Deep Dive
Shownotes Transcript
Sometimes I imagine human knowledge as a huge, convoluted membrane floating in space. And on one side is what we know, the whole of human scientific inquiry, and on the other, the vast unknown, and in some cases, the completely unknowable. Over time, that membrane grows in all sorts of weird ways, and that's where we at Quantum Magazine come in. I'm Samir Patel, the Editor-in-Chief.
At Quanta, we report from this frontier to bring you a greater understanding of the latest advances in basic or fundamental science and math, from astrophysics and quantum computing to molecular biology and classic unsolved math problems.
Quanta's been doing this for more than a decade across thousands of stories and videos, and yes, podcasts. We have the Quanta Science Podcast, which has translated many of our stories into audio form, and the Joy of Why, where mathematician Steven Strogatz and astrophysicist Jana Levin talk to specialists about their work. And today, we're launching something new.
The Quanta Podcast, a weekly show where I will be speaking with the editors and writers behind the magazine to offer another perspective on physics, math, biology, and computer science. First off, if you've been a fan of the Quanta Science Podcast, don't despair. It's going to continue here. Those are audio edition episodes, stories that we have on our site, and we're going to continue to do them as audio edition episodes running every two weeks.
I've been a science journalist for about 20 years. I've worked at magazines and websites. And one of the first things I noticed when I started as Qantas' second editor-in-chief was just how inspirational, dedicated, interesting, and eloquent our staff is when they talk about the subjects that they cover.
This podcast is really inspired by the conversations we've had in meetings, over coffee, at happy hour. For me, it adds a whole new dimension to our work, and I want to share some of that energy with you each week. For this trailer episode, I want to introduce you to one of those people, our executive editor, Michael Moyer, a very accomplished science journalist who has been with Quanta since almost the very beginning.
We're going to talk a little about what Quanta covers, how it's changed over time, and our latest big project. Michael, welcome. Hey, so excited to be here. How long have you been at Quanta? I've been at Quanta since the fall of 2014. So when you started, what was Quanta? What was it then? And how has it changed over the years? Right.
Right. So Quanta at the time was very small. I came on, I was the fourth staff member. There were two writers and then the founding editor-in-chief, Tom Lin. I came on from Scientific American Magazine where I was the physics editor at the time. And I had been talking with Tom Lin here and he had this editorial job open. And Quanta was known mostly for doing physics type stuff at the time. But he brings me on my first day in the office. He says, congratulations, you're now our biology editor. Yeah.
And I'm like, okay, switch. Yeah. So, but it was wonderful because I got to learn a lot about biology. We had a wonderful writer at the time, Emily Singer, who herself did know a lot about biology and she was able to teach me these things. So over the years, you know, I'm not formally trained in mathematics, right? But I've been had the privilege of being able to work with our mathematics writers and editors. And in doing so, I've been able to learn all these fascinating questions that people are really into. Yeah.
But then as now, we were dedicated to covering fundamental science, basic science, and mathematics, which not a lot of people did. And we've kind of really kept that focus over the past decade that I've been here. To the fresh listener, basic science, fundamental science. Right.
What does that mean? So that's as opposed to applied science. So the thinking was in starting Quantum Magazine that there's a lot of coverage out there about things that can be used in the world for good, which is great. A lot of coverage on medicine, right? A lot of coverage of technology, a lot of coverage of engineering and kind of how science is being used to improve people's lives. All that is wonderful, but we existed...
and started to complement that and not to reproduce it. So what we do is basic science. It's the fundamental questions people have about the rules by which the universe works is the way I think about it.
So in physics, it's what is going on at quantum scales? How do we reconcile quantum mechanics, which is the theories that they're very small with Einstein's theory of gravity, which works on very large scales. Those two theories don't play well with each other. And so there's a lot of scientists who are really curious about how to make them work. It's the biggest mystery in physics right now. We do a lot of stories, as you said, about molecular biology, about evolution. What are the rules through which evolution through natural selection works?
So we kind of cover what I feel is really just curiosity-driven science, people who just want to understand these kind of fundamental rules better. That's the way I'm often describing it to people because we use a term like
basic science in another context, right? That's like, oh, that's science 101. But that's not what this is. This is science at the very base levels of our understanding of how anything in the world or the universe or life works. Right. So when you described it as this membrane that stretches out, that's a metaphor that I've used over the years, which I love. And I love it because that intersection between what we know and what we don't know, right? This membrane is always kind of shifting and expanding over time. And
And as this membrane expands, the
division, the borderline actually gets larger, right? It's like a circle getting larger and larger and larger. So as we learn more, we continue to have even more questions. And that's where we live at Quantum Magazine. We're certainly doing something that feels to me very different and distinct from what other places are doing. Why do you think something like this podcast feels like a good next kind of evolutionary step for the way we communicate with our quantum readers? Well, as you say, I mean, my colleagues here are a
are amazing, right? They're some of the most intelligent, most curious people you'll ever meet. They really are dedicated to what they do. And every time I talk with them, I end up learning new things. They have this knack, I think, for in the written story, but also when we're talking, for making these sometimes pretty esoteric, complex, challenging topics feel new and fresh and wonderful. And so that's why I'm excited to do this too. Mm-hmm.
