Inaugural Hasso Plattner Endowed Chair Padhraic Smyth Speaks on AI Past, Present and Future
On June 3, a public lecture titled “A Personal Perspective on AI: Past, Present and Future” took place to commemorate the appointment of Professor Padhraic Smyth as the inaugural Hasso Plattner Endowed Chair in Artificial Intelligence at the University of California, Irvine. The endowed chair was established following a $2 million gift to the university from SAP SE, a German supplier of enterprise applications and business artificial intelligence software, and was named after SAP’s co-founder.
The event, held at the Beall Innovation Center on the UC Irvine campus, was hosted by Marios Papaefthymiou, Dean of the Donald Bren School of Information and Computer Sciences (ICS). Dean Papaefthymiou welcomed leadership from SAP’s offices in Newport Beach as well as Silicon Valley and Europe; and members of the UC Irvine community including faculty, staff, students and alumni.
“I had the privilege of meeting Hasso Plattner in person in September 2019, during the grand opening of SAP’s offices in Newport Beach,” Dean Papaefthymiou commented. “A striking statement he made during his public remarks was that one of the reasons that SAP opened its office in Newport Beach was proximity to the UC Irvine campus and the excellent research program in artificial intelligence.”
“[Hasso Plattner] embodies many of the characteristics we strive to instill in our students in ICS: a spirit of entrepreneurship and a determination to succeed through vision, passion and hard work.”
“Hasso Plattner has established SAP as one of the leading technology companies in the world and the most valuable company in Europe – a remarkable achievement. He embodies many of the characteristics we strive to instill in our students in ICS: a spirit of entrepreneurship and a determination to succeed through vision, passion and hard work. Having this endowed chair in our school is not only an honor, but it’s also an inspiration for our students – something we’re particularly grateful for.”
Professor Smyth began his lecture by advising caution about trying to predict the future of AI, an endeavor which is often incorrect and overpromised. Instead he preferred to focus on how we got to where we are and the current reality of AI, in order to learn from the past, responsibly manage the present and optimize the future. Smyth grouped the advent of AI into three waves.
Foundational Concepts
1854
- George Boole developed Boolean algebra, using a binary system (0s and 1s) to represent true and false values. This logic became the foundation of digital circuits and computer programming, enabling decision-making and data manipulation within systems.
1906
- Nobel Prize recipients Santiago Ramón y Cajal and Camillo Golgiwon were awarded the Nobel Prize for “essentially figuring out that in the brain, we had all these neurons talking with each other, communicating. This generated a lot of excitement.”
1940s
- Alan Turing, John von Neumann and Claude Shannon, “legendary people in the field founding theories of computation.”
- Warren McCulloch and Walter Pitts wrote a landmark paper in 1948 called “The Statistical Organization of Nervous Activity” in Biometrics, which fused the two threads of research. “This is actually a paper, if you’re a student, worth going back and reading,” said Smyth. “It’s amazingly well written, really fundamental stuff.”
- The Hixon Symposium at Caltech in 1948, on “Cerebral Mechanisms and Behavior,” was attended by McCulloch and von Neumann, where Von Neumann’s talk on automata attracted some skepticism: “…if you’re a grad student, don’t get discouraged. Even von Neumann couldn’t convince people that he was on the right track.”
First Wave – 1950s to 1970s: Early Efforts
- John McCarthy, who attended the Caltech 1948 Symposium as a student, is credited with coining the phrase “artificial intelligence” when he organized the Dartmouth Summer Research Project on Artificial Intelligence in 1956.
- The late 1950s and early 1960s saw excitement in AI research, with great hopes for systems such as the perceptron learning machines developed by Frank Rosenblatt – but interest was waning by the 1970s as real-world results failed to meet hyped expectations”
Second Wave – 1980s to 2010 – Rule-based Systems and Neural Networks
- Rule-based reasoning systems attracted academic and commercial attention in AI in the ’80s, but turned out to be difficult to manually develop at large scales – leading to renewed interest in systems that could learn directly from data.
- The International Conference on Machine Learning (ICML) was established in the United States in 1980. UC Irvine was heavily involved in machine learning very early and hosted the ICML in 1987.
- The first issue of the Journal of Machine Learning was edited by Pat Langley, who was a UC Irvine professor at the time.
- David Aha, then a PhD student at UC Irvine, created an FTP archive, which was widely used by students, educators and researchers. It still exists today as the UC Irvine Machine Learning Repository.
- The Neural Information Processing Conference (NeurIPS) began in 1987, chaired by two Caltech professors.
- The Netflix Prize, a competition to improve algorithms for automated recommendations, ran from 2006-2009 and led to new academic interest in large-scale data sets.
- The Imagenet dataset and competition in the late 2000s, with 10 million images, developed by Alex Berg (UC Irvine) and colleagues, was another landmark large dataset in machine learning.
Third Wave – 2010 to present – Deep Learning Comes to the Forefront
- Deep neural networks (a.k.a. deep learning) trained on very large datasets quickly led to significantly more accurate machine learning models.
- They were applied commercially in the mid 2010s in areas such as speech recognition and machine translation, and form the basis of modern large language models.
- They are also widely used in diverse areas such as biomedical image analysis, climate modeling and software engineering.
As well as the obvious challenges with AI, Smyth used examples to show how both AI and humans are often overconfident in AI’s ability to solve complex problems, failing to recognize the limitations of the data sets available and the systems’ inability to follow common sense principles. He also warned of the dangers in educational settings of the misuse of AI having the potential to diminish students’ ability to interact directly with the world. He concluded that AI is extremely useful but it’s important to be aware of its limitations and resist the hype.
“Remember that history…. AI is broader than deep learning, and you owe it to yourselves to educate yourselves more broadly about other aspects of AI and think also not just about the technical aspects but the societal implications,” Smyth said. “We shouldn’t just be asking can we do this with AI, but why are we doing this with AI. What is the long-term goal here?”
“We shouldn’t just be asking can we do this with AI, but why are we doing this with AI.”
As for future predictions, Smyth emphasized: “There’s a lot of work we need to do.” He said that researchers are addressing challenges with human-AI collaboration, adding symbolic knowledge, user studies and more.
To conclude the event, SAP Chief Technology Officer Philipp Herzig presented the Hasso Plattner Endowed Chair in Artificial Intelligence medal to Smyth. “We are very excited not only about this endowment, but also to really deepen our partnership with UCI,” said Herzig. “This appointment marks not just an academic honor. It’s a statement about the future that we want to build together.”
“This appointment marks not just an academic honor. It’s a statement about the future that we want to build together.”
The full lecture is available to view below and on the ICS Youtube Channel.
Additional Links: