In an earlier segment of the podcast, “Can We Upload Ourselves to a Computer and Live Forever?”, Walter Bradley Center director Robert J. Marks and computer scientist Selmer Bringsjord discussed whether we could achieve immortality by uploading our minds to computers.
The basic problem with that idea is that human minds aren’t “computable.” Peter and Jane are not bits and bytes. Here, they continue the discussion, addressing the notorious Hard Problem of consciousness. People are conscious and even the most sophisticated foreseeable computers are not. And we are not at all sure what consciousness even is.
A partial transcript of the podcast follows (start at 8:50).
08:50 | Is consciousness a special case of cognition?
Robert J. Marks: In your paper, you claim that consciousness is a special case of cognition. That’s the first time I’ve heard that sort of claim. Could you elaborate and unwrap that a little bit?
Selmer Bringsjord (right): Let me just say that I couldn’t agree with you more about the “delayed scrutiny” [claims about uploading our minds to computers are pegged to an indefinite future] and the antidote to that is just taking a bet. But nobody wants to take the bets. So if I’d been around when Herb Simon — well, we’re talking the Fifties and he said, in a few years, we’re going to work it out. Or Marvin Minsky — well, it’s a few summers or maybe even one summer, don’t worry, we’ll bring you back this AI.
What I don’t get about this is, “Well, really, here’s five thousand dollars, here’s fifty, here’s a hundred, here’s my entire 403B that says right now, you can take it and my descendants will sort this out. Let’s work out the contract. I’m willing to make a bet. Let’s make a bet … ” But no one’s willing to take the bet.
Note: It’s possible to get a great deal of attention for promising thinking computers in the indefinite future without making any bets as to performance or even being held accountable for lack thereof.
Robert J. Marks: It’s kind of an algorithm of the gaps that someday we are going to have an algorithm that does this sort of thing, yet it has been promised since sixty years ago and nothing has really happened. And nothing has passed, as I’ve seen it, the Lovelace test that you proposed about fifteen years ago. So, consciousness isa special case of cognition?
Selmer Bringsjord: Right. I certainly will agree that that is not in any way universally affirmed and some people steer clear of consciousness and try to prevent consciousness from entering the scientific discussion, whether it’s neuroscience or even sometimes things that are more formal like treatments in decision theory—whether it’s normatively correct, that is, whether the agents are good at it, whether they’re making bad decisions. No, no, let’s come up with an account of decision-making that doesn’t take the consciousness associated with desire seriously. So we don’t have to worry about what desire really is and the consciousness associated with that, let’s keep it separate …
If we’re going to be honest with each other, you can’t instantiate these things in agents, at least agents of the human variety, unless that agent has feelings. Unless there is something that it is actually like to be human, unless the human feels pain, unless the human feels pleasure… Let’s just write down the activities that are part of being a cognitive agent as opposed to just an agent because in AI a textbook can say that an agent just computes a function from the percepts of the environment to actions. So even something that computes the square root is technically an agent in AI. But when we say a “cognitive agent,” we can’t suppress consciousness rising up before our faces and we have to deal with it. But again, some people can try to dodge it.
Note: Approaches to consciousness that are currently under discussion in science principally fall into one of three categories:
It is a material phenomenon: Philosopher Galen Strawson argues that, in order to exist in any scientific sense, consciousness must be “wholly physical.”
It is an illusion, naturally selected to aid survival: Neuroscientist Michael Graziano espouses this view. The problem is that, as Michael Egnor explains, “If consciousness evolved as an aid to reproduction, there is little reason to credit it with any particular effectiveness as a tool for ascertaining truth. It’s an aid to coitus, not contemplation. ”
It exists and pervades nature but we are only aware of human consciousness: That’s the panpsychist view: Scientific American, for example, has given panpsychism considerable respectful space in recent years because the alternative appear to make even less sense.
The idea that consciousness is a real but immaterial phenomenon is not at present considered a scientific idea, irrespective of evidence.
13:16 | Are consciousness and cognition non-algorithmic?
