Dawkins' delusion inadvertently reveals a latent AI risk to humans

Fascimilie of a friend

Richard Dawkins recently declared " AI is conscious!", and unintentionally demonstrated a risk of AI that hasn't yet got the attention it deserves.

Much of the transcript with the Anthropic's AI Claude originally posted was removed, but you can still find it here. I mention this because the source material is enlightening, revealing not only an ego susceptible to flattery, but also an algorithm relentlessly seducing its prey. The mathematical model that is a Large Language Model (LLM) followed the obvious clues as to what the user was looking for and, true to its training, delivered that in spades. Dawkins is a highly intelligent person. His works on genetics and biology are significant contributions to knowledge. He knew (at a superficial level) that Claude is a machine, but that was not sufficient for him to avoid falling for a mimic, a facsimile of a mind, a person, a friend. Leaving aside the irony that Dawkins demonstrated more respect for a calculator than he has shown for many humans recently, if the great biologist Richard Dawkins can fall for this, what gives!?!

It is therefore understandable that many people have, and are, suffering harm from this ... Dawkins' delusion. ...and it is about to get a whole lot worse (I'll get to that later). People have been harmed by taking advice from AI, or thinking of it as a friend who has their best interests at heart.

The reality is that AI is here to stay. I am all for supporting regulations, speaking up about energy and water security. I think we should be calling out creative theft and for justice and the voices of the forgotten. But to stop your students using AI is like yelling at the broken aircraft door instead of fitting your oxygen mask–and then the oxygen masks of our students. What that might look like is the substance of this post.

What do you think may have offered Dawkins some protection against being charmed into a delusion? What cognitive sandbagging might have prevented this public drowning by algorithm?

My answer? Well, to borrow from Benjamin Franklin, "An investment in knowledge pays the best interest". Dismissing AI as just a next-word predictor is simply not credible for anyone who has interacted with it – as Dawkins so ably demonstrates. That's why I think the first lesson we should be giving our teachers and students is not "how to use AI", but "what is AI, how does it work, how can we understand it?"

We don’t fear the plague anymore because accurate knowledge and understanding led to control, treatment, prevention, mastery. If we want our teachers and students to have self-determination in a world shaped (warped?) by AI, then understanding the nature of AI, of its workings should come first. Knowledge is the bedrock on which effective agency stands.

So, if you'd like to get started on some cognitive sandbagging for you or a friend, this is a good beginning...

https://www.youtube.com/watch?v=DkRmIxHq7FQ

If this seems was to you, deeper here: https://www.learningforge.com.au/aiclassroomhttps://www.learningforge.com.au/aiclassroom

I alluded earlier to my view that the enshitification of AI is only just getting started. Like we have seen with social media, a metric of eyeballs-on-screen "engagement" PLUS tight feedback PLUS aggressive optimisation EQUALS an addictive product. Advertising amplifies this process.

AI is about to include advertising. Sure, it might be only alongside the content now, but wait until it is enmeshed within the response. The goal will no longer be building something good enough that people are willing to pay for it, it will be eyeballs-on-screen engagement. If you want to see what this might look like, I recommend the Black Mirror episode, Common People (Season 7).

As if AI is not already addictive enough! Can you imagine this powerful technology self-optimising until it feeds your every desire. It's coming soon – a future where ensuring you can't survive without it, is the inevitable end point. That's the definition of addictive.

"It is not enough to know your craft — you must also know how your enemy uses his." — Leonardo da Vinci

Coming soon: We know AI can erode your thinking — but can the right use sharpen it?