What I saw at NeurIPS 2024
Lots of genius, and not much ethical reflection
I spent four days at NeurIPS 2024 in Vancouver this month. If you don’t know, NeurIPS is short for Neural Information Processing Systems. It’s a big conference, and it’s quickly getting bigger.
Something like 14,000 people were there, most of them professionals or academics in the world of artificial intelligence, machine learning, computer science, math, and related fields.
In other words, it’s kind of a big deal.
And it was my first time there. You can read about what I did at NeurIPS in this other piece.
Here I want to tell you what I saw, not what I did.
I saw a very international crowd. Along with that, I heard a lot of languages being spoken. For me, this was a lot of fun.
I like people-watching, and I also like walking through crowds and listening to other languages. I had some good and educational conversations with people from around the world.
I saw a LOT of posters in the poster sessions. The poster sessions took place in two separate gigantic halls. They also rotated through presenters, so there must have been thousands of posters.
This is also kind of cool, since each poster usually represents a tremendous amount of thought and work and commitment on the part of many people.
It’s impressive to see such devotion to solving problems that might matter to only a few people. Sometimes such devotion changes the world.
And sometimes it creates new companies, and new products. I saw pretty much every big AI company at this conference: Meta, Apple, Google, Tesla, IBM, and a few dozen others were there, partly to show off what they are doing, and partly to hire some of the brilliant people who were presenting their research.
In the poster presentations I saw a lot of interesting work being done. Lots of it was mathematics-based and might not interest my readers. Some of it should interest anyone with a glimmer of curiosity about their world, though.
When I hear academics in the humanities talk about AI, they often mean something like LLM’s, generative AI like Chat GPT and Gemini, and a few other already well-known products. After all, those are the things we are most aware of our students working with.
NeurIPS is a good reminder that AI is far more than the tools that are well known, and that it is rapidly touching on almost every area of our lives.
Here are a few examples of some problems I saw people using AI to solve:
How do you reassemble a broken puzzle with missing pieces? Some grad students from Italy talked with me about how they’re using AI to help archaeologists take a few remaining pieces of an ancient fresco and turn it into a restored image.
If you have mass spectrometry data from plants, can you discover what molecules are in the plants that all other methods of searching have missed? This was especially interesting, since it has applications for clean water, pharmaceutical prospecting, environmental DNA, and probably a zillion other things I haven’t imagined.
I also saw (and heard) some things that were a bit alarming.
For example, I saw some AI-generated videos that looked like they were not AI-generated, which had me wondering about the effect AI might have on the human creative process, and on jobs like my own.
What jobs could AI replace, and what might that mean for people who have devoted their lives to preparing for and then doing those jobs?
When I’ve asked that question, people have often told me to look on the bright side: we will be free from toil, they say. I am not so sanguine.
It seems to me that we flourish when we have meaningful work. And yes, we can separate toil from labor, but the line of division between them is not always clear.
It’s also not clear that when we find ourselves without work that we automatically gravitate towards pursuits that are healthy for us and for our communities. Sometimes it is work that keeps us healthy, in body, mind, and spirit.
Another concern comes to mind: as Michael Sacasas pointed out in a recent essay, “To Hell With Good Intentions, Silicon Valley Edition,” some of the “solutions” big technology companies offer might be solutions that we are not free to refuse. Who gets to decide which work is not worth doing?
In the last week, two of my creative acquaintances expressed concerns about how AI was either diminishing their creative work or causing them alarm. One of them is a filmmaker and photographer who makes nature documentaries for publications like National Geographic and the BBC.
He commented that it used to take a long time to make a beautiful scientific film, and when it was done it was both educational and a work of art.
Now it can be made quickly, and without spending time in the field.
Surely that’s good, right? Not exactly. He said that everyone one of his colleagues in filmmaking is feeling burned out.
They work hard to understand nature and then to tell a good story about what they have studied, but there is no longer much of a market for their work.
He commented to me, “A world where machines do our own work for us — what kind of uninteresting world is that?”
My other friend, a philosophically sharp screenwriter, raised an interesting question about the outputs of generative AI. He spent some time playing with Gemini, trying to get it to write a script in which a major prophet advocated for Coke rather than Pepsi. It took a lot of prompts, but in the end he was able to get it to write a dialogue between Socrates and Jesus. Gemini made sure to insert several disclaimers indicating that it was definitely not endorsing one product or disparaging the other.
Both of us noticed that Gemini seemed more concerned about misrepresenting Coke or Pepsi than about misrepresenting Jesus or Socrates. That tells us something about Gemini, and about what matters to the people who have built it.
Throughout the conference I heard people saying things like “We are about to see a shift in human history.”
What was strange to me as someone who was invited there to talk about religion and ethics, is how little discussion there was about religion and ethics.
Surely a shift in human history calls for us to think about what we are doing. And why we are doing it.
One speaker did point out that AI companies should pay people to actually work on ethics, but I am not holding my breath.
Another speaker was asked what we should do to motivate people to build AIs with attention to rights. The speaker acknowledged that was important, but was reluctant to comment more.
Francis Bacon, in his Advancement of Learning (2.vii.3) says that we might benefit from separating our discussions of purpose and meaning from our research into energy and matter.
In just a few words, Bacon moved the scientific method away from Aristotle’s old idea that science is the knowledge of causes, including formal and final causes.
“Move fast and break things”
The result, for Bacon and for us as his heirs, is a method of science that can move more quickly by attending to changes in matter rather than to those eternal causes like purpose and meaning.
And he was right. When we bracket our conversations about ethics and metaphysics, science can move more quickly.
And there is money to be made in moving more quickly, and in breaking things.
Just ask Oppenheimer, and his collaborators on the Manhattan Project. They got things to move very fast, and they enabled us to break things like we had never done before.
I don’t mean to say that AI will create a new kind of weapon. We might use it for that, or we might use it for putting together broken things, like ancient frescoes.
But I do note that after Oppenheimer saw the results of his work, in at least one interview he found himself returning to one of the world’s great scriptures, the Bhagavad Gita.
I find AI mathematically fascinating, and I loved walking around and seeing the posters. It might be that we will use AI to solve some big problems that we need to solve.
But perhaps, while we work on those problems together, it would be good for all of us — especially if we think we might be on the cusp of generating a shift in human history — to spend more time with such scriptures, and not less.
If you’re curious, you can see more of my writing on Medium here: https://medium.com/@davoh
If you’re not yet a Medium member, I’d be glad to have you sign up through this link that will also support my writing: https://davoh.medium.com/membership
