How my dog made me understand intelligence, memory, and AI
I’m not one for New Year Resolutions, and it’s pretty much for the same reasons everyone else rolls their eyes at them. But I do have an exception: I set a goal every year for how many books I’m going to read. Usually, I reach the goal around September. My goal last year was to read 40 books. I only read 32.
The reason? In October, I swapped out my poetic novella of choice for what I felt represented a physics textbook: Max Tegmark’s “Life 3.0.” The professors I was working for at Lehigh University told me it was the best book on Artificial Intelligence out there. We were teaching general AI to journalism majors, so I trusted their judgement on the book. But every time I tried reading a chapter, I’d realize my mind was on a totally different planet and I comprehended nothing. So then I’d start over. I like to finish every book I read (because sometimes the ending is the best part) but I couldn’t make myself finish “Life 3.0.”
Months passed and I continued my research on AI, reading articles and teaching classes. I was understanding everything enough that “Life 3.0” just collected dust on my to-be-read shelf, and I mostly forgot about it. And then last week, my mom convinced me to foster a dog for her dog rescue. And it’s made me really consider my understanding of intelligence, as well as how my peers and I can get involved with developing AI.
My foster dog, Uno, is only a year old. But he listens well, hardly barks, and my boyfriend and I taught him how to sit in less than five minutes. “We really lucked out with this dog,” I say to anyone who will listen. “I don’t know how he got so smart.”
Which then leads to the question of how we measure intelligence. Can I really say he’s so smart if he tries to pee on every bush in town when we’re on a walk? Is he smart if he’s afraid of walking up stairs? But wouldn’t we consider Uno more intelligent than a dog who, all other things being equal, could never learn their own name? Understanding general intelligence is a fundamental step in grasping how it can be Artificial. So I decided to revisit “Life 3.0.”
The author concedes right off that there’s no single hard-and-fast way to define intelligence. But he breaks intelligence down into four parts: memory, rational thinking, learning and consciousness. So I started trying to comprehend what these all mean for me, in the context of myself, Uno, and AI. I start with memory.
A dog can usually remember their name or what it means to grab their leash before a walk. They can usually remember past trauma. But they can’t remember much else, especially compared to a human.
If you compared human intelligence to that of a machine’s, the contrast is stark. I don’t remember what time I scheduled my next hair appointment, but I don’t have to remember, because I put it in my Google Calendar. I use machines as a crutch, because their capacity to remember things will always be better than humans. I make machines work for me in this way, because their memory aspect of intelligence is far greater than mine. In this way, AI is way above humans.
Memory may be the most straight-forward aspect of four areas of intelligence, so it’s maybe the easiest for Artificial Intelligence, Google, etc. to surpass human capability. But is this the only one that easy? And why is getting AI to think for itself so difficult? Is Uno smarter than a calculator? Does he have a greater capacity for rational thought than Artificial Intelligence, and if so, what’s all the hype about?
Next week we’ll try and answer these questions about thinking and see where AI stacks up… and hopefully in a way that’s accessible enough that it won’t collect dust on your to-be-read pile.
Every week, I’ll be aggregating some stories and recent happenings in the world of AI that I come across in my research.
- The development of deepfakes can be used to create really poisonous misinformation. But in this case, it’s combined two of my favorite things: TikTok and Tom Cruise.
- Dogs can be trained to smell diseases like cancer and maybe even COVID. Now, a bunch of researchers are using a machine learning process to mimic that capability, and make it even more sensitive.
Remember when Jonathan Safran Foer, the author of Extremely Loud and Incredibly Close, left his wife so that he could be with Natalie Portman? And he didn’t even tell Portman, and she totally rejected him? I hope that you, too, can feel as blissfully confident as Safran Foer did in 2016.