i would start with the abstract to one of your papers, then keep iterating over it and simplifying it until you think your target audience can comprehend it, th

i would start with the abstract to one of your papers, then keep iterating over it and simplifying it until you think your target audience can comprehend it, then say that sentence to him/her that way, you stand a chance of preserving at least a part of what you're actually doing rather than stating a hand-wavy "I help computers to think"