Dawn of AI

What we can expect from the next phase of evolution.

Lately I’ve been doing a lot reading regarding the different views around the development of Machine learning and Artificial Intelligence. If you have read anything at all on this topic you’re probably already aware of all the amazing (and frightening) stories on what the future might hold for us. This blog post is a summary of a couple interesting articles, debates and prophecies on the super intelligent era.

If you’re interested in learning more about the subject you can find more in depth reading and videos at the bottom of this post.

 

What is Artificial Intelligence? (AI)

We can narrow it down to three categories. Cool, super cool and frighteningly cool.

The cool one.
This is the state we currently find ourselves and have been for quite some years.
This is called Artificial Narrow Intelligence (ANI). At this level our computer can be incredibly good at one thing. Every calculator could be referred to as an ANI. Recently an ANI built by Google beat the best Go* player in the world. Something many expected would take decades to achieve. *Go is considered to be the most complex game in the world.

The super cool one.
Artificial General Intelligence, or AGI as we will call it, is as smart as a human. But since digital processing is so much faster than biological processing this will surpass human intelligence.

Think of AGI as an infant but with an incredibly fast learning curve. AGI can think abstractly, plan and learn from previous experiences. When we reach this level of Artificial Intelligence development can go incredibly quickly.

The frighteningly cool one.
Artificial Superintelligence (ASI) is smarter, more creative and better socially than humans. When we reach this level of AI there is no telling how quick or how far development can go. An ASI could potentially be millions or billions times smarter than humans. This is what has caused debate in the science and technology world. An ASI with good intentions could possibly cure every illness in our world or even make the humans the first race to reach immortality. An ASI with bad or wrongly programmed missions could end humanity before we even saw it coming.

Sounds far fetched? Not according to some..

“With artificial intelligence we are summoning the demon” Elon Musk 

“The development of full artificial intelligence could spell the end of the human race” Stephen Hawking

“Machine intelligence is the last invention humans will ever make” Nick Bostrom

 

So what exactly is happening now?

At the moment we’re most likely somewhere in the late era of ANI with a rapid increase in how smart our ANI’s are becoming. At the beginning of ANI we had calculators that were as good as any mathematician. Today we have cars that drive themselves as well as any driver could. Better than most drivers in fact – statistically. ANI’s at hospitals are making more precise diagnosis and Elon Musk’s Space X rockets can now reach the International Space Station and return to make a landing on an X in the middle of the ocean. These things are all incredibly cool and amazing but we still can’t have a computer read a book and have it understand the meaning of it. This is what’s difficult to achieve and there are a multitude of companies trying to get there first. Google and Microsoft are among some of the companies investing billions of dollars into developing Artificial Intelligence.

 

– Yeh yeh. Tell me when skynet is coming.
Skynet is probably not happening and we will most likely not have AI that sets out with bad intentions. The AI would more likely be programmed to have the best of intentions. The concern a lot of people have is how AI would interpret what would be best for humanity or how a super intelligent AI would value life compared to its goal.

Sam Harris talks about this in his Ted Talk: Can we build an AI without losing control over it?

He makes the comparison on how we view ants. Most people wouldn’t consider themselves evil if they decided to build their house on top of what would be the home of millions of ants. Ants don’t live for more than a summer anyway you would probably argue. And they shouldn’t come in the way of achieving your goal of having your new house built.

How would an ASI view humans if it turned out we are the thing standing in the way for the ASI to complete its goal? Tim Urban has written a scenario on how the first encounter with an ASI with a good intention might turn out. Read his article on ASI here (quite the read…!)

This “intelligence staircase” can give you a sense of why people are a bit worried about how an ASI would view humans. Don’t take this one too seriously 🙂


As you can see we’re way smarter than ants. But unfortunately for us (if the AI goes rogue) the ASI will most likely be infinitely smarter than us. And considering how lesser we view species with less intelligence than us, it’ss likely an ASI could view us as something that’s just in its way of achieving it’s goal.

Back to the question of when we can expect to see such an intelligent machine. Most scientists and technology experts believe it will happen within the next 100 years. Half of the community believes it will happen before 2050 and about a third believes it will happen before 2030. What everyone agrees on is that when we reach ASI the acceleration will be absolutely mental.

Here is a scenario: Imagine the first AI reaching human level intelligence by 10 am on a Sunday morning. Five minutes later it would be twice as smart as a human. One minute later it would be ten times as smart. Ten seconds after that it would be a hundred times as smart. By the time it had dawned on us that we’d created something a thousands times smarter than us the ASI would most likely have a counter measure to everything we could do to shut it down! Mind blowing stuff indeed.

So when we reach the point of no return where we are able to create an ASI (whenever that might be) we can only hope that it’s done in a way that benefits us rather than sends us into extinction. If done correctly we can, in theory, unlock all the secrets of the universe.

 

Learn more

Here’s a list of some interesting stories, articles and videos I’ve found. If you’re interested in learning more about the topic I recommend you to check out some of these links.

Tim Urban – Artificial Intelligence revolution

Sam Harris – Can we build AI without losing control over it video

Nick Bostrom – What happens when our computers get smarter than we are? video

Google AI beats Chinese master in ancient game of Go

Facebook shuts down robots after they invent their own language