Now, our math writers and editors and our biology writers and editors, everybody actually from our art staff to our audience staff took part in a new project that we actually launched the other day. This is separate from the podcast. Can you tell me a little bit more about this thing that we just did?
So the other day we came out with a really huge project that we've been working on for a long time. It's called Science, Promise and Peril in the Age of AI. And what we really wanted to do was look at the way that AI is intersecting with science.
And what we do, basic science and mathematics in a really complex and interesting way. Science is obviously why AI exists, right? And not just for the obvious reasons of because computer scientists came up with it, but it has its roots in not just neuroscience and the way that people have thought about the brain, but also in fundamental physics theories going back all the way to the 80s.
But at the same time, now we're having this feedback effect where AI is really changing the way that science is being done in a lot of different areas, right? It's not just that people are using Chat Sheet BT to do literature reviews or anything like that. It's that they're now using generative AI to come up with new questions to ask in science. They're coming up with new ways to think about how we might do mathematical proofs.
Right. And they're really leading to a lot of soul searching in science and a lot of thinking about what science is going to be and how it's going to change over the next five to 10 years. This really feels like kind of an important sort of forward looking moment for this because so many people interact with AI on a daily basis now. It's easy when you're reading science news to see all the different ways. Oh, they're using AI in this project. You guys are using AI in that project. Just last year, two of the Nobel Prizes won.
We're both very closely related to AI and demonstrate like the two sides of that, right? Where one of them was given for the fundamental science, which was physics, that led to the development of neural networks. And the other one for an application in biology and protein folding that shows exactly how it's changing the field. So I think this seems like a great opportunity for us to take a bit of a step back and
and think about, okay, like, what does this actually mean for the way that people do science? It's a demonstration of what fundamental science means and then something that's actually altering fundamental science in really interesting ways. Math, too. So, Michael, what was the most surprising, craziest thing you learned in the course of working on this package of stories? So, I really enjoyed learning about what's called interpretability research. And what interpretability is, is...
the way that researchers are trying to figure out what's going on inside a neural network, the things that power chat GPT and image generators and all sorts of things that we have today. And inside this neural network, we have many, many interconnected what are called neurons. In reality, they're just little mathematical functions, but they all have relationships with one another and the output of one goes into the input for the other and they exist in many different layers. And
And out of this complicated structure and with enough training data, you're able to get these really amazing behaviors. But what is actually happening inside to make these amazing behaviors come out on the outside is still very much a mystery. We called an entire section of this package the black box because that's often the analogy people use.
And I think people might be surprised that we actually don't know how chat GPT or some of these image generation models work inside. They're doing things that surprise us. Right. And they're doing all these things that surprise us, despite the fact that we can go in and interpret
ability, researchers really can go in and look and see, okay, what every little neuron is doing. This is an ability that every neuroscientist would kill for, right? To be able to have a full map of the brain with every strength of every connection between every neuron in the brain. But even though we have it, it still doesn't get us anywhere closer to really being able to solve this problem of how all these interconnected mathematical functions end up giving us something like
ChatGPT. So I was really surprised to learn that there are ways that you could then go in and not just look at what individual neurons are doing, but actually start tweaking each one, right? Playing with the knobs and then seeing what the output is on the other side. And by doing that, being able to create some sort of a map, right, of what's going on and how these single little things, when put together in a complicated enough manner, are able to emerge with this really incredible behavior.
And that's a really fundamental way that science is often conducted when you have a complicated system or when you have a bunch of circuit breakers in your house. You turn off one and see if it turned off the hallway light. But doing this over the billions of digital neurons, parameters, whatever it is inside these AIs is what they're doing to try and get to another level of interpretability. I mean, there's an old saying in physics that
that more is different, right? And what that means is when you get a complicated enough system, the same rules that you were using to look at one or two or three particles no longer apply, that you have to have new sets of rules and new understandings. And right now they're just trying to build up that understanding piece by piece.
And I think that readers, when they look at this package, will piece by piece get a bigger, deeper understanding, not just of how AI is changing the sciences, but really how it works, how it came about, the fundamental science that contributed to it.
So our first episode coming on May 20th, we're going to dig a little deeper with one of our writers into one of these stories that's in this package. I'm very excited in the coming weeks to talk to just about everybody that we've got here. So, Michael, thank you for joining us. I'm looking forward to the next one. Thank you for having me.