But why wouldn’t it happen? We know it is physically possible for genius level IQs to exist. Terrence Tao is an amazing mathematician. People like Einstein or Witten are great physicists. Paul Krugman is an amazing economist. We already know it is physically possible to be a groundbreaking genius in every field on earth because people have done it.
So even if the ‘best’ we can ever hope for is genius level talent in math, science, technology, social skills, etc becoming available to the masses it will still revolutionize the world.
But I’m not sure why that would be the ceiling. If intelligence rests on things like working memory, crystallized intelligence, pattern recognition, creative solutions to achieve goals, etc. then those should go far higher than we are biologically capable of.
In my view, it is more a question of when we transcend our biological limits to intelligence.
I’m glad to see this is getting shot down by the research community. My initial response upon hearing about the cat brain simulation was “Wow, that’s amazing!” Then I went and actually read what they’d done and I was like “Why is this news?” Hooking up a bunch of simple neurons is not simulating a cat brain, no matter how many there are!
To the bolded: Duh. But that’s irrelevant, we’re talking about machine intelligences not violating physical limits. Some processes in your body do emit and absorb photons, so you do emit light to some degree, these processes can be regulated, it’s merely a matter of arranging molecules. Right now what is holding up Nanotechnology is the talent more than anything. People are going to be dissecting molecules and working out how they can use them to make machines, like legos. That’s just a matter of figuring out individual molecules, and there’s a whole lot of that to be done.
There is a bit of sight one has based on sustained performances, if we couldn’t we wouldn’t build anything at all. So we have to base our predictions on our imaginings of future performance. But that’s what information is for, certain things can be relied upon. There are a lot of known unknowns out there to be explored, and we can see it envision it. So it’s not like we are talking about things that are impossible, like violating physics, we are talking about merely theoretical, as in practicum hasn’t caught up with theory yet, but it’s more of a logistics problem than it is an intellectual one as we already understand more or less how the individual elements work on their own. This sort of work is already being done. Miniaturization adds layers of complexity, if something can be built at the molecular level you can build a very sophisticated machine.
It really isn’t that far fetched. But I think people are misinterpreting the idea of the Singularity. He isn’t really predicting what will happen when we achieve artificial intelligence, he’s just predicting that we’re close to it. You’re right, he may be wrong, but where he isn’t wrong is that the next 50 years will be paradigm changing events based on technologies that ALREADY EXIST, but have yet to find their time to reach a market that wants these items or have just recently been proven and production isn’t ramped up yet, for whatever reason. There’s a lot of it out there.
Remember, focusing light depends on the shape and arrangement of your lenses and how you control the direction of the light source that gets through the lenses. If you have smaller lenses, you can focus light more precisely.
This has already been stated in different forms but wild predictions like Kurzweil’s tend to get stymied by some unknown limitation in the universe. Kelvin (sort of) predicted that soon physics would explain everything in the world; 20+ years later Heisenberg was proving that certain things were unknowable. Hilbert was on his way to grounding mathematics until Godel’s incompleteness theorems proved it was impossible to have consistent mathematics. I can’t help but feel that we’ll hit a similar barrier.
Kurzweil certainly seems technically competent, and sounds very smart, but does that make someone one of the best inventors of our time? I’m not sure who I would put in that category, but the things listed on the wiki page seem like straightforward ideas (as in, many of the tech people I know including myself have done many of those same types of things, before ever hearing about other people doing them).
Seems like a top inventor would be coming up with new ground-breaking ideas. Or, maybe I’m wrong, maybe there are few new ideas, it’s really all about execution.
Actually, nano tech will do nothing about the Moore’s law issue. The issue with Moore’s law is that the circuits on the processors are getting so small that the wavelengths used to etch the circuits are already in the ultraviolet range. Can’t get much smaller than that. At a certain point it becomes impossible to make things smaller.
3-D chips and some other things may push it along for a little while but there is a physical limit that we are going to hit. Claiming that someone, somehow, will find a way to continue the exponential growth once the physical limit is hit is a very large bet. People will find ways to use things more efficiently, but you can’t beat the laws of nature.
