The Semantics Differentiation of Minds and Machines

John David Funge writes "In Dr David Ellerman's book Intellectual Trespassing as a Way of Life there are a number of interesting essays. But there is one particular essay, entitled "The Semantics Differentiation of Minds and Machines," that caught my attention and which should be of interest to Slashdot readers. In that essay Dr Ellerman claims that "after several decades of debate, a definitive differentiation between minds and machines seems to be emerging into view." In particular, Dr Ellerman argues that the distinction between minds and machines is that while machines (i.e., computers) make excellent symbol manipulation devices, only minds have the additional capacity to ascribe semantics to symbols." Read the rest of John's review. Intellectual Trespassing as a Way of Life author David P. Ellerman pages 290 pages publisher Rowman & Littlefield Publishers, Inc. rating 7 reviewer John David Funge ISBN 0847679322 summary Dramatic changes or revolutions in a field of science are often made by outsiders or "trespassers".

However, Dr Ellerman's argument appears circular. In particular, Dr Ellerman seems to have decided that, by definition, the only possible semantic interpretation for any collection of wires, capacitors, transistors, etc. that we would commonly refer to as a "computer" is as nothing more than a symbol manipulation device. While a computer is indeed (at the very least) a symbol manipulation device, what is there to prevent another mind ascribing additional semantic interpretations to the collection of wires, capacitors, transistors, etc. that we commonly refer to as a "computer"? In particular, what if my mind were willing to make the semantic interpretation that a computer is a device that can both manipulate symbols and can also ascribe semantics to symbols.

Moreover, what if I one day met a collection of blood vessels, skin, bones, etc. called Dr Ellerman? What would prevent me from ascribing to him the semantic interpretation that he is nothing more than a symbolic manipulation device? After all, Dr Ellerman concedes that their may be no way of distinguishing minds from machines purely on the basis of behavior. That is he specifically acknowledges that computers may one day pass the Turing test. So why would my mind not then be able to legitimately ascribe any semantic interpretation (that fits the observed behavior) I see fit to either humans or machines?

It seems that Dr Ellerman's essay considers two different types of physical devices that are potentially indistinguishable on the basis of behavior. Then arbitrarily defines one type of device (computers) to correspond to nothing more than symbolic manipulation and the other (human brains) to have the additional ability to ascribe semantics. Upon adopting these two axioms, he is then (somewhat unsurprisingly) able to conclude there is a distinction! But the distinction simply arises from the fact that he has arbitrarily defined a distinction in the first place.

In another essay in the collection, entitled "Trespassing against the Happy Consciousness of Orthodox Economics," Dr Ellerman argues that modern Western societies are not as free from slavery as orthodox economics would have us believe. In particular, he concludes that work in non-democratic firms is nothing less than a form of "temporary voluntary slavery". It would be ironic therefore if his essay on minds and machines were one day used to justify the slavery of (non-human) machines. Indeed, Dr Ellerman's characterization of the supposed intrinsic differences between humans and machines is sadly reminiscent of the despicable and unscientific arguments about intrinsic racial differences that were once used to justify human slavery."
You can purchase Intellectual Trespassing as a Way of Life from bn.com. Slashdot welcomes readers' book reviews -- to see your own review here, read the book review guidelines, then visit the submission page.

Comments

semiotics

[gEvil (beta)]

I believe the proper term for this field is semiotics, the study of the assignation of meaning to symbols and signs.

Just like Organic vs. Inorganic chem.

[Vengeance]

While there was long believed to be some sort of mystical special quality to organic molecules, eventually we figured out that chemistry is chemistry, and that simply by using Carbon we get interesting possiblities.

I (so far) have not seen any reason to suppose that the difference between 'thought' and 'computing' is any different. Incorporate enough complexity in the right sort of organizational framework, and the two should be interchangable.

Machine Learning of Semantic Relations

[Baldrson]

Peter Turney's Learning Analogies and Semantic Relations falsifies the Ellerman's assertion that semantics is out of the reach of engineering. Turney's more recent Human-Level Performance on Word Analogy Questions by Latent Relational Analysis (Warning: PDF) shows an engine performing about as well as college-bound seniors taking the SAT verbal analogies test.

