Screentime correlates more with kids' brains than IQ; Matt Yglesias gets Dan Dennett wrong; Why literary fiction awards keep falling for AI scams; & more
Desiderata #40
The Desiderata series is a regular roundup of links and thoughts for paid subscribers, and an open thread for the community.
Contents
Please remember, Dan Dennett said LLMs aren’t conscious
Why literary fiction awards keep falling for AI scams
If AI is so smart, why do its apps suck?
All the colors your screen can’t show you
Screentime correlates more with kids’ brains than IQ
From the archives
Comment, share anything, ask anything
1. Please remember, Dan Dennett said LLMs aren’t conscious
As we shift from focusing on intelligence to focusing on consciousness, there have been a bunch of think pieces these past months about whether LLMs are conscious from both professional outlets and the blogger class.
Some of these takes have been… pretty bad. Like some people need to be taken away to philosophy jail.
Which is actually fine? Doesn’t this happen with everything? Five years ago, basically no one would have been able to tell you what “Moravec’s Paradox” for robotics was. Now, bloggers and commentators are often well-versed in the difference between AI and AGI, and how deep learning works, and ably use words like “transformers” and “next-token-prediction,” and discuss the previous history of the “AI winter” and scaling laws—there is an entire panoply of AI-related vocabulary that is now spat out with facility by the pundit class. Over the next decade, this same process will play out with consciousness.
But for now, there will be a lot of mistakes.
For instance, earlier this month the popular blogger Matthew Yglesias wrote a piece saying:
I think… that as the experience of conversing with a chatbot converges on the experience of conversing with a very patient human, we should assume the chatbot is having an experience similar to being a very patient human.
Yglesias primarily cites popular philosopher Dan Dennett as someone whose work would support Yglesias’ own view here.
Why Dennett? The thing is, everyone is looking around for a standard-bearer for the Yglesias position, which is to assume that because a chatbot can converse, it is therefore likely having subjective experiences the way a person holding a conversation would. They want to call this position “functionalism.” But that’s not functionalism! Functionalism is not that two systems behave the same, therefore, they have the same consciousnesses. Functionalism is the idea that what grounds minds in physical systems are the causal roles played by the mental states, between input/output (and so behavior), but also between other states, and that these causal roles are substrate-independent and could in theory be implemented in silicon or what have you. Basically, if you are a functionalist, you probably believe that artificial consciousness (AC) is physically possible (we could build it, somehow, with some advanced or futuristic technology). But you definitely are not committed to LLM consciousness.
Dennett himself (arguably the great inheritor of functionalism after Putnam) firmly rejected LLM consciousness.
Here’s Dennett:
LLMs are not people: They’re counterfeit people.… I want to suggest that counterfeit people are more dangerous, more potentially destructive, of human civilization than counterfeit money ever was.
Dennett was strongly against any talk of consciousness when it came to LLMs, found it absurd, dismissed it with a wave of a hand, and regularly said that even if artificial consciousness were to be discovered in the future (presumably, via some architecture that looks quite different from an LLM), humanity should collectively agree not to build it, and that we should instead focus on fashioning AI into intellectual tools. And remember, Dennett was around to see various advanced models, including GPT-4. This wasn’t an argument from ignorance.
So I have bad news: the reason why it’s tough to find a famous and popular modern philosopher of mind who is a good standard-bearer for the position of “It holds a conversation like a person, therefore we should attribute it a person-like consciousness”… is because it’s a bad position!
2. Why literary fiction awards keep falling for AI scams
Meanwhile, the anonymous authors Claude and ChatGPT are hard at work as emerging writers, collecting accolades for amorphous slop. Last month the Commonwealth prize, published by Granta, was (almost certainly) given to an AI-generated story, “The Serpent in the Grove” which contained banger lines like:
They called her Zoongie. Maybe it was a name; maybe rain took a shape and decided to keep it. She had the kind of walking that made benches become men.
Fresh off this debacle, the exact same thing seems to have just happened again (there’s supporting evidence, including AI-detector scores).
Anyone with a reasonably good internal AI-detector should have it pinging about a mile a minute while reading either of these “prize-winning” stories. In fact, the latest case literally re-uses a lot of similar language and tropes as the Granta-case winner, which are all heavy favorites of AI.
AI gravitates precisely toward this genre of short story, where the narrator floats around disembodied of everything but remembrances of things past, like memories of olive groves (or whatever), and the hands of their grandmother, and so on.
But I think the hard truth is that this genre of short story was always bad, and its success was entirely its legibility, not its content. You can see this by transposing away all its mystery. Imagine if I wrote a story in which…
The entrance to Dunkin Donuts hit with a blast of cold air, the door opening to reveal its always-earthy and always-yeasty smell, cut by an atmosphere made by machines. Like a hospital for sugar. My grandmother, her skin as thin as paper, handled the change at the register with reverence, counting each quarter, for a moment the girl she had been back in the Great Depression. It was the same order we got every time—herself a coffee, black, and me a cruller, also black, a chocolate rope I bit into hastily. The tantalizing scent of the cruller wafted over me, as it had in New England for hundreds of years.
‘Erik,’ her ancient voice whispered as we slid into our booth.
‘Why we whispering gran’mama?’
She patted the orange-lined table.
‘Because of this place. Because some places always remember.’