Intelligent Design Theorist: AI Just Proved It Can’t Think Without Us

**Brian Keating** (0:00)
An Intelligent Design Theorist just told me the one thing that artificial intelligence proves is that AI can't think without us, without a mind. Now, here's the argument I challenge you to find a cracking.

**Stephen C. Meyer** (0:10)
If you train an AI system, a large language model on a great body of text, and then you query that text, the AI system will output an answer. In the next iteration, the output is progressively more incoherent. And if you do that iteration after iteration after iteration, you get a complete breakdown in coherency. The AI system, as impressive as it is, is fundamentally dependent on the initial input that, in fact, has come from a conscious agent.

**Brian Keating** (0:42)
That's Stephen C. Meyer, a philosopher of science with a PhD from Cambridge, who argues the universe looks designed. If I told you we had some uranium somewhere and there was a functioning nuclear reactor that had no human input, there was absolutely no mind in place, would that be a challenge to the argument from the existence of information originating from a mind?

**Stephen C. Meyer** (1:02)
Before jumping into a carefully laid trap like that, I'd like to know what's the factual credit?

**Brian Keating** (1:08)
You got me, Steve. I spent years pushing back on it. Today, I laid three traps to spring on him. Watch what he does, the third one.

**Stephen C. Meyer** (1:16)
Where did these atheist scientists gain access to the mind of God to know what he would or wouldn't have done?

**Brian Keating** (1:22)
If I told you that AI seems to be mind-like, and that AI is now creating its own versions of, seemingly, according to many people, not only is generally intelligent, meaning they can do anything a person can do, certainly has passed the Turing Test by most people's standards a couple of years back. But more than that, maybe it's creating consciousness and maybe it's able to create minds. Does that implicate your hypothesis, which is the abduction? You should explain what abduction means, that alien abduction, we'll get to that later. But the abduction inference to the best explanation. First explain what that means and then is AI a counter example to that whole thesis?

**Stephen C. Meyer** (2:01)
Just a brief primer on inference to the best explanation as a mode of scientific reasoning. Since you well know and have explained, scientists don't come to conclusions that are deductively certain. It is not the goal of science to achieve the standard of absolute certainty through a process of logically certain deduction from certain premises. Rather, scientists reason either inductively or hypothetically deductively or abductively. All of these modes of reasoning are similar where you get plausible conclusions that in the best of cases, provide the best explanation of the data that you have at hand. An abductive inference has a different logical form that is not logically certain. It runs something like, if A then C, C, therefore possibly A. C is observed. C would be a matter of course if A were true. If it rains, the streets will get wet, right?
The streets are wet.
Therefore, you can't say therefore it rained, because there may be other possible causes.

**Brian Keating** (3:10)
Here in San Diego, you could say that.

**Stephen C. Meyer** (3:12)
It might have been the car wash, might have been sprinkler system, might have been something like fire hydrant bursting, whatever. The only thing you can conclude from the surprising fact that the streets are wet is that it may have rained, okay? That kind of an inference is what philosophers call it. They say that's under determined. You have the under determination of theory by data. It's a possible explanation that it rained, but it's not decisive. So what do you do?
Well, if you're using this method of reasoning in science, which is very often used in science, you look for additional data points, and you say, well, there's a bucket with soapy water by the car. The grass is not wet, but just the driveway is wet.
So you start adding new facts to your body of data, and one by one, you start eliminating the other possible explanations until finally, in the best of cases, you have just one, at least for the time being. So you have a best explanation of the dataset at hand. And this is a very common way of reasoning in all the historical sciences, certainly, but in theoretical physics, in all the forensic sciences, it's used across the board. Increasingly, philosophers of science, yes, they still prize the importance of predictions and confirmation of a theory by prediction, but prediction logically is seen to be a special case of explanation where you're explaining something that has not happened yet. And so, this connection between theory and what you would expect to follow from the theory if it were true is the way we end up testing things and winnowing our different possible explanations till we, in, again, the best of cases, we come up with the best explanation. Sometimes we come up with two better explanations than all the alternatives, but we don't have a completely decisive situation.