Nope, I just said that you are able to turn on your microwave because you want to. The microwave then runs through a list of instructions. The two actions are different, however you wish to label them.
Straight Dope 3/3/23: Followup - What are the chances artificial intelligence will destroy humanity?
The two actions are both computation.
Not likely. The microwave has a known list of tasks it can it can perform. Everything it does goes through the adder. There is no adder in your brain.
It is possible to come up with definitions that many may accept, but that deny or trivialize many attributes, including those fundamental to basic humanity. Most doctors consider themselves scientists and are very good at evaluating evidence. “The brain is a computer because it can do arithmetic, use forms of logic, and use heuristics and rules of thumb” may be true but shows the dangers of making complex things too simple, rather than as simple as possible. Reduced to such oversimplifications, the characterization becomes much less useful.
Many participants in this thread are completely ignorant of the basic concepts in computer science and AI that we’re trying to discuss. Yet Discourse tells me:
2 people have followed the link to the Science & Speculation paper in the OP (one of them was me).
1 person has followed the link to the Michael Cohen paper that was central to the OP (that was me).
1 person has followed the link to the Wikipedia article that I provided early in the thread that summarizes the field of AI alignment that many people expressing strong views in this thread have never even heard of.
1 person has followed the link to the Wikipedia article on AI safety, ditto.
This is very weak, and I’m done with listening to the hot takes of people who apparently have no interest in fighting their own ignorance. Much of this conversation is frankly at about the same level as the Gish Gallop of religious creationists who are clueless about the most basic concepts of evolutionary theory. SDMB is supposed to be better than this.
What’s going on here is an attempt to make brains be computers. But brains are electrochemical analog control systems and what we call computers are stored program digital processors. There is no equivalency.
I thought my definition was better than that. Nothing Creationist about adders.
It is certainly possible to reduce and accurately describe people as “43 kg of oxygen, 16 kg of carbon, 7 kg of hydrogen, 1.8 kg of nitrogen, 1kg of calcium, and a rag and a bone and a hank of hair. But the reverse is not true. It is possible to be more than the sum of the parts. One suspects megasupercorporations would prefer to value people at a buck-o-five each. But they are unique and much more valuable.
As for not understanding your view, did I not print the article which summarizes it? If not, what part of that article do you disagree with?
Every neuron in the brain has a potential at which it fires. It is connected to other neurons through synapses. When potentials from enough synapses together reach the activation threshold of the neuron, it fires.
This process is called ‘addition’. Every neuron in your brain is an adder.
Put enough together in the right order, and it’s multiplication. Or any other arbitrary sequential calculation.
In artificial neural nets, the potentials of synapses are represented as weights in the parameters. They function in exactly the same way.
Sure, there are sixteen or so logic gates to mimic NANDs and NORs. No one disputes the brain can do computation. This is different from the more trivializing The brain is just a computer. I’ll believe that when the brain is understood sufficiently well enough to build one that is indistinguishable in almost every aspect.
Except for rare specialties computers do not contain hardware neural nets. ‘Neural nets’ are software constructs that reside as lists of instructions in the computer memory. The neural net never exists as a physical entity like groups of neurons in a brain. So there is no place in the computer where judgement or understanding or other brain like activity might reside. With the possible exception of the adder since everything passes through the adder.
There is no comparison between neural nets and neurons. They are two entirely different things. The output of neurons is an electrochemical wave that is interpreted by other neurons. The output of a software neural net is a binary value that is ultimately interpreted by a human.
If a computer is other than an adder, microcode and data management then there may be a place where something brain like could reside. If you believe that to be the case then we should explore the possibility. What definition of a computer would allow such a thing? Or maybe I misunderstand your point.
Where, exactly, in the brain does judgement or understanding reside?
You are the only one suggesting that there is anything trivial about it.
A perfect understanding of what the brain is doing is not required for a fundamental understanding of the kind of thing that it is doing. I would ask
(a) Do you offer any hypothesis for what kind of thing you think the brain might be doing that is not computation?
(b) Do you have equal skepticism about the fact that the brain is a product of evolution? Because such skepticism is a similar logical consequence to your reasoning here: we do not perfectly understand the way the brain works, therefore we cannot conclude that [ it is the product of evolution / it is a computer ], even though we offer no alternative hypothesis (let alone evidence) about [ how it came to be / what it is doing ].
Do you understand Turing’s “point”? You have given no indication that you’re familiar with it, yet that’s what you appear to be so vehemently disputing.
Is the brain digital? Can neurons, which after an action potential may freeze or speed up or slow down, really transmit strings of information? Does the brain incorporate quantum mechanical methods? Or do we not know? What makes us wake or woke? Stuporous or alert?
The reticular activating system is thought to be the seat of consciousness. What algorithms cause that? If there are behaviours we cannot explain and do not really understand, and there are certainly behaviours unlike what we would expect from a computer, this implies that though computation may play some role there is more going on. The fact some elements of neuroscience are akin to input-output or feedback control systems, and that the brain clearly processes information, does not mean much more is going on besides computation. Calling the brain a “supercomputer” hardly explains that difference. I can accept some equivalence between chemical and electrical signals, to a point, but electrochemistry hardly explains the whole complexity of chemistry or biochemistry, merely some aspects. It is a big ask to assume all emotions and all behaviours follow processes in the same manner as computers.
Shannon demonstrated that a universal Turing Machine could be represented digitally by a single two input adder with a single data output - a one bit serial computer. All computations that can be achieved by an adder can be accomplished with a one bit serial Turing machine given enough time. Which means that all digital computers no matter how large or complex are at heart an adder. They are all just adding machines.
So, the Turing Machine description implies that if brains are computational then a Turing Machine can do everything that a brain does. The problem is that brains are not computational. I have ample evidence to determine that to my satisfaction. But for an authoritative answer you’d have to talk to HMHW or others at their level of academia.
Yes, I understand Turings’ point very well. It does not apply to brains because brains are not computational.
That is a question for neurologists not a computer geek. I have read that researchers have identified areas of the brain where various activities take place.
Care to share some of this ample evidence?
Because all I’ve seen you do is to adopt a restricted definition of computation that does not correspond to the definition that is relevant to what a universal Turing machine can simulate.
Again, you seem to be treating this as though this is a purely semantic debate about the “best” definition of computer.
But the basis of this discussion is the specific question of what a machine intelligence that is a computer (a universal Turing machine) can theoretically simulate, since that speaks to whether AGI is possible. It can simulate all other computation.
So the definition that matters here is what is included in that claim. The notion that “everything is computation” is not just a weak attempt to justify the claim that a brain is a computer via semantic trickery. It is a strong statement that there is an uber-broad definition of computation that is included in what a universal Turing machine can simulate.
Nothing you have mentioned falls outside that definition. It is not restricted to binary computation, it is not restricted to electronic substrates. It includes everything that neurons do, all biochemical processes.
What about Penrose’s contention that Turing complete computers nonetheless cannot deal with non-computable problems like the Halting problem? Another person making similar claims is Robert J. Marks II in his book “Non-Computable You: What You Do That Artificial Intelligence Never Will”