Also, just because something has been growing exponentially in the past is by no means proof that it will keep growing that way. Many a fool has lost money on that idea.
Perhaps true. But I took Pat Winston’s AI class at MIT 37 years ago, and back then AI was going to happen in 15 - 20 years. Almost all of the projects being worked on - the block world, generating directions, doing integrals, exist today as products. But there has been almost no progress in anything approaching real AI since then. If progress keeps up at this rate, it will take hundreds of years.
Computing power has grown exponentially, and you can’t design a computer nowadays without lots of computer power. But there is no AI component in EDA tools used in computer design, unless you count fairly simple state space search algorithms. I’m not aware of any tools that even use genetic algorithms, outside of universities, that is.
When I took AI, one of our books was from 1959, which shows you how long this has been going on. I remember reading USA today about the introduction of the 386 - a 16 bit processor was sure to lead to AI real soon now. Don’t hold your breath, people. AI researchers have done a great job developing heuristics to mimic much of the stuff we do, but they’ve done a terrible job tying it together into anything like AI. When the first intelligent machine is created, there will be plenty of tools it can use to do human like things, but those are the easy things.
I’m still betting that the first intelligent computer will be running a human brain simulator.
Well you make a valid point though to be pedantic, a microprocessor is nanotech. But yes, if you read the thread you will see that Wesley Clark answered this question. People seem to be obsessed with Moore’s Law, which is merely an EXAMPLE of the theory in action, and isn’t the only thing holding the theory up.
But the theory isn’t about Moore’s Law, it is about the way technology progresses. I posted above precisely what you need to know regarding this, how he addresses the problem that you and about 50% of the other people in the thread have presented. It’s in the first two chapters of “The Singularity is Near”. He recognizes such hard limits and considers this plateauing point to be a place where some other paradigmatic change occurs. He’s not talking simply about Moore’s Law. He’s talking about the evolution of ALL technology. For instance how Moore’s Law helped decode the Genome faster, and the more of the Genome that was decoded, the more of the genome they could decode in less time, along with computers becoming better and faster. Then when there is a market for the technology that can be produced with these scientific discoveries, more people sink more money and more effort into R&D thus driving the rate of advancement even further.
Sure, fine Moore’s Law is reaching it’s Zenith, but that’s beside the point, it’s merely one part of the theory.
But the point is not that microprocessors will always improve. They will improve other parts of the computer when they reach some kind of hard limit. For instance they put in multiple cores, and start building up the processors in the third dimension. All addressed in the first two chapters of his book. But it’s just about computing power, it’s about how one technology facilitates the advancement of another. Like how Nanotech and Biotech advancements help each other as people use DNA fragments to create protein scaffolds and bind them to other molecules. The more we learn about what different portions of DNA do for different creatures we can insert it into a cell modularly. The DNA can alter the functionality of the machinery of that cell. Once you can build a cell, you can build a cellular matrix, and bind different cell types into a larger organism or alter particular cells by inserting DNA into them to alter the pattern slightly.
Many a fool has lost money predicting the future, Ray Kurzweil has gotten rich doing it. People seem to be hyperfocused on the Moore’s Law thing, but Moore’s Law is merely one example that is widely known.
Warning against the possibilities of Black Swan’s can go either way, either a technology that is a game changer, like the PC, Cell Phones and the World Wide Web might occur and proliferate in a short time frame after having existed for a long time. Cellular phones have been around since the 70s, but we only really started achieving market saturation in the oughts. This is another example of his theory. The growth of the cell phone market saturation has been growing exponentially for 40 years, but only in the late 90s did it begin to achieve explosive growth.
The theory has three aspects to it:
The steady improvement of technology by using the same technology to improve itself. Like the Microprocessor.
The steady improvement of technology by using parallel technologies to improve each other. Like the Microprocessor and Nanotech, or the Microprocessor and the Genome decoding. Like the Genome decoding and Biotech, like Biotech and DNA protein scaffolding.
