May 11, 1965

Mr. D. D. Tunnicliff
Analytical Department
Shell Development Co.
Emeryville, California

Dear Mr. Tunnicliff:

My thanks for your letter of 28 April, which gives me a much clearer picture
of the standing of your program in dealing with some of the woolier situations.
I have to agree that no precise solution may be imaginable for some situations,
and I had in mind not only the possibility of essentially identical spectra,
but perhaps less easy to anticipate, cases where different combinations of
spectra might be expected to yield substantially the same sums. Whatever the
hypothetical possibilities of the occurrences of these pitfalls, some practical
experience would be the surest guide to evaluating their importance, and it is
for Just this reason that I included in my review comment an opinion that your
further discussion, which was elaborated in the letter, would be of value to
just those readers who would be particularly interested in the subject at hand.
It would of course be quite impractical to consider every possible combination
of the items in a large library of spectra, but we might imagine the development
of some heuristics, at least to find the most comfusing potential combinations
of spectra that would be virtually certain to lead to an unresolved ambiguity.

We would like to try mounting your program on our 7090 under IBSYS. Am I asking
a great trouble on your part to furnish a card deck or a tape of the FORTRAN
program for us to work over for this purpose, or Has the revision already been
made? We would of course repay any costs involved in preparing a copy. Your
writeup seems quite clear and I hope I would not have to burden you with too
many questions.

The utility of the program for our immediate purposes—-evaluation of amino acid
mixtures, partially resolved by GLC-~is somewhat speculative, but I feel the
experience would be instructtre.

Many thanks,

Joshua Lederberg
Professor of Genetics

Encl: DENDRAL I
dL 120
Artificial Intelligence Project, Memo No. 30 (Feigenbaum & Watson)

aie! AdIT DIANA