Aardvark DailyNew Zealand's longest-running online daily news and commentary publication, now in its 25th year. The opinion pieces presented here are not purported to be fact but reasonable effort is made to ensure accuracy.
Content copyright © 1995 - 2019 to Bruce Simpson (aka Aardvark), the logo was kindly created for Aardvark Daily by the folks at aardvark.co.uk
Please visit the sponsor!
The human brain has around 86 billion neurons.
The original 4004 microprocessor chip had just 2,300 transistors.
Now while a transistor is not identical in function to a neuron, it can be considered a comparable metric, well at least that's what I'm going to use today.
For many years, we (that's a collective "we"... as in "mankind", not as in the whole-wall urinal you find at the local pub) have attempted to create artificial intelligence by way of machine-learning systems.
Quite some time ago these systems were attempted on hardware that was purpose-built as a "neural network" of artificial neurons. These days however, we tend to use conventional CPUs and create those learning networks in software.
Having said that, NVidia has more recently come out with AI-specific processors and even has something called a "tensor core" in its GPU cards.
So how close are we to creating our own synthetic brain with consciousness?
Well if transistor-count is anything to go by, we must be getting pretty damned close.
Back when the 4004 was first designed, 86 billion neurons must have seemed like an incredible number. To get the same number of transistors using those first microprocessors would have required over 37 million of those devices.
A somewhat impractical concept.
Today however, we've come a long way since those early days... a very long way.
The AMD Epyc processor has just short of 40 billion transistors, for example. That's half a human brain's worth, in a single CPU!
With fabrication capabilities now providing for such astounding numbers, what happens if we build a neural net CPU with this sort of capacity?
What happens if we then connect that CPU to the internet and leave it to learn, all by itself?
Would this even be a remotely good idea?
It is becoming very clear that true AI and the potential for such a device to develop a consciousness may be a lot closer than we'd care to admit.
Sure, individual neurons have a far greater number of network connections than a single transistor offers -- but it's only a few orders of magnitude more and that could mean that a rack of top-end CPUs could come close to matching the raw processing and learning power of a human brain.
Give us another decade or so and we'll likely have all that power in a single CPU, at which time "artificial intelligence" will likely take on a whole new level of importance.
Although we already have some huge systems (such as the Google online community assets) currently being administered by AI/machine-learning systems, chances are that these will be unrecognisably smarter within the decade and may even develop their own "personality traits", whether we want that or not.
SkyNet is currently the realm of science fiction -- but for how long?
How far away is that milestone day when a machine first becomes self-aware and what will we do when that happens?
Have we adequately thought through the ethical and moral implications of creating machines that have a consciousness and are therefore "alive"?
The future is a scary place sometimes and we're on a journey right into the heart of it with no way to stop or get off.
Have a nice day :-)
Please visit the sponsor!
Have your say in the Aardvark Forums.