Proposal for the Dartmouth Summer Research Project on Artificial Intelligence
John McCarthy presents his proposal for a summer research project on artificial intelligence, aiming to convince his peers to collaborate and secure funding for what would become the birth of AI as a
Setting
A wood-paneled conference room on the second floor of Dartmouth College's mathematics department, with a large rectangular table at its center and tall windows overlooking the campus green.
Characters
John McCarthy
primary
A lean, wiry man in his late twenties with sharp features, a high forehead, and dark brown hair combed neatly to the side. His piercing eyes behind round wire-frame glasses convey intense focus. He has a slight forward-leaning posture that suggests intellectual engagement.
Marvin Minsky
primary
A wiry, intense man in his late 20s with sharp features, thick dark hair combed back without a part, and penetrating eyes behind round-framed glasses. His posture suggests restless energy barely contained.
Nathaniel Rochester
secondary
A lean man in his mid-30s with sharp facial features, wire-rimmed glasses, and neatly combed brown hair. His posture suggests a man accustomed to precision, with long fingers that frequently tap against surfaces when deep in thought.
Claude Shannon
secondary
A lean, middle-aged man with sharp features, thick-rimmed glasses, and neatly combed dark hair showing the first signs of gray at the temples. His piercing eyes convey both curiosity and analytical precision.
Graduate Assistant
background
A young man in his mid-20s with a lean build and short, neatly combed brown hair. His wire-rimmed glasses perch on a narrow nose, and his clean-shaven face shows focused concentration. His posture suggests both academic rigor and deference to senior scholars.
Dialog
John McCarthy
Therefore, if we can simulate every aspect of learning or any other feature of intelligence—consequently, we ought to be able to construct a machine to do so.
Marvin Minsky
But doesn't that presuppose intelligence is reducible to symbol manipulation? What if we're dealing with emergent properties like—
Nathaniel Rochester
No—wait, let me put it this way: Your 'thinking machine' would need more memory than our entire 704 installation. We're bottlenecked at 4K words here.
John McCarthy
Precisely why we must separate the theoretical framework from current hardware limitations—unless you believe computation won't advance beyond vacuum tubes?
Marvin Minsky
Forget tubes versus transistors—what's your operational definition of 'learning' here? If it's just curve-fitting, we're back to Wiener's cybernetics.
John McCarthy
I propose we define it recursively: any process where performance improves through exposure to data patterns—though naturally, the devil's in the recursion.
Nathaniel Rochester
Devil's in the core dumps, more like. But fine—if we can get IBM to spring for drum storage, I'll code your heuristic search.