Market forces, as saturation hits, demand increases more companies crop up, more money is made, as more money is made more money is spent on R&D and thus the particular technology is improved more quickly.
The point is not about any one technology improving ad infinitum it’s about technologies that cause paradigm shifts and the overall rate of advancement of all technology. When Moore’s Law reaches its plateau we will advance in other ways. But computers will continue to get faster. He talks about the chipset problem early on in the book talking about certain materials processes that were at the time he wrote the book still experimental. I don’t know what the time between R&D and the market is on microprocessors.
But really, we already have. Could any one human grasp or perform, say, science as a whole? Every time a bunch of people form a team or organization to handle a problem beyond the capacity of any single mind, they are creating a “greater intelligence”. Sloppy, clumsy, and poorly integrated, but that’s what they are really doing. Rather as if you had to make a brain out of clumps of brain cells with only slow, limited connections between each clump; you’d get something more powerful than the individual clumps, but nowhere near as good as a human style brain.
I really don’t see any reason to think that our own personal intelligence is even near whatever limits actually exist. Any more than we are the strongest or fastest or toughest things that can exist.
I’d go further and say that we don’t even really tap very well into our intelligence as it exists because of poor pedagogy. Every single one of us is far more intelligent than we operate due to loss and inefficiency as a result of poor training.
I don’t know if this is still true, but in the 1990s and early 2000s people were applying for and getting patents for the most patently obvious stuff. A guy I worked with (at WorldCom!) got a patent on the process of replacing a laptop. It describes such innovations as the order in which to do things (first back-up the users documents onto a zip drive, THEN wipe the hard drive clean).
Another guy got a patent on how to handle help desk calls. You know, first have the help desk guys trained to handle the simplest issues, and recognize complicated issues to escalate to second level support, then third level and so on.
We had a whole staff of in house IP lawyers to file this crap. And yes, it was so that we could say we have x,000 patents. By that time we had stopped actually innovating in telecom. Actually WorldCom never did. They bought companies who were innovative, but the innovation would stop soon after the acquisition.
In case it hasn’t been mentioned; these days the US Patent Office is funded entirely out of patent fees. So naturally, it’s pretty easy to get a patent for something.
That assertion seems pretty much where most people see difficulties – will we? I’ve briefly mentioned above that there are hard limits to computation – you can’t store more information in a volume of space than the Bekenstein bound allows you to, you can’t flip bits faster than quantum mechanics permits, you can’t transfer information faster than light speed. The thing is, up until these bounds are hit, it’s in principle possible to make Kurzweil’s arguments in exactly analogous form, though it’s clear that they will fail. Thus, his arguments don’t have any real predictive value – if they fail then, they may fail now, just like arguments implying that we ought to move faster than light speed and have colonised the solar system by now did.
And besides, his arguments, if they hold, could be made at an earlier time for any non-human species, who have gone through the bulk of evolutionary history exactly the same way we did, going through some billion years of revolutions alongside our particular branch of the tree of life. Yet, they apparently failed to continue their course – they never walked upright, invented language or the printing press, as have some isolated populations of humans. So apparently, no matter how long your curve has followed the exponential growth, there’s no guarantee that it’ll continue to do so; in fact, almost all lines so far safe our own have petered out eventually.
Yes, on our computers in 1981. It wasn’t hard. You feed numbers to a digital to analog converter to produce sound. Which numbers? Depends on what you want it to sound like. I could make more notes, sounds and sound effects with that than with the actual synthesizer that we owned. It could even produce human speech (which sounded like that crappy computer talk, but hey it worked).
My point wasn’t to take away from Kurzweil’s execution of his projects. And, as I said in my previous post, he appears to be smart and technically competent. But when you say top inventor, to me that says novel ideas. Maybe you and I have different ideas of what would qualify for a top inventor.
Either way, I do agree he is not just some guy off the street that has no knowledge of these topics, which appears to be the point you are trying to make.
