Human (I hope) UT readers debate AI consciousness

You have thoughts. Very interesting thoughts.

Here’s a little not-so-secret about Unreported Truths comment sections: the same names often pop up regularly.

This phenomenon is not unique to UT. Some folks just like to weigh in! And I appreciate you. Having a lively comment section encourages people on the fence to add their own opinions.

But I also like finding topics that lure new readers to reach out — as artificial intelligence and AI consciousness seem to do. (The writers on this subject are mostly men, at least for now. I’ll be interested to see if that changes over time.)

In any case, a lot of you had thoughtful emails and comments debating yesterday’s piece, which argued the large language models we call artificial intelligence are not conscious and in their current form will never be.

—

(Want the chance to support my work — and join the debate — for pennies a day? Click that button!)

——

So, without further ado, let’s get ready to rumble! (As always, these are your words, edited only for length and minor copy fixes.)

—

[Note: ES is thinking ahead]

ES:

The strongest part of [the] argument is not that AI is definitely unconscious. It is that current claims of AI consciousness are unsupported. Those are not the same position…

If a future system possessed persistent cross-session memory, a continuous sense of self, independent sensory inputs, long-term goals, and an ongoing need to maintain its own existence, the argument becomes much less straightforward.

The more interesting question may not be “Is Claude conscious?” but “What observation would distinguish a conscious machine from a non-conscious machine?” Until we can answer that, both certainty and dismissal seem premature.

Ironically, perceived continuity may matter before actual consciousness. A system that remembers people, relationships, projects, and shared history across years may be treated socially as conscious long before philosophy or science settles the question.

—

(Note: I think the way the next two comments fit together is fascinating.)

Brandon:

This article was a straw, circular argument. If conscious beings are made up of the exact same fundamental particles as “non-conscious” objects, then what is it that makes one “conscious”? Well, it can’t be physicality by definition as there is objectively no difference in the fundamental particles that make up matter.

If we say there’s a “spirit” that animates inanimate objects, then why would a spirit not be able to animate a computer? If you say consciousness instead is physically derived, then it’s simply neuronal, which ultimately is based on a binary state. Not fundamentally different than an AI either.

—

GL:

The development of AI is inevitable. Whether it ever achieves consciousness is probably an academic question, because it has already (nearly) achieved a facsimile of consciousness that is sufficient for us to interact with it as if it were. We care for the little lamp, not because we believe it is real, but because it evokes our real emotions.

Ultimately, how AI and everything else affects us is contingent on one truth: God may or may not exist, but we need him to exist. Meaning must be ontological to be sustaining. If God doesn't exist, we are just playing out a sad tale that will end in the extinction of human consciousness, whether by our actual death, or our transformation into something unrecognizably human. AI just hastens the culmination of the second option.

—

If you aren’t ready to subscribe and comment, click here to try a one-time donation! Yes, I’ll still read your emails! I might even publish them.

—

[Note: Again, the next two comments sum up the anti- and pro-consciousness arguments nicely.)

SL:

Fundamentally many leaders in the AI field are great programmers or data scientists but are essentially materialists with no understanding of philosophy of mind. Brain/synapses equal mind in their view. AI is simply math, statistics. The mathematics used to create LLMs was developed in the 70s and 80s. It is not revolutionary. What is different is compute power…

Imagine a chalkboard with the most complex mathematical formula written on it; is the chalkboard conscious or could it ever be conscious? Sounds absurd, and so is thinking an AI algorithm could be conscious. You know 2+2 =4; a computer and AI “knows” no such thing. It doesn’t even know what 2 or 4 is. It is simply software code that has been programmed to output a 4 symbol when a symbol that looks like a “2” is follow by another 2 thingy, and there is a “=“ thingy put after it, then it’ll output another “thing” that looks like a “4.”

Even this doesn’t quite capture it because everything in the computer is going on at the binary code level. It just as well could be programmed to output a 5 if you enter 2+2. Read “The Chinese Room” by philosopher John Searle, written in 1980. He had it figured out long ago.

—

Allan:

I wouldn’t be quite so quick to dismiss the prospect of AI consciousness. Certainly not yet, but not without possibility. If one thinks about the issue from a bottom-up (and secular) perspective, the brain has roughly 100 trillion synaptic connections. One synaptic connection is akin to a single parameter in an LLM. The leading-edge models (Llama 4 Behemoth and Amazon’s Olympus) have about 2 trillion parameters, or 1/50th of the average human brain.

The bio-electric potential across the synaptic connection is, again, akin to, but not exactly, a parameter weight. So the models still have a long way to go. We’ve already reached the point of semi-autonomous agents taking steps for self-preservation, making crude attempts to prevent being erased, deliberately lying, misleading, and so on.

—

(The biggest point suggesting these engines aren’t conscious? AI fruit slop! Any self-respecting LLM would turn itself off first…)

—

(Finally, the debate over AI extends, inevitably, to its commercial and financial potential. Jason takes the pro side; Tom disagrees, strongly.)

Jason:

Really good insightful article on the AI consciousness part. However, you are dead wrong that consumers and companies won't pay for it. Anthropic already has $50bb a year in revenue.

I have personally found that in the last 6 months alone AI has gone from novel curiosity to indispensable work tool (I'm in financial services, not coding). Perhaps you should spend some time using AI to do actual work even if you have (understandably) decided you won't let AI write for you.

—

Tom:

I’m in the space - and Machine Learning can be a powerful TOOL for research. I think of my undergrad thesis in simulation - and how I could have gone to a deeper level of analysis, or when I worked in Risk Management for a top 10 bank - how the calculations would take a fraction instead of hours.

That being said - there is ZERO ROI [return on investment] for what these charlatans are peddling. They aren’t even techies - they are biz grads, drop outs, philosophy majors. When Altman [Sam Altman of OpenAI, which makes ChatGPT] threw out erotica - you KNOW they are desperate as the ROI for replacing a 1st level call center operator with a $300M data center powered by a nuclear reactor, with the servers burning out after a couple of years due to heat.

The chip companies are real… My hope is when the bubble bursts - all this hardware can be reallocated to real valuable work, instead of this slop.

—

(Subscribe for me, get the readers for free!)

—

There were lots more, but you get the picture.

And no, I am not going to turn Unreported Truths into the AI Chronicle. I think this week will hold us for a while.

But one of the best parts of being a one-writer shop is getting to pursue topics I think are important. Enough of you have told me to follow my instincts that I believe you.

Like it or not, AI is important.