Once AGI is created, we humans face a sheer vertical wall that calls to mind the setpiece from the legendary “Princess Bride”: The Cliffs of Insanity!!!! So steep, so tall, that they simply were incomprehensible to most all humans (except, of course, the Dread Pirate Roberts, our delightful antihero). This is the allegorical equivalent of what we face: the foretold Intelligence Explosion.
I struggle daily to put this — the imminent arrival of AGI, or Artificial General Intelligence — into proper perspective for my friends and family. But along the way, I’ve found a few key quotes and insights that help to frame it quite well.
So, here they are, for your edification and enjoyment:
The Harbingers of the Intelligence Explosion
“When you are driving a car as a new driver: you drive, and you learn.
During that time, you improve your skills. But also during that time, only one person is getting better at driving: YOU. When the AI drives a car, it learns too. The difference is this:
When one AI learns, ALL the AIs learn… not just the ones on the road today, but all the AIs that have yet to be born, too. The osmosis is instantaneous. And with each new AI on the road, driving and learning, the rate of learning increases that much more…”
— Sebastian Thrun
head of Stanford AI Lab
former head of the Google self-driving car project
The next is a short story by one of my favorite thinkers on the topic, Sam Harris. Here, I paraphrase a key anecdote from his seminal TED talk… talking about the Intelligence Explosion waaaaaaaay back in September 2016. Truly, he was one of the first to grasp the magnitude of what was about to happen:
“Let’s say that we succeed in creating this thing…
general-purpose AI, “AGI”…
not a superhuman AI,
but simply an AI that’s “as smart as us…”
say, as intelligent as the average professor at MIT.
When we do that, we copy and paste it a few times — Make simple digital duplicates of the codebase, and allocate sufficient electrical power and computing resources — so that we have 10 copies.
And we set up lines of communication between them.
Now we have the digital equivalent of a conference room with a team of 10 MIT Professors collaborating within it. A formidable intellectual force, but nothing that humans haven’t seen before.
Except for one…
simple…
fact:
Speed.
The nanoscopic transistors that make up the AI brains — due to the physics of electronics and silicon circuitry versus the inherent limitations of our biochemical “wetware” brains, which power our thoughts — perform their digital thoughts at approximately 100,000 times the speed of human neurons.
What that means is this:
Within the first 7 days of this AI running,
it has already thought, ideated, collaborated and experimented for
the equivalent of 20,000 human years.
Can you imagine what 10 MIT Professors
could think up and execute,
given 20,000 years to do it in?
That’s what will happen
…In 7 very short cycles of the sun rising and setting.
In the first seven days of the Age of AI.
That’s why,
once this thing is finally launched (unleashed?);
things are going to happen.
Very.
Very.
Quickly.
So, what that does, is it puts the global tech giants, and the global political superpowers, into a de facto arms race.
Because… in this race,
if you’re 3 months behind in launching your AI,
you’re effectively 500,000 years — conservative estimate — behind the winner.
And the prize?
The prize is, The World.
Winner takes All.
In other words, if you’re not first, you haven’t just lost, you’ve already been decimated…
is it any wonder, then, that Google, Microsoft, Amazon, Tesla and Meta are each throwing well north of $10 billion a year at AI “research”? This isn’t research… this is a race.
The question is, will humans win?“
—Sam Harris
TED Talk 2016
“Can We Build AI Without Losing Control?”