AI is drastically speeding up the process of formally communicating research findings in the computer science field, according to the founding dean of UW-Madison’s College of Computing and Artificial Intelligence.
Remzi Arpaci-Dusseau spoke this week during a seminar hosted by the university’s School of Medicine and Public Health. He discussed how AI has largely taken over writing all computer code as the technology becomes more deeply entrenched.
Major tech firms in Silicon Valley just last year transitioned “almost overnight” from writing most computer code by hand, to delegating this work to the latest generation of AI agents. And Arpaci-Dusseau said the industry is never going back.
“There’s not going to be a world where we say, ‘you know what, that was really fun to do it much more easily. Let’s go back to doing it the hard way,’” he said. “Doing code by hand will be viewed as this period between 1950 and 2025 … however long humans are around, there were these 75 years where people laid out code brick by brick, line by line, that will not exist beyond this time.”
Along with writing the actual computer code, AI models are transforming the method of producing research papers, Arpaci-Dusseau said. One of his graduate students used the technology to assess hundreds of other published scientific papers, build a guide for writing them effectively and ultimately produce a high-quality end product.
“Our papers usually would take a year, like from beginning, idea to end. You know, build, analyze, write the paper,” he said. “And we did this whole paper in probably four or five weeks, like it’s crazy how much faster we could go.”
He acknowledged concerns about human versus artificial authorship for works produced this way, noting it raises questions around what is being created by a person and what is the product of the tool. In this case, he said “it was clear the person was still the important part” of this process.
“There’s going to be this bifurcation of people who are using these tools are starting to go really fast at the things they do … and the people that don’t are going to go at the same speed as before,” Arpaci-Dusseau said.
He also briefly touched on concerns around the huge data centers that are powering this AI revolution, emphasizing the “incredibly large” energy demand for these projects.
“Only time anybody ever yelled at me about something I said is, I said something on Wisconsin Public Radio like, hey I wouldn’t want a data center in my backyard either … it’s a serious question we have to ask ourselves as a society,” he said. “Do we want to invest in these sorts of things, and how do we invest in them and how do we do that in a responsible way for the world going forward?”
And while academic leaders like Arpaci-Dusseau are considering these and other questions about the rapid rise of AI, huge forces are moving ahead with the buildout to sustain this growth.
Last year alone, the tech industry spent three-quarters of a trillion dollars on capital expenditures to support AI, Arpaci-Dusseau said. That’s expected to rise to $1 trillion for this year.
“That’s coming from four companies,” he said. “That amount is more than the next … Fortune 500 companies spent on anything. It’s the biggest investment in almost anything probably in human history, going on right now.”