Woah, I have some smart colleagues researching AI at ASU!

Arizona State University is known for it's leadership in using generative AI. But what about research into the cutting edge of artificial intelligence? Digging into this, I was impressed.

Image: Midjourney

Arizona State University likes to be Number One. We’re one of the largest universities in the US with over 145,000 students.¹ We’re just been ranked #1 in innovation yet again — for the tenth year running no less — by US News and World Report. And we were one of the first universities to start collaborating with OpenAI on exploring how generative AI can transform everything from learning to research.

In fact it’s fair to say that ASU is leading the way on how AI could transform learning and education as it explores ways of using the technology to extend and enhance learning environments, approaches, and opportunities.

This is all very exciting. But as my work around emerging technologies is as much about what’s coming next, and how future generations of breakthroughs will potentially transform society, I was interested to find out more about what my colleagues are doing at the cutting edge of AI systems and capabilities research, not just AI implementation.²

I should, of course, have a handle on this already as someone who’s in the thick of the university’s work around AI. But ASU’s a big place, and sometimes the really cool stuff gets lost in the mix.

So to fill some of the gaps in my knowledge, I turned to our School of Computing and Augmented Intelligence, or SCAI.³

First things first, you noticed that the “AI” in SCAI doesn’t stand for “artificial intelligence,” but “augmented intelligence.” The name represents a strategic move going back a few years that reflects a focus on using computer science to enhance and extend human intelligence, not necessarily to replace it.

That said, this is where some of the most interesting research (although not all of it by any means) on artificial intelligence is, in fact, occurring.

For starters, I was interested to see that ASU (based on SCAI’s programs) has just been ranked #19 by US News and World Report for the Undergrad Computer Science speciality of AI. This is pretty impressive considering the number of AI-focused programs available around the country.

It’s even more interesting when coupled with developments at ASU around our new School of Medicine and Advanced Medical Engineering which will be bringing advanced AI capabilities to research and learning at the cutting edge of medicine.

But I was particularly interested in research being carried out at ASU at the cutting edge of AI-related science and technology. And to get a sense of this I turned to my good colleague and SCAI researcher Chitta Baral.

Talking with Chitta, the first thing he drew my attention to is the Computer Science Rankings web page.

I’m guessing that this is something that every good computer science/AI researcher is very familiar with. But to a lowly physicist/transdisciplinarian like myself it was new — and it was a revelation.

The website ranks institutions using conference papers published by computer scientists, and allows relative rankings to be explored in a range of different sub-areas.

It also dives deep into individual researchers’ profiles, publications, and publication areas.

The website is a quite wonderful although rather distracting rabbit hole of computer science data, which partly explains why I got so side tracked after Chitta introduced me to it!

Just looking at rankings based on AI-specific publications (one of many categories) over the past ten years, I was intrigued and (I admit) quite proud to see that ASU ranks #3 in the US:

Chitta did warn me that this is a little misleading (see below), but I was impressed nevertheless.

Digging deeper, the website ranks the top-publishing computer science researchers in this category at each institution. This is what it looks like for ASU:

Dig deeper still, and the domains each researcher is publishing in are revealed, which provide a more nuanced perspective on where their expertise lies:

As I mentioned above, just looking at the AI rankings is somewhat misleading as they’re based on just two publication venues: The American Association of Artificial Intelligence conference (AAAI), and the International Joint Conferences on Artificial Intelligence (IJCAI).

As Chitta pointed out to me, “many researchers these days prefer to publish in venues of their sub-field of AI, such as Machine Learning, Vision, or NLP”. He he also noted that the general AI venues of AAAI and IJCAI still are go to places for all of AI, including subfields such as knowledge representation and automated planning that do not have a direct presence in the rankings shown above.

In other words, the rankings here need to taken with a good dose of context. But they still indicate the prominent of ASU research at the cutting edge of AI.

For example, drawing from the lists above, Huan Liu is recognized around the world for his pioneering work in AI. Subbarao Kambhampati (Rao) is a highly impactful AI researcher and a former president of the Association for the Advancement of Artificial Intelligence. Yanjie Fu’s work is pushing the boundaries of different modes of machine learning in messy data/information environments. And Siddharth Srivastava and his group are making waves around autonomous agents and “intelligent robots that plan and act under uncertainty to accomplish complex tasks.”

I’d also add YooJung Choi, Yezhou (YZ) Yang, and Hannah Kerner to the list of SCAI researchers who’s work particularly caught my eye — although there are many others in the school doing ground breaking research.

Yoojung for instance is doing really interesting research on trustworthiness and AI. YZ’s Active Perception Group is working at the intersection of generative AI, deep learning and computer vision to better-enable machines to interact with humans in physical environments (something that is increasingly relevant given the rise in interest in humanoid robots). And Hanna is doing fantastic research at the intersection of AI, environmental systems, and working toward a more sustainable, responsible, and fair future for all. Her research also led to her being listed on Forbes’ “30 under 30” in 2021!

But I want to come back to Chitta’s work, and his research into natural language processing and large language models in particular.

I’d heard rumors that one of Chitta’s former PhD students was involved in research that formed an important part of the development of today’s powerful LLMs such as ChatGPT. But I didn’t know the details until we chatted.

The research in question, it turns out, was published in a pivotal paper by Swaroop Mishra, along with Chitta and others, on cross-task generalization in AI models. Chitta was Swaroop’s PhD advisor while he was working on this.

The paper was the first to develop the idea of “instruction tuning,” where AI models are developed to learn how to carry out more general tasks through broad instructions rather than rigidly trained to complete very specific tasks. And it allowed LLM developers to make the jump to the capabilities that are now common in text-based generative AI platforms.

While I’m sure this is common knowledge within AI research circles, I was impressed to discover just how pivotal ASU-based research was in the emergence of the current wave of generative AI.

Swaroop is now a research scientists at Google Deep Mind working on reasoning within AI models — another testament to the impacts of ASU’s graduates in the field!

What I ultimately took away from my conversation with Chitta though is that, while what we are doing at ASU around developing and using AI tools is impressive, we are also at the cutting edge of more fundamental AI research and innovation.

And as this is where the most disruptive and transformative advances in AI will occur over the next decade or so, I was excited to discover that my colleagues are significant leaders and contributors here.

Of course, the cutting edge of AI is very much a team sport, with advances coming through national and global collaboration and partnerships. And so rather than focus too much on rankings, it’s important to understand how the work of my colleagues in SCAI are part of a much bigger community of researchers who are quite literally creating understanding and technologies that are poised to transform the future.

But just as being part of a university that consistently ranks #1 in innovation makes me a little proud, I’m also rather chuffed to be part of a community that is at the forefront the bounds of what is possible in artificial intelligence, and its beneficial — and responsible — development.

1. These are the Fall 2023 figures — while fall 2024 figures haven’t been officially released yet, they are significantly higher!

2. Anyone familiar with my work will know that I’m interested in broad/transdisciplinary approaches to advanced technologies, society, and the future. However in this deep dive I was very specifically interested in what’s going on at ASU at the cutting edge of research into AI systems and capabilities — just in case anyone’s wondering why I don’t mention work around ethical, responsible, mindful, principled, and other socially-oriented framings of AI. We do a lot of this as well — but I’m leaving that for another day!

3. A quick shout-out to SCAI’s current director Ross Maciejewski who is doing a fantastic job of leading the school as AI becomes increasingly prominent.