or skyscrapers are. However, they can't learn that humans engineered the skyscrapers. For them, something of that scale is simply part of nature. Not only is it beyond their comprehension to build a skyscraper, the very idea that someone can build it is unfathomable with their neural hardware. 




                    

Artificial superintelligence(ASI), is an agent that excels at everything over humans. ASI isn't just good at trading stocks or just self driving - it excels at everything at the same time, including concepts that humans have never thought of nor comprehend (much like how we're "better" at building nuclear weapons than monkeys). 

The crux of the issue is that with advancements in AI happening at a break-neck pace, we barely have any regulations or discourse around it. There are several reasons why this is such a difficult problem to address. 


AGI: Artificial General Intelligence (aka robot butlers)

1) ASI is a winner-take-all scenario. Once you create an ASI, even just seconds before your closest competitor, the first mover's advantage is infinite. Countries like US, Russia and China have their own agendas and tolerance for regulation. If US imposes regulations on our own development because were "morally superior", then whats preventing the other superpowers from creating it first?

2) Ok then don't impose any regulations so we can be the first to create an ASI right? The issue here is that creating an ASI whose morals are perfectly aligned with ours is a problem we've yet to solve. Optimizing for speed of development will surely lead to catastrophic results (ex: King Midas Problem, AI rights).

3) Ok then create a prototype, then iterate until we have a "bug-free" ASI. After all, this is the ideal way to ship software right? This doesn't work because an ASI is defined by its ability to self-teach, plan, and act. Once we have even a prototype of that, the pandora's box has already been opened. We can make denuclearization treaties, but we can't shove ASI back into its box. 

Due to incompetencies of the public sector, people like Elon Musk has taken the problem upon themselves, but this just opens another can of worms. 

Our ability to bring awareness on ASI and its implications may very well dictate if our last invention will lead to our immortality or extinction.


For the interested:
Ted Talk  <- highly recommend
AI revolution: The Road to Superintelligence


No comments:

Post a Comment

Note: Only a member of this blog may post a comment.