The Paper Test Shows Why AGI Is Not Possible: And Why The Brain Is Different In Kind

We previously, and briefly, discussed The Paper Test (see the articles on limitations at bottom here), which is an argument that computers, including those laden with AI, are not alive, not thinking, not conscious, not intelligent and do not have minds. Thus there will be no time in which computers reach Artificial General Intelligence, or AGI. I posed a variant of this argument on Twitter. The responses were interesting, and one of them deserves to be answered.

THE PAPER TEST

1. A computer is a collection of states, 1s and 0s, laid out in known planned order. At a set time of the clock, get a big piece of paper and write them all down, in the same ordered way. The paper will thus be an exact replication of the computer’s internal state.
2. A computer’s state changes at clock ticks. It does this based on programmed rules, given and constrained by the computer’s architecture. Move the clock forward one tick. On a new piece of paper, placed over the old one, write down the new state of the computer.
3. Do this again for the next clock tick. Then again and again, etc.
4. At the end you have on paper the same sequence that is on the computer, one after the other. Now flip those pages, like you did in notebooks as a kid, creating animations. There is no difference between the paper and computer in content.
5. If the flipped paper alive? Intelligent? Does it have a will? Has it reached true general intelligence? Is this a mind? Is it conscious? If so, where is it? In the flipping? In the paper? The ink? Just where?
6. Obviously none of this in intelligence. But the computer is just the same. Exactly the same, but instead of you flipping pages, electricity does it. Which changes nothing.
7. AI will not reach AGI.
8. Thus, there is no reason for panic or even “concern”.

Years ago we did the same experiment with an abacus, imagining the changing state of the wooden beads was powered by sluices of water. The inference is the same.

My own guess is that those who think computers can come alive and possess intelligence is because of the electricity, which to many is mysterious. The men who designed the computer hardware, its electronics, would find the idea hilarious, because they know there is nothing in a changing voltage level that thinks.

A lot of coders have only a passing familiarity with electronics. Sure, they know how to use electronics better than most. But knowing its innermost secrets, like how the equivalent series resistance of a capacitor changes with frequency and the like, is not something they spend much time thinking about.

Which means they can allow, in their minds, electricity to do strange and wonderful things, like produce this occult power called “emergence”, which is the unknown force that somehow, nobody knows how, transforms those voltage levels into life. Of course, all the thinking behind those changing the voltage levels (which are those 1s and 0s) took place outside the computer, in the minds of men.

That is all the proof we need to show computers, and code, do not possess consciousness or life. And will not, no matter how many transistors are added. One or ten mega-trillion, it’s all the same. Waving your hands around and mumbling about “emergence” is proof of nothing.

Enter the objection many made (like this common example):

I hesitate to recognize this as even a midwit argument. It is depressingly void. Your brain consists of atoms. Is an atom conscious? Now list all these atoms on one piece of paper per nanosecond. You can’t? Perhaps because there are too many?

(Anger, incidentally, and insults seemed to be almost the default retort. As I’ve said many times, people cherish their fears and cannot bear to have them removed.)

It’s not that there are too many “atoms” to write down which is the problem, because size doesn’t matter to us. Neither does speed. Making discrete calculations faster does not create life. A transistor, or a whole bank of them, suitably hooked to an oscillator, which turn on and off faster and faster does not reach sentience after 32 GHz, or whatever. Speed is relative. Lack of paper is also not a problem. We can imagine our paper is provided by a government agency. The problem with “brain” as machine is that it is false. And it is false because of several reasons.

One of which is the brain is not all of us, but only part of us. We are wholes, substances. It’s the whole body acting at once, as a complete unit, that does its thinking. But ignore that and say it’s all brain. Close enough.

Computers have discrete states that change at ticks of a clock. Brains do not. Brains (we are supposing for the sake of argument that it is only brains that count in thinking) do not have discrete states. Further, there is no clock tick. Changes happen along continuums of space and time. All of the matter in the brain is combined states of potentiality and actuality, only some of it is a lot more potentiality than actuality. And these states cannot be written down discretely. (Review this article about substance if these terms are unfamiliar.)

It would be like trying to write down the position of an electron in quantum mechanics. Sure, you can jot down a formula which takes certain conditions and then gives a probabilistic prediction of where the electron interaction happens (with whatever thing is being interacted with), but that’s not even close to writing down its exact discrete location and value. And this equation says nothing about the causes of the ultimate location, or indeed of the causes of the exact mix of potentiality and actuality of the pre-interacting electron.

