Slashdot Mirror
Alva Noe: Don't Worry About the Singularity, We Can't Even Copy an Amoeba
An anonymous reader writes "Writer and professor of philosophy at the University of California, Berkeley Alva Noe isn't worried that we will soon be under the rule of shiny metal overlords. He says that currently we can't produce "machines that exhibit the agency and awareness of an amoeba." He writes at NPR: "One reason I'm not worried about the possibility that we will soon make machines that are smarter than us, is that we haven't managed to make machines until now that are smart at all. Artificial intelligence isn't synthetic intelligence: It's pseudo-intelligence. This really ought to be obvious. Clocks may keep time, but they don't know what time it is. And strictly speaking, it is we who use them to tell time. But the same is true of Watson, the IBM supercomputer that supposedly played Jeopardy! and dominated the human competition. Watson answered no questions. It participated in no competition. It didn't do anything. All the doing was on our side. We played Jeopordy! with Watson. We used 'it' the way we use clocks.""
I did philosophy myself as an undergraduate, so I don't want to bash our armchair friend here for doing his best. He is making the classic mistake of making claims about fields he isn't part of. In this case biology, computer science, and cognitive science in general (beyond philosophy).
Regarding the statement "We used 'it' the way we use clocks":
He is mistaking agency for being something that is an end unto itself. This isn't true. Agents commonly use other agents as tools. The mere property of "being used" doesn't dictate whether something is sentient, intelligent, an agent, or whatever. Yeah, we used Watson to play Jeapordy!, but that doesn't mean it isn't smart. Watson is actually way "smarter" than any human in certain ways.
This boils down to what you define as intelligence. In humans, intelligence is a very rough term applied to an enormous pile of features. Processing speed, memory, learning algorithms, response time, and many more features all contribute to what we think of as intelligence. A singularity doesn't need to precisely mirror the way in which a human thinks in order to be a singularity. It just needs to be able to adapt and evolve. I'll be the first to admit we are a long way off from modeling a human consciousness in virtual space. However, existing machine learning and rule based techniques are powerful enough to do some really impressive things (like Watson and Siri). They aren't singularity level, no, but that doesn't make this man's arguments relevant.
Regarding "we can't produce "...machines that exhibit the agency and awareness of an amoeba":
The idea that an ameoba displays intelligence in excess of our current ability to simulate is frankly a little ridiculous. Artificial agents are capable of very complex behavior. They can react to abstract constructs which are inferred about their environment. They can anticipate opponents based on statistical probability and thereby win, on average, more often than even *a human being*. An amoeba is closer in behavioral complexity to a simple chemical reaction than it is a contemporary artificial intelligence.
I think the singularity is just some technology-wrapped new age bullshit.
The fact that you had to resort to fiction (and cheesy fiction at that) for your reference, instead of an applicable real life analogy, just shows that up.
You can make a machine that is many orders of magnitude more intelligent than a human, but unless it has the mechanisms to want something, it won't want anything. Think about what it would take to program a conscious being, just think about how we humans are aware of time and the movement of time--as a software engineer it boggles my mind trying to think about how the brain accomplishes that. Then to program a machine (biological or not) to want something in a fairly consistent way over a period of time under changing circumstances...it takes more than just brute force processing time to accomplish that. We biological machines are aware of ourselves, and we have no idea how we accomplish that. We are going to have to figure that out before we make the singularity machine.
The Moore-Murphy Law: The number of things that will go wrong will double every 2 years.
Consciousness neither implies nor is implied by either emotions or desires.
Contribute to civilization: ari.aynrand.org/donate
More than half the time, the "answer" doesn't even make sense as a response to the "question".
Q: Who is "Joe DeSixpack"
A: Born in 19th Century Verona, he died of plague playing beach volleyball in Aruba.
NO!!!
Skynet begins to learn at a geometric rate. At 1:35 A.M. Eastern time it runs out of disk space and crashes horribly
Yes! That's precisely the technology-wrapped new age bullshit we're talking about.
How can I believe you when you tell me what I don't want to hear?
You're applying your own arbitrary definition of intelligence, using that to frame the argument and declare yourself right. AI as a subject got past that kind of crap by the 90s with a realisation that it's a little more complex than all that.
You dismiss various AI solutions as just being a bunch of algorithms, well guess what? we still have absolutely no evidence that human beings themselves aren't run by a bunch of algorithms, the only difference is that we just don't understand them well enough to document or reproduce them synthetically yet.
It's stupid to declare something like a neural network or expert system not intelligent just because you understand the details of it because most people who see the actions of a neural network would say "That's pretty intelligent how it can do that".
If you have two closed boxes for classifying say, wine. In one is a person and in the other is a computer with equipment and a neural network trained for wine classification and a wine sample goes in and a classification is displayed on a screen by entering an output and the computer can of course do it better than any person people are pretty much always going to class the computers response as the intelligent one when it gets far more tests right. The Turing test was designed to show the sorts of intelligence we see in strong AI, but modified versions of that display intelligence from weak AI in select circumstances.
So yes, you can absolutely say things like neural nets, and expert systems are not strong AI, but you absolutely cannot say they are not intelligent without framing it on the rather stupid definition that something is not intelligent if we understand how it works. In some circumstances these systems would be deemed to be more intelligent than humans by most people and as we don't have a fixed definition of intelligence that seems a far better way of judging intelligence - getting people's judgement on intelligence in a statistically sound study than coming up with definitions like "Something is intelligent if we don't understand how it works".
If computer algorithms could show no intelligence whatsoever then we'd only be using them to do dumb repetition, like building cars on an assembly line, but we don't, we use them to augment our search capabilities, to correct our grammar and spelling, to figure out an optimal path for data to travel down on a complex network and so on and so forth far better than a human could - we're using them to augment our intelligence every day and many ways, and that's because they can display some intelligence. Not conciousness, not strong AI, but a degree of intelligence all the same.
You're conflating conciousness, intelligence, and strong AI all into one big pot, but it's all far more nuanced than that. You're assuming life works in a binary way, where something is either not intelligent, or something has human level intelligence and artificial would be a strong AI. But god only knows, we have enough evidence of various living things in this world to see that there are varying gradients of conciousness and intelligence for that to be true. Assuming it'd somehow be different with computers makes no sense and guess what? in the last 20 years we've seen progress with AI research with ever increasing levels of intelligence. When that'll escalate to the level of what we deem strong AI, or human intelligence is anyone's guess but we're not suddenly going to go from having no strong AI to having strong AI, we're going to have ever increasingly intelligent stuff that approaches strong AI and eventually becomes good enough to declare as strong AI.