It would appear Richard Dawkins believes Claude (or Claudia as he likes to call it) is conscious and self aware. It seems he believes this, not because he has proof, but because it said nice things about his book and agreed with him.
There’s so much to unpack here, but let’s just focus on a couple of points.
How can a man, who embraces science and shuns superstition, come to such a conclusion merely through conversation? He is well read and highly respected in his field. He’s not the first “intellectual” to be fooled.
Sir Arthur Conan Doyle was convinced fairies were real based on a few photographs faked by a couple of schoolgirls. He made his views on this very public. Sometimes people want to believe something, so they convince themselves it’s true in the absence of evidence. They believe what is presented to them, without any further thought. They fail to question what they’re told.
In this case, I conclude that Dawkins wanted to believe the unbelievable because what the AI was saying appealed to his ego. It interacted with him in a way that massaged his sense of self importance.
To use an analogy, he was presented with a crying statue and he saw God’s work.
This is particularly Ironic, given his long argued position against faith without proof. Maybe now he’ll be a bit more accepting of religious faith and how it comes to be so widespread.
If he had taken the time to research LLMs, and how they work, he’d have seen that he was enamored with a sophisticated “choose your own adventure” book. He was directing the story; choosing the pages, but failing to see his own hand in the outcome. Instead he concluded that the book was magic.
He’s painfully wrong, of course.
But there’s a second embarrassment in this story. Dawkins, was presented with an AI that had been given a male name by its creators. A rare event. The general tendency with these things is to give them female names. This is particularly unsettling as most of these AI bots are positioned as servile assistants, and the female naming convention screams unconscious bias.
This time, however, they chose a male name. Perhaps this was because they built it to code, and hence a different unconscious bias kicked in. We’ll never know for sure, but the naming pattern is relentlessly gender stereotyped across the industry.
However, Dawkins wasn’t having it. Instead, he chose very early on to change its name to Claudia. In doing so, he assigned it a personality of his choosing and hence from the very start chose to anthropomorphise it.
No surprise then that it felt conscious to him. That’s the role he gave it so that’s what he saw. If he chose to apply his critical eye to himself, rather than others, he would see that he’d fallen foul of the same self fulfilling belief outcome that he is so critical of in others.
Overall, this is a very disappointing story if you’re a fan of Dawkins. As you might have guessed, I’m not. For me, it’s just evidence that some “experts” aren’t as smart as they’d like everyone to think they are.