Keynote Lecture: Been Kim

Wednesday, August 7, 11:45 am - 12:45 pm, Kresge Hall

Enabling human learning from machines: creating machines that work for us

Been KimBeen Kim, Google DeepMind

Senior Staff Research Scientist

Our relationship with artificial intelligence is rapidly evolving. As machines demonstrate increasing capabilities, surpassing human performance e.g., Go and Chess, the question of how we communicate with them takes on new urgency. In this talk, we will discuss communication between humans and machines as the main means to make this relationship ‘work’ – to benefit from them. While current AI’s natural language interface provides an illusional feeling of communication, it is widely known that we can manipulate outputs through seemingly insignificant changes in input, highlighting the fragility and potential for misinterpretation in such interactions. This talk advocates for a shift in perspective, proposing to bridge the communication gap by actively expanding our own knowledge and capabilities. Doing so enables us to also expand what machines know – by building better aligned machines. I share why considering the communication gap is crucial and provide an example of bridging the gap.