April 9, 1970

Dr. Myron Tribus
Aesistant Secretary of Commerce
Washington, D. C.   20230

Dear Myron:

Thank you for your letters of March 31 and April 1.
Our correspondence has so my rich issues in it that it is bound to be
somewhat disorderly.  If I wait to take up every question that you ask or
that you provoke I will never get around to answering your letters at all,
and that would be tragic.

I will, then have to include in one letter a set of very diverse questions,
and hope that you also do not feel compelled to deal with all of them bebore
answering any.

As to fluoridation (your letter of April 1) my quggestion about fluoride weeks
was partly a serious proposal - but really a too gimmicky one; more importantly
a satire, or, if you like, thought-experimelt on 4he question. I am not the
only critic who regrets the use of satire now and then, and I am in fact col-
lecting a file of maa culpa8 of my journalistic colleagues who publish such
regrets now and again.  I did, however, think that exposing the situation in
this way would bring most people to realite that they wuld not take any effort
to protect themselves from a supplement of fluoride if it were delivered In the
way I indicated.  On the other hand, the epidemiological work that has been done
on chronic fluoride exposure is probably about as,good as can be managed at the
present time, which means that it is basically inadequate to meet the demands
of the situation.

Thank you for your comments on DENBRAL (your March 31). We have explored and
often used a number of approaches to introduce efficient tree-pruning, and I
aspect that our write-up is somewhat deficient In announcing all of them. We
are quite sensitive to the futility of the comfpletely deterministic enumeration
of the possibilitiee for larger molecules, and have explored a number of strategies
that would at least attempt to match the performance of the human heuristics in
this field.  We do reach the point where human capability is equally frustrated,
and have been following, mainly, two approaches.

(1) To classify the domain8 of confusion so that partial solutions can be
expressed in terms of meaningful genera.  For example, it can be anticipated
that mass spectrometry will have a difficult, if not impossible, job of dlsccim-
inating between isomeric alkyl chains if these are at all large. The search
strategy can be simplified if we then extract the concept "alkylt', deal with it
as a superatom in the system,
as a later refinment,     and leave the issue of disttngufshing among alkyls
           to be pursued only if the partial solution is unacceptable
and if there is some expectation that the further resolution is possible.


Dr. Myron Tribus

April 9, 1970
Page 2

(2) We also anticipate the need to call in other analytical approaches - for
example, NMR, or IR.  we have a paper in press now that illustrate8 the enomouB
truncation of effort that is possible when such auxiliary information is provided.

While these steps are quite essential for the efficient operation of the DENDRAL
program as a chemical assistant, I have peraronally become rather bored with the
effort of further manual reffning of the system.  I am instead trying to push
for the construction of an executive program that will make Its own tentative
judgments about the organization of the Search strategy and itself look for,
identify, aud evaluate all of the kinds of proposals fof improving the program
that we have been diacuesing.  This is a meta-DENDRAL of a somewhat different
character than the program that will attempt to amplify the chemical theory that
                                                                  --
is at the Center of the hypothesis-evaluator.

1 feel quite strongly that we have gone almost to the end of the rope la flrat-
level programming, and that we have to learn vays of dealing with subroutines as
hypOtheSe8 that need to be induced by the master program itself, perhaps roughly
in the same way as hypothetical organic structures are Induced by DENDRAL.

I also appreciate your note about the more systematic use of chemical Stability
(free energy) and I can only answer your question with the apology that we have
simply nev%r gotten around to ueing it more 8ystematfcally, although racognizlng
the advantage of it.  I will look for the Souders Tabler that you mentioned, and
it might indeed be quite useful to us.  We have used this approach in a rather
ad hoc way in a few places, but have never put.together the costplete table in a
way that wuld make good meuse.

Sincerely yours,

Joshua Lederberg
&OfeSSOr of Genetics