Please xerocopy for J.L.
3/20/73
Dear Tom -- SUMEX S.V. lessons
We have still not found and justified a credible game for our robot to play. Yet as I tried to tell Klaus the brain needs a hand as well as an eye to promote its own growth in a reality-oriented context. I should articulate more clearly to the study section (1) this orientation of our effort and (2) that there are still also prospects of early biomedical utility. Perceptions (e.g. speech recognition may be a better model of successive, reciprocal refinements of hypothesis and data acquired. There is certainly much more simulation that could be set up within DENDRAL to justify this approach -- e.g. "intelligent" search of the dictionaries of solved subproblems which is tantamount to a dynamic manipulation of GOODLIST oriented by data reference.
So whatever the outcome of SUMEX, there is still a lot of introspective research that can be done, the "real time" instrument being a tape drive.
[END PAGE ONE]
[BEGIN PAGE TWO]
But I would like to see us put more imagination into building "hands." I think NSF would be receptive to proposals for new kinds of instruments; and if we have a lull now we should be thinking of more definitive pre-designs for digitally controllable mass spectrometers. These will certainly entail new approaches to internal calibration. I wonder about trying to take something like PFK, seek PFK plus plus, and filter out the near-integral masses (or the PFK plus one that would interfere with a given sample prob [?]) with a low resolution m.s. that gates the reference ions into the beam
For metastable work we need to learn how to improve m plus yields at low ionization voltages; or alternatively to de-excite the excited states (collisions with ?)
[END PAGE TWO]
[BEGIN PAGE THREE]
Tom -- let's have a session soon say Monday 2 30 to extract lessions [sic] and positive directions from the SUMEX S.V.
(1) What is most chemistry we can do with minimum new hardware e.g. in re Genetics Res. Ctr.
(2) (Simulation mode of real time Dendral) [note to the right, i.e. external memory simulations instrument]
(3) (Better part A Effects) [note to the right, plans for]
[line to, Financial problem?]
(4) hooking onto other PDPIOS on net.
(5) new instrument ideas on ms
controllable machine
photoimage recording
continuous internal calibration for HRMS
may need "intelligent control" for gating reference gas
2-D metastable scans
controlling energy of excitation laser inputs in flight? consult Gilda
(6) OTHER Parameters; OTHER AREAS of analysis that might better exercise our ideas of Real Time AI. (What is a more exciting instrument than GC/ms?)
(over)         Josh.
[END PAGE THREE]
[BEGIN PAGE FOUR]
whoever can come of the server level
ECL     EAF     TR
Bruce B
Dennis
Carl Dj. (opt. for him but urge [?] him if he has some definite ideas
Alan D.
also we must schedule a SIGBIG soonest for general briefing. Friday 23d. at noon [mark to note, ask Ruth to verify whether my AR date is 3/23 or 3/27] or at 4? or 3/27 or 3/28.
Can Dennis arrange a demo of DENDRAL (any part) for lecture 3/29 8 05 pm at or near Rm 20 University? (Lois Durham ext 4480 solicited it)