Artificial intelligence runaway growth…


I can tell you this straight up that trading is essentially stealing other people’s money. It doesn’t contribute much to productivity in the global economy. Artificial intelligence will bring about singularity. Infinite growth, resources and production will be happen through artificial intelligence. I want to highlight a scenario of how we can have massive runaway growth. These days, we regard 3% productivity growth a year as a lot in the United States. But I think with cheap artificial intelligence beyond human levels we can get growth rates in excess of quadrillion fold a year or even googol fold a year. To give an idea it takes over 1,100 years of growth at 3% a year to reach quadrillion fold. Poverty would be instantly eliminated within a few months of cheap artificial intelligence. I believe that investments in artificial intelligence, robotics and quantum computing will far outperform investments in other fields.

I wanted to point this out that technology is getting dramatically cheaper at an accelerating pace. A $1,000 iPhone today would have cost $1 trillion+ to make 50 years ago. Let’s take this as an example. Right now productivity per capita in the United States is about $50 per hour worked. Let’s say for instance you have a robot that is able to work 140 hours per week that can match or exceed human productivity per hour. Let’s say half the time the robots are spent making more robots and the other half the time the robots are spent making goods and services for humans. So the robot is doing $3,500 of work a week for humans and $3,500 of work a week to build more robots. Let’s make another assumption, it costs $3,500 to build the same robot. So the robot economy would keep doubling every week. Under this growth scenario an economy of 1,000 robots that cost $3,500,000 to make initially would produce 1,000 quadrillion other robots over a year while producing millions quadrillions of goods and services for humans.

What does it take to building cheap artificial intelligence? Right now we have hardware limitations. Simulating human intelligence probably requires at least 10^16 calculations per second along with a terabyte of hard disk space. Right now Google’s Director of Engineering Ray Kurzweil predicts we will be able to produce machines with such power for under $1,000 within a decade. Beyond the hardware requirements, the software that simulates human intelligence will have to be produced. This is where it gets a little bit tricky. Right now we have produced software that can beat us on complex games such as chess, Starcraft, mathematics and beyond. I should note that our brains are electrochemical while computers are electrical. Our brains are essentially a complex network of trillions of neurons that fire yes/no signals just as computers fire 1/0 signals. We will likely develop technologies that copy what our brain does into a computer. Yet we will enhance that with more processing power and electrical efficiency.

Leave a Reply

Your email address will not be published. Required fields are marked *