For a review of Peter Turney's group's accomplishment see "AI Breakthrough or the Mismeasure of Machine?"

False presumption

[blair1q]

Semantics are associations between symbols.

So whatever this guy is on about, he's got it wrong.

Computers are perfectly capable of making fuzzy inferences from loose associations.

With a greater understanding of real connections, they will be better able to weed out the fuzzy associations and strengthen the remaining ones.

This is how intellectual learning works.

And there's no reason a computer can't simulate it better than a human can.

Review seems poorly written

[nexarias]

I don't think the reviewer has demonstrated adequate mastery of the subject (artificial intelligence) and its present studies. For example, the problem of assigning meaning to symbols is a BIG one, and the defining of computers as symbol manipulators is NOT arbitrary. This problem first arose when Thomas Hobbes talked of the mind as a symbol manipulator and Descartes rubbished his argument, pointing out the problem of Original Meaning (how symbols come to indicate this or that in the first place).

Computers as symbol manipulators is also an idea that arose from John Searle's "Chinese Room argument". Perhaps one of the best contemporary discussions is by John Haugeland in his book "Ariticial Intelligence: The Very Idea".

Overall, a seemingly immature review of the book. Disappointing.

Symbolic vs semantic

[chriss]

Basically: a symbol is a variable and can hold any value. If a system knows that Dolly is a sheep and that sheeps are animals and that animals eat, it can guess that Dolly eats. But it cannot tell if Dolly is a plane, unless someone somewhere made that relation (planes are machines, machines are not living beings, animals are living beings, so Dolly can't be a plane). They would need an unlimited amount of rules.

A human "knows" about the meaning (semantic) of the symbol "sheep". Although this has never been discussed, he could answer that a sheep will not stand still if set on fire. The question is how the human is able to tell this. He does not need a sharp line of arguments.

But maybe he simply uses an enormous amount of small rules that seem to form something more complex called semantic in the sense of the article. The OpenCyc project assumes this and tries to teach a machine millions of small rules (assertions and concepts) to create sort of common sense based on a real world view (requiring to "know" about the world) in software.

The real difference between humans and machines

[ENOENT]

When a human makes a mistake, it immediately pours massive processing power into either formulating arguments about why it's not a mistake or finding someone else to blame for it.

Machines tend not to do this.

"Only humans can..."? Can even humans?

[Impy the Impiuos Imp]

The AI community has suggested that what humans believe is some kind of "deep understanding" is nothing of the sort. We have just learned to push symbols around, too.

Consider the "deep understanding" of simple mathematics. But is your instant recall of 6 x 8 (assuming you can) anything deep, or just memorized, along with the symbol pushing to mechanically figure out tougher problems?

The problem lies in tying up a "symbol" in the mind (which may be more than literally a string of characters. However, it is an object) and something "out there". That's the tough issue, not the symbol pushing itself, necessarily.

Pot, meet Kettle

[Artifakt]

"After all, Dr Ellerman concedes that their may be no way of distinguishing minds from machines purely on the basis of behavior."
"It seems that Dr Ellerman's essay considers two different types of
physical devices that are potentially indistinguishable on the basis
of behavior. "

It seems that the reviewer considers both mind and brain to both be purely physical things, and indeed synonyms - Physical devices that are thus potentially indistinguishable on the basis of behavior. Upon adopting this axiom, he is then (somewhat unsurprisingly) able to conclude there is no distinction! But the lack of a distinction simply arises from the fact that he has arbitrarily defined amind and brain into a single category in the first place.

Review translated: Trust me, I don't have any underlieing assumptions like he does, so I'm right and he's wrong, PH33R MY L33T PH1L0S0PHY SKILZ!

11 year old book of crap reviewed here? Why?

[Dr. Spork]

Sorry, I teach philosophy in college and I read student essays like this every semester. This one seems reasonably insightful, probably B+ (though I haven't read the book myself, so I can't say whether it misrepresents the position).

