What Will Big Brains Do?

In the last fifty years nearly all predictions for technology in the next fifty years rest upon the reasonable assumption that computer power will increase. Smart machines play a role in just about every scenario of the future we have, including most dystopian ones. The apocalyptic worlds of the Terminator or the Matrix or Bladerunner are scary precisely because smart things have gone amok. This universal expectation of smarter machines is based on the eerily steady and hard-to-ignore rise of computing speed over the last 50 years.

As many observers of technology have pointed out, the rise in computation is not just increasing, the rate of its increase is increasing, which means simply the powers of computers are accelerating. So relentless is this acceleration that if it were to continue for much longer, the kinds of advances we've seen from the birth of computers till now would repeat itself in only a few years, then a few months, and finally a few days. This means that from our view the changes in computer power would seem to be growing infinitely fast.

Setting aside the possibility of whether computational technology could ever reach infinite growth, at some point before this stage computers would certainly be many millions of times as powerful as they are now. Given our experience of computer growth in the last 50 years, it is perfectly reasonable to accept the proposition of future computers several millions of times faster than computers now, and not too crazy to imagine that threshold achieved in our lifetime.

But while we find it easy to accept this premise, I've found it extremely hard for anyone (even artificial intelligence experts) to imagine what precisely a smarter computer would be like. If you try to describe intelligence smarter than a human you normally go blank after the common first idea that it would think faster. Once it thought faster, would it have different thoughts, would it have a different type of intelligence, and is there any way we can imagine what a more powerful mind might do?

The difficulty of peeking into this alien world of higher intelligences is one reason the Singularity metaphor has caught on. A cosmological singularity, such as a black hole, prevents outsiders from gaining information about what happens beyond the black hole's boundaries (although the strict validity of this notion is now under revision). A technological singularity means that a near infinite acceleration of change prevents us from forecasting or even guessing what happens on the other side of this change paradigm. In this metaphor, our lives are so slow compared to the speed - on the other side - that our minds are incapable of comprehending a super fast, super powerful super intelligence.

It's a good theory, but it is probably wrong for a number of reasons. For one thing we already have experiences with brains bigger than ours, with intelligences smarter than us, and with intelligences different than ours. It is worth investigating the nature of these alien intelligences because what the inarguable acceleration of computation points to is a future where technology becomes more like a mind. If technology wants to be more mind like, what can we surmise about greater minds?

Another way of stating the quest: Everything we know about the current trajectory of technology today suggests it is headed towards becoming very intelligent in the future. What can we say about how greater intelligence works, and what it might want?

Posted on November 16, 2004 at 12:05 AM

Comments

I don’t know that I agree that a faster computer is necessarily ‘smarter’

How does a computer learn? How does it find new ideas and build on them? How does it make connections?

I see a faster computer as being possible to explore every possible chess move, but how would it invent the game in the first place?

Posted by Daryn on March 28, 2007 at 6:25 PM

It will be like magic. Propose a goal and it will make it true in a way we will be unable to imagine. Like magic. Repair the car. Operate the cat. Open that safe. Make a gold ring. I imagine a mind like Feynman’s, with memory of the raw materials required, where they could be found, how to use them, able to work out a plan of how the requested thing could be achieved.

Posted by jaimito on May 14, 2005 at 5:37 PM

I don’t think that speed is the only factor to consider. The main limitation of our brain is focus. We can only consider one thing at a time. An advanced system could consider many things at once. I think that omniscience will mark the begining of the “singularity”. After that we will just be left to ponder the meaning of our existence.

Posted by Greg Stephens on April 18, 2005 at 8:32 PM

It seems to me that the hallmark of a more powerful intellegence is not it’s quickness, but how better it is at picking out patterns in space and time. To assist this, wouldn’t an advanced and evolvable intelligence desire a better sensorium? Give me five channel color vision instead of three, sure, and more across ever increasing spectrums. We use our bodies to compute problems, thinking kinesthetically (think how you trace the movements of a gear train with your finger) - so certainly they’d want to manifest themselves physically, to be as dextrous as possible. But new senses as well, to directly perceive previously invisble patterns. So we can look to nature to get ideas here, from slime molds on up. (I’m wondering what forms their synesthesias would take, what their Kandinskys would paint).

When I wrote patterns in time I was really thinking about capabilities in planning, not about some great Tralfamadorian perspective from the fourth dimension.

Wouldn’t their languages include ones more ambigious than ours, to make for more profound serendipity, for better poetry? Isn’t intelligence all about metaphors and models?

When I start to think of intelligent devices, one of my first questions is: what is the nature and extent of my relationship with this device? Normally I just want the device to serve me; now I have to consider how it wants me to serve it in turn. I don’t mean this in a negative sense. With intelligent humans, I want to amuse and provoke, to contribute and to earn respect.

How do I make this device laugh?

Great fun, the Technium. Thanks for thinking out loud.

Posted by Randy Fischer on December 9, 2004 at 6:07 AM

Post a Comment

Name:


Email Address:


URL:

Remember personal info?
YesNo
Comments:

Captcha:

Type the characters you see in the picture above.