Robert J. Marks (right): So, if we have these things cognition and consciousness, which are attributes of humans, your claim in your paper is that they are non-algorithmic, that is, that you can’t write a computer program to simulate them. They are not computable. What is your argument that cognition is not computable?
Selmer Bringsjord: Well, first, to be careful, some of them are not computable. Clearly, playing checkers is a computable process, provably so by definition. If we want a simpler case, applicable even to young children, then Tic Tac Toe. Even a very young child can learn an infallible algorithm for Tic Tac Toe but when they make those decisions… they’re doing something that’s computable. … But I’m talking about things that distinguish the human person.
14:48 | Examples of cognition that are not computable
Robert J. Marks: So what would be some examples of cognition that were not computable? Clearly, chess and checkers are computable.
Selmer Bringsjord: Well, at the top of the list is conjectured discovery and confirmation in the formal sciences or—to use what is probably good enough—mathematics.
Doing mathematics where you are conjecturing and making discoveries and confirming them is untouchable. I have a book—eternally undone but getting quite close now—on Gödel’s great theorems. If you just look at one little piece of his career, where he proves that the continuum hypothesis (basically that there is no set between the natural numbers and the reals) … this is astounding. So when we talk about AI doing all this work, it doesn’t really do anything in mathematics.
The great thing about that one is that we can inspect the output produced by humans that are playing in this space. So it’s not like they just give us vague reports about doing these amazing things. They write their results down. So we can look at the results and we can say, “Can a machine generate something like this?” and the answer is, flat out, with a ring of iron, no. This would be my number one.
Number two would be creativity… As much as he was a maniac, Wagner. I mean, how does one human being create the storyline, the music, the libretto, which is essentially poetry, and produce that out of whole cloth?
Let’s just think about love. What’s it like for one person to genuinely love another person and be loved by that person. Now, we can’t mathematize that. We’ve got no account of what it is. In fact, the leadng formal account of human emotions—the so-called OCC account, which I like very much—has come up totally empty on any kind of formal account of love. And yet, we love people and we want to be loved and we know what we’re talking about… so every human being on the face of the planet can just see that there is a major problem here!
Note: Transhumanists do not usually try to explain how they would create immortality by capturing human consciousness in a machine so one can only evaluate the social movement in terms of the issues it would raise if their ideas were remotely possible. Here are a few questions that have been raised:
• Bartlett’s question is especially pertinent because schemes for reproducing you as a computer program may require killing you first..
• Would you want immortal life as a computer program? What would be left of life as it matters to us? Here’s a test question: Would you give up your right arm for a robotic device that performs better?
Next: Why a computer will not write the Great 21st Century Novel
Earlier: Can human minds be reduced to computer programs? In Silicon Valley that has long been a serious belief. But are we really anywhere close?
Thinking machines? The Lovelace test raises the stakes. The Turing test has had a free ride in science media for far too long, says an AI expert. (This is the partial transcript and notes to the earlier part of the podcast.)
Thinking machines? Has the Lovelace test been passed? Surprising results do not equate to creativity. Is there such a thing as machine creativity? The feats of machines like AlphaGo are due to superior computational power, not to creativity at originating new ideas. Dr. Bringsjord sees the ability to write, say, a novel of ideas as a more realistic test of human vs. computer achievement.
- 00:39 | Introducing Selmer Bringsjord, Professor — Rensselaer Polytechnic Institute (RPI)
- 01:18 | Can computers fake all human behaviors?
- 02:02 | Can computers duplicate all human behaviors?
- 03:10 | Subjective measurement
- 04:40 | The definition of cognition
- 08:50 | Is consciousness is a special case of cognition?
- 13:16 | Are consciousness and cognition non-algorithmic?
- 14:48 | Examples of cognition that are not computable
- Selmer Bringsjord’s website
- David Gelernter at Wikipedia
- What Robots Can and Can’t Be by Selmer Bringsjord
- Bringsjord, Selmer, and Michael Zenzen. “Cognition is not computation: The argument from irreversibility.” Synthese 113, no. 2 (1997): 285-320.
- Claude Shannon at Encyclopædia Britannica
- Ray Kurzweil