Whereas in a computer we know all the causes of the changes of discrete states, even if we don’t know all the causes in the underlying electronics. And don’t need to, because we know them well enough to engineer circuits that do our bidding. It is those discrete states that count, the on and off of the transistors, and not the strata in which they reside that count, that’s why the paper or abacus also count as computers.

The brain has a lot more than just one electron. The state of the brain cannot be written down. If this is not obvious, read on.

Here is a decent cartoon (source) of a vault protein “docking” with a microtubule.

Quoting Wokepedia, “The vault is a large cytoplasmic ribonucleoprotein, a non-membrane-bound organelle in most eukaryotic cells whose function is not yet fully understood….Most human cells have around 10,000 vaults…”

These are in brain cells, too.

Some say microtubules, of which the brain has plenty, are information carriers. Quoting one well known paper:

Microtubules are major architectural elements without which the neuron could not achieve or maintain its exaggerated shape. In addition to serving as structural elements, microtubules are railways along which molecular motor proteins convey cargo. Microtubule arrays in axons, dendrites, growth cones, and migratory neurons are tightly organized with respect to the intrinsic polarity of the microtubule, which is relevant to both its assembly and transport properties. 

You cannot write down everything that is happening in this small picture, even with all the paper in the world. And not because of the size, but because there aren’t many discrete states to write down. This stuff is not sitting still. Some of it is not yet fully actual, like the electron, and exist in quantum states (which are more potentiality than actuality). Writing the state down is impossible.

If you don’t believe that yet, here’s a picture (source) of actin filaments, or microfilaments, that are responsible for the continuous movement in the shape of cells (like in the brain):

Lastly, here’s a new picture of one square millimeter of the brain (lo res; see source for full glory; some 57,000 cells and 150,000,000 synapses pictured):

As one man commented:

Even worse – each neuron that contributes to your cognition and consciousness might contain 100,000 Microtubules – each Microtubule Operating at 10,000,000 oscillations per second – and being Quantum photonic memristers

Good luck with the AGI tho

(Photonic memristers are quantum objects something like a continuous transistor which uses light instead of electrons.)

You might argue that all of this is like the doped silicon in transistors inside computers. Lots of “quantum stuff” going on inside them, too. But we don’t need to know most of this, because we’re only interested in whether the transistor is on or off. Then you might say that all those things like microtubules, vault proteins, microfilaments, and much much more beside, only function to turn neurons on or off. And thus it is neurons that are like transistors.

But neurons are not on or off. And neurons are not the whole of things.

At this link is a small animation of a just one neuron and its connections. The picture had this apt quote:

“I remember this moment, going into the map and looking at one individual synapse from this woman’s brain, and then zooming out into these other millions of pixels,” said Viren Jain, a senior scientist at Google, in Nature Magazine. “It felt sort of spiritual.”

None of this is on-off. There are continuums which act differently at different frequencies. The brain is seething continuous action. It is impossible to “freeze” a moment, as we can do with trivial ease in a computer, and write down the state of a brain. And this is so even if we could freeze a moment in time, because there are not just discrete states here, but continuous states of which we have only the barest glimmer of cause in only a few things, and in any “quantum” object we can never know the cause of change.

You will see lots of cartoons of neurons acting like non-linear regressions, summing weighted inputs and the like, hence “neural nets”. But that’s not real neurons. That’s models! Horribly crude ones at that. The Deadly Sin of Reification hits people hard in AI.

In short, you can’t do the same Paper Test with a brain as with a computer. We are not simple machines. We are not machines at all. And we are lot more than just brains.

All this is before arguments proving that immaterial nature of intellection (in us, not computers). We’ve covered these only briefly before, and will do so in more depth later. But for now, assume they are true: intellection is not material, though it does interact with “brains” (not in a Cartesian way). Then again, it is not possible, which is to say it is impossible, to write down the state of a person’s thoughts.

Or, even more importantly, to know the causes. Even if you can imagine learning the causes of the continuous changes in neurons, you won’t in immaterial thought.

More

• AI Cannot Hallucinate Nor Lie
• “AI Will Kill Us All!” Say People Programming AI To Kill Us All
• AI Limitations (and Strengths): Predictions
• The Limitations Of AI: General Or Real Intelligence Is Not Possible
• Why The Recent Claims Of AGI Fail: Plus, A New Test For AGI
• AI Needs To Feed On Fresh Meat, Or It Dies
• AI Is Driving Many Crazy: Our Latest Panic

Here are the various ways to support this work:

• Subscribe at Substack (paid or free)
• Cash App: $WilliamMBriggs
• Zelle: use email: matt@wmbriggs.com
• Buy me a coffee
• Paypal
• Other credit card subscription or single donations
• Hire me
• Subscribe at YouTube
• PASS POSTS ON TO OTHERS