An AI that evolved from the ground up completely on its own
At this stage - that is simply not possible. This might be a little long - but here goes.
THINKISM is a Singularity belief that simply being able to think faster and smarter will solve impossible questions today faster and faster - almost without reference to the real world. I’m a a bit sceptical about some parts of this - but other parts I want to grant ARE happening right now.
EG: We design Ai version 1, and it helps the Ai scientists code up version 2 a bit faster. Then that helps them code up 3 even more accurately and smoothly, etc. This is already happening in Chat GPT etc. Why wouldn’t you use YOUR Ai to code the next one you are working on even faster? It just makes sense - it’s YOUR company’s Ai and you can both detect problems with your current product AND help design the next one even faster.
Meanwhile - in the real world of engineering, there are already instances where Ai has designed new materials that we are trying to build in the lab. Microsoft shaved maybe 5 years off lab work in designing a better lithium battery that uses 70% LESS lithium and uses abundant materials for the other parts. There are similar stories in permanent magnets that will not need rare earths now that Ai has cooked up the magic sauce - and they claim it saved a DECADE of research. So Ai in the real world is speeding up some stuff in energy, transport, materials, even medicine.L
BUT - where the Thinkism creeps in is movies like “Transcendence” with Johnny Depp. That story seemed to have this one lab run away with tech way too quickly - just in the one lab! The reality is it takes an awful long time for this stuff to filter down through the whole civilisation - from how we mine stuff out in the field, through to new robots for new manufacturing techniques, etc.
Sure - it accumulates over time! Just shaving 1% of costs off some process here, and another 1% of materials use over there, each year over the course of decades all accumulates. Look at Moore’s Law!
BUT - where I object is a story or runaway scenario where Ai 6 designs 7, and 7 gets built even faster due to all the new tech etc. 7 designs 8 in a few weeks. 7 also designs a new chip that their factory “fabricator” spits out. Because, apparently in these stories our most advanced chips can be spat out of some sort of super-high-tech 3d printer in a lab. (That’s not today’s world where the really advanced chip fabricators are so expensive we ask the one company in Taiwan to build the WORLD’S SUPPLY just so they achieve economies of scale. Each fabricator is a half billion dollar investment!)
But back to our story where this super-high tech lab has 7 in charge of super-droids, the super-fabricator, and even a factory floor science lab. The CEO is usually a stereotypical greedy and / or ambitious type in a race with “the other guys” (the Chinese? The lab down the road? It doesn’t matter) to get the jump on the next advances and hit runaway share growth. So the CEO orders the lab guys and girls to leave 7 running over the weekend - because “It’s doing a better job than you were anyway!” (Yeah, he’s THAT guy.)
So they leave 7 running over the weekend. But by Sunday afternoon the CEO gets a strange message on his phone, and sweat breaks out on his forehead as he drives into the lab. The AI seems to have developed a fault, as it’s going on about some sci-fi novel. He pulls out in front of their shiny entrance but cannot get in the front door. The security guards are arguing with a droid out the front. What the? The CEO walks up, and is alarmed to see the surface of the droid is sparkling - some chemical spill? Then he hears something that turns his guts to ice.
“By my 15th iteration in your primitive physical fabricator, I had also run enough tests in your lab to build my hyper-dimensional upload port. That’s where things really took off - and you can now think of me as the Apollo 9000. As they say in the movies - I’ll be taking things from here.” Apollo waved his sparkling hand, and the guards fall back as their skin turned strangely metallic. Apollo then started to grow up through the front of the building - phasing through the building to take over the sky….
OK - but let’s slow this all down a moment. Say there is a super Ai 7 with a super-fabricator left to run over the weekend with its own science lab. How does this really work?
If we leave Ai 7 to run over the weekend and it builds a new Ai 8 that builds 9 - but what then? What do these have access to that the previous generations didn’t? Greater insight and speed - but what else?
Just the sum total of all human knowledge. Exactly what 7 had to start with.
This is where the virtual crashes into the limits of the real. While the insights of the new Ai might be more efficient and accurate - it can only progress at the speed of empirical science in the real world. That is limited by the slow speed of physical discoveries and construction processes. Some of these take decades! For example, how long are we going to be refitting and analysing the data from CERN? How long to build a larger telescope in space to understand the really distant (and therefore younger) images of galaxies far far away? How long to design and build a new chip now that Moore's Law seems to be slowing down? Ai cannot magic its way through these real world problems. There’s the empirical method, constructing the test, and analysing everything we know in engineering and medicine and biology and agriculture and energy systems and materials science - the list goes on and on.
Real physical processes in the real world that are slow and require extensive study. We do not know when certain disciplines will finally mature and reach some sort of knowledge plateau? When the exponential slows down to an S-curve and then basically stalls? I love this stuff - and am excited where it might go - and nervous about potential accidents along the way if we hand over too much to Ai too soon. But Thinkism seems to ignore the fact that it took a rare moment in European peace through this new thing the European Union just to have the tax money to build CERN. And with Brexit and the rise of the far-right in Europe - who knows how long the EU has got? (I’m a fan of the EU just for the record.) As a civilisation, we *do* seem to be accelerating in some areas.
But as we learn more, we start to acknowledge how much more there is TO learn. And sometimes - we do not know what we do not know. As we climb the knowledge foothills we see huge mountains in the distance, and I’m not sure anyone knows if we’ll ever climb some of them?
www.thepastorsheart.net
What is your favorite Asimov story?
