Yuan-Sen Ting

How A.I. is Changing Science

All drawings are A.I.-generated

Disclaimer: A.I. is a big umbrella and can mean many things

Artificial Intelligence

Modern A.I. can seem scary

A.I. in Math Olympiads

A.I. in Coding Olympiads

Source: x.com/OpenAI

If you believe in IQ tests (you shouldn't)

Source: e.g., MaximumTruth.org,
LifeArchitect.ai,
binaryverseai.com

2021: GPT-3

The End of Human Research?

OSU Professor

Biopharmadive.com

Axios.com

Futurism.com

But should we really be worried?

The Plot Twist

At times,
A.I. can seem silly

Just this summer, first autonomous A.I. soccer championship

June 2025

Source: The Independent

Moravec's Paradox (1988) -

"It is comparatively easy to make computers exhibit adult level performance on intelligence tests or playing checkers, and difficult or impossible to give them the skills of a one-year-old when it comes to perception and mobility"

Reversing the evolution of "intelligence"

Evolution Timeline: What came first vs. last

Conversational and logical abilities are the easiest to imitate

A lot of our holistic abilities were developed much earlier

Not so soon. Scientific research
is a holistic endeavor

The Reality Check

A.I. still struggles with many tasks that are easy for humans

Princeton Language and Intelligence Lab, June 2024

Can A.I. reason about scientific charts?

ARC Prize Foundation (ARC-AGI-2, 2025)

Spatial Pattern Reasoning

Reproducing this with GPT-5

I asked GPT-5 and Claude to make slides for this talk

GPT-5

Easy-for-AI

Complex calculations

Easy-for-Human

Logical inference (?)

Memorizing information

Language

Coding

Spatial reasoning

Common sense physics
(water flows downhill)

Basic motor skills​ 

Visual reasoning

Understanding context

Working With What We Have

To mitigate the many "blindspots" of A.I.

We need to equip A.I.
to interact with
domain-specific scientific tools

In materials science

In chemistry

In astronomy (also here at OSU)

Sun, YST, et al. 2024

James Webb
Space Telescope

A.I. Agent

Designing fully self-driven labs can be frustrating

How I feel all the time

The scope of fully self-driven, autonomous, research is limited

Enthusiasm, not hype

A.I. as Co-Pilot

While fully autonomous research is likely not immediate,

A.I. has changed how I do science completely

Morning routine: "Interrogating" papers of interest with A.I.

My afternoon routine: Intuitive coding with AI assistance

On Education

How shall we prepare the next generation?
Scientific education needs to adapt accordingly

The purpose of education has not changed

- be THE EXPERT in your field

In my own teaching here at OSU - A.I. Tutor

In my own teaching here at OSU - A.I.-Administered Quiz

Students need to learn how to think critically about the output of A.I. 

Not only as consumers, but also builders of A.I.

The Ripple Effects

The changing landscape of work

@ stanford

The Deeper Question

Identity:

What makes a scientist
a scientist

Ship of Theseus

How many planks one can remove such that
it is still the same ship

How many planks one can remove such that
it is still the same ship

Policing AI usage is largely futile (and ill-defined)

Publisher
reviewer

Me

Which
makes you,
you?

Flipping The Lens

A.I. as our mirror

A.I. as our mirror

New way of exploring human behavior

Not just A.I. for science, science of A.I.

Steam engines come before
the laws of thermodynamics

The new "neural" science

Stepping Back

A humbling lesson

Yet, plenty of reasons to be proud

Employment

Identity

Education

Merit

Science, like chess

Science, like chess

Science, like chess

A.I. has no intention to understand the Universe,

Extra Slides