Sir

Several graduate students would like to suggest a modest proposal to the authors of the recent article “Functional genomic hypothesis generation and experimentation by a robot scientist” (Nature 427, 247–252; 2004).

We've been doing a little research of our own and have generated an exciting hypothesis: we can build a Virtual Principal Investigator.

The software platform will consist of a tiny amount of background knowledge about a completely different system from the one being actively investigated, a buzzword-invention engine, coffee-consumption apparatus, a grant-writing node and a cat-o'-nine-tails that will tie everything together.

For much less money than an actual investigator, we created a machine that resides in an office, drinking coffee at a steady rate. At bi-monthly intervals, the Virtual Principal Investigator spits out grants that are derivatives of highly successful grants from the past. Between these time points, it will play an endless loop of inspirational messages and test the effectiveness of the whip as a method of controlling subservient organisms.

Initial tests of our Virtual Principal Investigator have revealed no differences from our existing model of an investigative scientist, but we have yet to perform an experiment that was not naive or random. We look forward to discussing our results with you.

Several non-robot scientists (names supplied)