But what's really on my mind is this: Read the table of contents - this book could not possibly be anything but crap. I mean, what sense does it make to have one chapter called "Chapter 3: The Libertarian Case for Slavery" and once you're done with musings on economic theory, you toss off a Chapter 7 where you casually present your solution to the question about the difference between minds and machines? How promising is that? Not very. So while the review author may have torn this chapter a new orifice (and the thesis surely has many other problems to boot), I must say that I do not toast his choice of reading. This is crap that was ignored in 1995, and just because it's a $2.95 special at the used book store doesn't mean we need to hear the following on Slashdot:

Newsflash: Some crank wrote a stupid book 11 years ago and I found there is a problem with one of the chapters!!!!! Read on!!!!!

I'd have more sympathy if the text were available online so we could RTFA and have a substantive discussion, but in the absence of that, our only option is to flame the responsible.

This is religion, not science

[dmoen]

The "intelligent design" crowd is a group of people who, for religious reasons, refuse to believe that human beings and animals belong to the same category. Since it's inconceivable that humans evolved from non-human animals, the theory of evolution must be overthrown, and another theory erected in its place.

There is a similar thing going on with people who study how the human mind works. Some people, for religious reasons, refuse to believe that human beings and machines belong to the same category. Humans have souls, and machines do not. Therefore, a computer can never be programmed to have all the qualities of the human mind. It's harder to see this as a religious issue, since some of the people who hold this position are atheists who claim not to believe in souls or the supernatural. But what makes this a religious issue is that there is no amount of scientific evidence that can ever convince these people otherwise.

Anyway, the two camps have been arguing about this forever. It's impossible for a member of one camp to "convert" a member of the opposite camp using rational argument. So they resort to insults. People in the "strong ai" camp accuse the other camp of being Cartesian dualists, or believing in a supernatural soul. People in the "dualist" or "mysterian" camp accuse the strong ai folks of denying the existence of human consciousness and self awareness. According to the dualists, strong ai folk believe that humans are just machines, so humans can't be conscious in any real sense, don't have free will, and can't be morally responsible for their own actions. Some (stupid) strong ai folks even agree with these insults directed against them, which makes the debate more complicated, and more infuriating. The issue of moral responsibility, which is always bubbling under the surface of these debates, shows how this is really a religious issue at a deeper level.

For the record, I am a strong ai person who believes that human beings are deterministic machines who have consciousness, free will, and moral responsibility.

If you would like to read some good books that back up my position, see:
- How the brain works, by Pinker
- Freedom evolves, by Dennett

Doug Moen

All in how you look at it....

[NiteShaed]

Talking about machine intelligence is tricky in that we generally only consider *human* intellegence (which makes sense considering that's what we are). In John Varley's "Steel Beach", he suggested The Invaders (a mysterious species of aliens) might not consider humans an intelligent species, but looked at us as just another engineering species like bees, meaning intelligence is really dependant on your point of view. What we're really talking about when most people say Artificial Intelligence is actually more an issue of Artificial Humanity.

Re: is the brain a digital computer?

[johnrpenner]

this point has been made before,
by cognitive scientist john searle in his paper:

    is the brain a digital computer?

in the summary, searle puts it this way:

--| Summary of the Argument |---

This brief argument has a simple logical structure and I will lay it out:

On the standard textbook definition, computation is defined syntactically in terms of symbol manipulation.

But syntax and symbols are not defined in terms of physics. Though symbol tokens are always physical tokens, "symbol" and "same symbol" are not defined in terms of physical features. Syntax, in short, is not intrinsic to physics.

This has the consequence that computation is not discovered in the physics, it is assigned to it. Certain physical phenomena are assigned or used or programmed or interpreted syntactically. Syntax and symbols are observer relative.

It follows that you could not discover that the brain or anything else was intrinsically a digital computer, although you could assign a computational interpretation to it as you could to anything else. The point is not that the claim "The brain is a digital computer" is false. Rather it does not get up to the level of falsehood. It does not have a clear sense. You will have misunderstood my account if you think that I am arguing that it is simply false that the brain is a digital computer. The question "Is the brain a digital computer?" is as ill defined as the questions "Is it an abacus?", "Is it a book?", or "Is it a set of symbols?", "Is it a set of mathematical formulae?"

