Editor’s note: This is part of a series of virtual visits with NCSA thought leaders on current topics impacting the field of high-performance computing.
We’re in an exciting time of AI advancement and NCSA is poised to lead the way. With faster and more powerful computing than ever before, it’s now easier to experiment with creative and innovative ideas. This presents windows to unforeseen possibilities and opens the doors to reaching them.
Bill Gropp, NCSA Director
In the annals of human history, no technological advancement has garnered as much awe and apprehension as artificial intelligence (AI). Over the past few decades, this cutting-edge field has grown exponentially, revolutionizing industries, altering societal dynamics, and challenging our perception of what is truly possible. AI, the science of imbuing machines with cognitive abilities akin to human intelligence, has made strides that were once confined to the realm of science fiction, and its capabilities continue to evolve at a remarkable pace.
Those aren’t my words, but rather those of ChatGPT, an AI chatbot developed by OpenAI and launched almost a year ago, responding to my prompt – and I don’t happen to agree with everything it said. But surely you’ve heard of it. While ChatGPT is currently one of the more prominent applications of AI, it is but one in a long list of uses and interpretations that catch the attention of the public and become markers in pop culture history.
One of those examples has a fictional origin story that takes place at the University of Illinois – the very place NCSA calls home. In the film “2001: A Space Odyssey,” HAL9000 is depicted as a sentient artificial intelligence computer that controls functions of a spacecraft and communicates with the crew in a monotone, deadpan and human-like voice. HAL remains one of the most famous portrayals of AI, but don’t worry, you won’t find any spoilers here – even if the film and novel came out more than 50 years ago.
And that’s not where the connection between artificial intelligence and our campus ends. More recently, NCSA received a $10 million award from the National Science Foundation to deploy DeltaAI, an advanced computing and data resource that will be a companion system to Delta. DeltaAI will triple our AI-focused computing capacity and greatly expand those available within the NSF-funded advanced computing ecosystem. This system will provide powerful capabilities for simulation and data science with a strong emphasis on support for AI, which is in growing demand across many fields of science and engineering. Utilizing artificial intelligence in advanced computing is a priority for the United States – in both the research community and government – and NCSA is thrilled about this expansion in our ability to support AI research.
In 2019, we launched the Center for Artificial Intelligence Innovation (CAII) to bring together faculty and industry expertise in AI to create new ways it can be used in big-data research, industry and technology. In March 2020, as a member of the C3.ai Digital Transformation Institute, NCSA helped award $5.4 million to accelerate AI research to mitigate the COVID-19 pandemic. Last year, we unveiled our new HOLL-I (Highly Optimized Logical Learning Instrument) system targeted at training large AI models on extreme-scale datasets.
In the short run, we’ll see AI applied in innovative ways to more applications. We’ll also see a greater emphasis on addressing the limitations of current AI, such as bias in the training and thus in the output, and what are charmingly called “hallucinations” – plausible, compelling and fictitious output. One surprising feature of AI is how powerful it is even though the underlying system has little comprehension (as we would understand it) of the world around it. We see the challenges that AIs face when confronted with something beyond their training, which is particularly apparent in autonomous vehicles.
Make no mistake, AI is a powerful tool and it has and will continue to transform everything to which it is applied. But, like all tools, it has its limitations and understanding those limitations and working within them will be key in the successful application of AI. It is unknown whether the current approaches to AI can solve these problems. New methods, or combinations of methods, may be required. This is why research in AI is so important. Research that not only will give us “explainable AI,” but which will help us understand the limitations of the current approaches and point us in the direction to make AI more fair, effective and trustworthy.