Last week CSAIL hosted the first “Hot Topics in Computing” speaker series, a new monthly forum where computing experts hold discussions with community members on various topics in the computer science field.
MIT professor Josh Tenenbaum gave the first talk of the series on artificial intelligence (AI) and common-sense reasoning. During his presentation he discussed the importance of creating AI systems that can learn quickly, how intelligence is more than just pattern recognition, and why the field of AI needs algorithms for common-sense reasoning. He proposed an approach using simulation as an engine for scene understanding, where we can begin to understand how to teach machines to learn like children. He further detailed that with a model-based system, we can advance one-shot learning and help machines make better predictions, while emphasizing the idea of learning from small data.
After his remarks, the audience debated the engineering difficulties in the realm of AI and flexible systems, the role of data and models, and the use of probabilistic programming tools. There was discussion about the capabilities and limitations of deep learning engines, differences in model-based and neural network approaches to pattern recognition, and which applications and domains are amenable to various AI tools.
In the future, the forum will focus on a different topic on a monthly basis, featuring a short presentation from an expert in that field, and a moderated fireside chat followed by an open discussion with the wider community.
“'Hot Topics in Computing’ represents a chance for our community to hear from and debate with our experts in computing,” says Daniela Rus, Andrew and Erna Viterbi Professor of Electrical Engineering and Computer Science at MIT. “This is an opportunity to engage in a conversation about what is the potential, what is actually achievable, and what problems need to be solved to get to the holy grail.”