It's not what we think, but how we thought it that fascinates Aaron Baughman. His interest in how the human brain operates began as a boy, playing video games that appeared to think and reason. “I’d build this virtual civilization as the computer created complex economic or military strategies,” he says. “It was amazing. My pursuit of computer science, and of understanding our own brain, probably began right there.”
In college (which included two summers as an IBM intern), Aaron explored the intersection of computing and neuroscience, such as projects to learn how people recover from strokes or manage Parkinson’s disease. This evolved to his focus on neural models applied to computing, which would shape his career. As an IBM consultant, he helped government clients develop biometric systems, and later to adapt IBM Research’s Jeopardy! system for their needs. Since then, his application of technology to sports and entertainment is responsible for high-profile IBM projects such as ESPN Fantasy Football, and generative AI systems to support the Grammy Awards, the U.S. Open, Wimbledon and the Masters Tournament. These combine natural language processing, text summarization, machine learning, data mining, deep learning and computer vision to create new capabilities for broadcasters, reaching fans worldwide and earning an Emmy Award last year for Aaron, his teams and IBM.
“It’s fun to see our work reach so many people through sports broadcasts or fantasy football, but it also prompts many to consider how these machines might augment how they experience the world in other contexts," Aaron says. "If we can predict how a football player will perform, perhaps we can help retailers predict what products will be in demand. Our work can and does influence healthcare, finance and much more.“
As an IBM Fellow, Aaron will lead the design and implementation of AI techniques within the Business Transformation Services portfolio and collaborate across IBM, while continuing to lead the development of AI assets for IBM Live Events. He credits mentors, family and friends for propelling his work—and although he’s an INFORMS Franz Edelman Laureate with 360 patents to his credit, Aaron believes there’s much more to accomplish.
“The speed of innovation is unreal, and the growing availability of AI systems is creating opportunities I couldn’t have imagined even 5 years ago," he says. "I once played games and wondered how they worked, but today my daughter in fifth grade is using AI within a school project and her younger sister understands the concepts too. It’s fascinating to watch all of this and an honor to be a part of it.”
Volunteering helped Kush Varshney expand his vision of technology helping society, and focus his work on helping AI systems become safe, trustworthy and transparent.
“When I was in college, machine learning was still an abstract mathematical problem that struggled to make meaningful predictions in messy, real-world contexts,” he says. “I joined IBM Research in part because it was working on people-related problems, which hadn’t even crossed my mind as a use for machine learning. Then in 2013, I volunteered with an organization that connects data scientists with social change organizations. We helped one nonprofit identify villages in Kenya and Uganda most likely to benefit from support. We helped another that installs solar panels on homes in parts of India where the power grid isn’t reliable. This experience started bringing everything together for me: fairness, transparency, trust and social good.”
Kush shared his volunteer impact with IBM colleagues, which led to his co-founding IBM’s Science for Social Good initiative in 2016. It applies data science to address hunger, poverty, health and inequality through partnerships with social enterprises. Kush also became a pioneer in safe, trustworthy and socially responsible machine learning and artificial intelligence, publishing seminal papers and the first book on the topic.
“My early IBM projects were related to human capital management and healthcare, and they were revealing because it became clear that accurate predictions aren’t enough," Kush says. "These systems have consequences on people’s lives, so there must be consideration for fairness, and our models need to be understandable. This is true not just for social enterprises, but for all of IBM’s clients as we develop models that everyone can have confidence in.”
Pursuing that goal, Kush led development of the open-source toolkits AI Fairness 360, AI Explainability 360, and Uncertainty Quantification 360. He also invented algorithms for mitigating unwanted bias, and created factsheets for AI development lifecycle transparency. These are helping IBM clients across sectors to make systems more trustworthy using watsonx.governance.
“I believe we still need to incorporate a greater understanding of moral psychology and ethics," Kush says. "What are the human values we want these systems to express, how do we empower communities to do so, and where does that knowledge reside? It’s in laws, but also in folklore and fables, and all sorts of places. That’s not easy to encode in a computer model. There’s a lot of work needed, but there’s also a great team of smart and caring people I’m privileged to share that effort with at IBM."
As IBM Fellow for AI Governance, Kush will focus on new approaches to mitigating harms and to safely governing large language models, as well as their incorporation into the watsonx platform and other AI-infused products.
Trent Gray-Donald says he’s something like a plumber. “I’m the builder, someone who knows where all the pipes and wiring should be and how it all operates. I can work with anyone to create systems that are robust and scalable, to make sure the architecture is right and deeply understood. I help transform inventions into successful products."
Trent built this capability on a lifelong, irresistible need to know how things work, down to the last detail. After university, he spent 16 years building language runtimes, especially the IBM J9 JVM. “Working in assembly language, building a JIT compiler target, debugging race conditions, all of it taught me how computers actually function near the metal. It was a fundamental realization that everything’s explainable, and there’s no magic in computing."
Trent’s career shifted to data and AI at IBM, where his deep expertise and holistic approach to every challenge led to a series of achievements. He led the evolution of IBM’s diverse Data and AI product portfolio into a leading platform, both as software and SaaS. He is the lead architect and strategist behind IBM’s approach to Foundation Models for code across Ansible, COBOL, Java and general-purpose programming (including WCA@IBM). Trent was also responsible for the overall architecture and implementation of Watsonx Code Assistant, Data Fabric, Watson OpenScale, and IBM Analytics Engine—and his platform vision ties Cloud Pak for Data together.
In his new role as chief architect of Watsonx Code Assistant, Trent will lead the evolution of products that employ generative AI to help IT teams create high-quality code. The strategy is to build on foundational models to create assistants in specific areas where IBM has specialized expertise, such as COBOL and Ansible. IBM is also working with financial clients who created their own programming languages to develop tools for those unique requirements.
His advice to other technical leaders is to establish foundational skills, then proceed in a spirit of cooperation. “I’m always learning something, but always teaching something at the same time," Trent says. "I’m willing to say I don’t understand this, so let’s all figure out what’s really needed and what steps will move us from here to there. I don’t think you need to be aggressive to succeed. It’s better to build relationships, to influence and guide without asserting absolute control. You want to be someone who helps problems go away rather than make them appear.”