Some physical systems facilitate the computational use much better than others. That is why we build, program, and use them. In such cases we are the homunculus in the system interpreting the physics in both syntactical and semantic terms.

But the causal explanations we then give do not cite causal properties different from the physics of the implementation and the intentionality of the homunculus.

The standard, though tacit, way out of this is to commit the homunculus fallacy. The humunculus fallacy is endemic to computational models of cognition and cannot be removed by the standard recursive decomposition arguments. They are addressed to a different question.

We cannot avoid the foregoing results by supposing that the brain is doing "information processing". The brain, as far as its intrinsic operations are concerned, does no information processing. It is a specific biological organ and its specific neurobiological processes cause specific forms of intentionality. In the brain, intrinsically, there are neurobiological processes and sometimes they cause consciousness. But that is the end of the story.\**

--

regards,
j.

One of many examples.

[jd]

Humans are excellent at differentiating between things that are really the same, or inventing totally new layers of reality because of flawed assumptions about the way the world works. Today, I think we've gone beyond needing to think of fire, earth, air and water as being the four elements from which all physical matter is constructed, and light does not need an aether to "travel through".

For that reason, any attempt to differentiate the mind and computers by using comparisons that aren't really meaningful or applicable should be thrown out. Maybe computer-based intelligence will never exist, but if that is the case, it won't be for the reasons we're being given.

For example, looking at the high-level functionality of the brain and comparing it with the transistors of a computer is an absolute give-away that the author isn't going to let the facts get in the way of a good story. The low-level mechanics of the brain (the chemical and electrical signalling) can be reasonably compared to the low-level mechanics of a computer, because it is valid to compare like with like. For the same reason, it would be fair to compare the Operating System of a computer to the ancient "reptilian" core of the brain. Both are designed for housekeeping operations and are used by the higher levels to mask implementation details. And so on up through the layers.

It should also be kept in mind that the human brain is capable of almost ten times the throughput of a top-of-the-line supercomputer. Given that one of the limiting factors of parallel architectures is the interconnect, it does prove that our networking technology is still extremely primitive. This is important, because it is going to be hard to build a machine that can "think" like a human if we have the "neural" interconnects of a Diplodocus.

At the current rate of technological progress, I do not believe we will have a computer powerful enough to model the human brain until 2015 or 2020. Even then, it'll be a Government-owned supercomputer likely used for weapon simulation. We won't see Strong AI researchers get hold of such machines until maybe 2060 and (if the usual development patterns hold) nobody will have any idea how to turn the raw computing power into something useful until 2100 at the earliest.

So, really, the earliest we could possibly really know (for certain) that the mind is (or isn't) like a machine is 2100. Anything stated with certainty before then is pure ego-stroking for the group of supporters attached with one camp or the other. Doubly so when it is provably and obviously intended to be deceptive.

The only problem I see with debating the matter from an intellectually honest standpoint until then is that current global warming models put most of the planet under water or under rainforest by 2100, which means that we might never really know the results of the research anyway.

It's the "soul" fallacy

[pieterh]

"Humans are different from X because they can do Y", where X is variously "animals", "machines", and Y is variously, "make tools", "use language", "play chess", "murder", or whatever.

It's a silly exercise because there is nothing specific about humans except their ability to interbreed with other humans. That is all that technically defines us as a species, and even that definition is fuzzy, ignoring people who are sterile, too old or young to breed, or who never leave their keyboards long enough to look for a mate.

When it comes to the mind, emerging consensus is that it consists of a large number of well-designed tools, not some fuzzy blob of software. Most likely, each of these mental tools can be perfectly implemented as software. There are simply a huge number, and some are very, very subtle.

We will, eventually, be able to simulate the whole human mind in software, in the same way as we'll eventually be able to construct robotic bodies that work as well as human bodies, by re-implementing the structures that our genes build, one by one. The best way to construct a robotic hand that works like a human hand is to reimplement a human hand. The best way to construct a robotic mind that works like a human mind is to reimplement a human mind. This is perhaps self-evident but it's not always been accepted.

As for the arbitrary distinctions, this is just a belief system, an attempt to create a soul, however you phrase it.