Stuffilike.net

The Singularity Started With the Wheel March 16, 2015

As difficult as it may be to comprehend, the wheel is the basic unit of technology. It made the repetitive business of carrying stuff easier. When tasks can be easily repeated (preferably automated), they can also be tweaked to do them better, maybe each time.

With computer controls, these tweaking steps can be automated, and the results don’t even have to be seen by a human.  These results can be used to produce new methods to experiment, ad infinitum. This is precisely why we should not allow AIs any autonomy whatever in creating new AIs.

But I digress.

The tools of automation are now cheaply available, giving everyone who wants it access to finely-controlled stepper motors which can be used in the trial and error methods heretofore mentioned.  Cheap microcontroller systems to run them combined with said stepper motors give us robotic assemblers, 3D printers and molecular assemblers.

Yeah, you heard me.

Usually, small-molecule synthesis usually relies on procedures that are highly customized for each target. Martin Burke, a Howard Hughes Medical Institute (HHMI) early career scientist at the University of Illinois at Urbana-Champaign, used a single, fully-automated process to synthesize fourteen distinct classes of small molecules from a common set of building blocks.

A broadly applicable automated process could greatly increase the accessibility of [this class of compounds] to enable investigations of their practical potential. More broadly, these findings illuminate an actionable roadmap to a more general and automated approach for small-molecule synthesis (he used Csp3-rich polycyclic natural product frameworks and developed a catch-and-release chromatographic purification method).

As a former chemist, I must say this is plenty difficult and detailed…but it only has to be done once and this genie is not going back into the bottle.  This will step up the pace of novel moiety experimentation, especially now that we have computational chemistry on a sound footing.  Picture this: computer cranks out theoretical molecule families for research.  Magic chemistry machine makes them.  Another automated machine tests them.  Potential drug candidates can be screened without human intervention, for conditions that currently have no treatment, but do have a good theoretical model.

Honestly, I have been thinking of this for thirty-five years, when one of my classmates described the room-temperature chemistry that was just being used for automated peptide synthesis, a hot subject in my college years¹.

Now, with  automated synthesis producing testable quantities of continuously-varying drugs, we can start continuously comparing them with standard drugs for, say, antibiotic activity in a Petri-like environment (I hope it is no surprise that this technology exists already, although it is not in concert with the aforementioned molecular assembler), quickly finding optimal candidates in what could be an entirely automated process.  Promising candidates’ structures can be continuously varied by the molecular assembler under the watchful eye of an expert system (it is fun to imagine the expert system eventually deciding that chlorine bleach is the optimal antibiotic; obviously safety trials against mammalian cell lines need to run in parallel).

Aha, I hear you cry, what about diagnosis?  I’m pretty sure I covered this already², when I  talked about brute-force cracking the human medical condition through big data: thousands of tests administered cheaply, regularly through millions of peoples’ lifetimes.  This data would be trawled for correlations between medical conditions and test results, telling us things clinicians would miss just because human heads can’t hold that kind of data well enough to draw statistical conclusions, or even reasonable inferences…but computers can.  Frustratingly, the legal problems here are beyond human comprehension as well; the intellectual property costs to create this many tests would be astronomical, although once acquired it could be quite cheap to administer (this is already possible, just not done for greed’s sake).  This will require a revolution in thinking which is not, alas, forthcoming soon³.

Other science can be brute-forced in a similar fashion by automation in other chemical reactions; I picked drug discovery for illustration since that’s where the most money can be found currently.

These are delightful speculations and become even more possible as long as things continue the way they are going, at least in terms of physical possibility.  Cheaper, faster processors make it possible to control all manner of laboratory  and industrial devices, not just your toaster, son.

It all makes me wish I were a better writer, because these ideas deserve better advocacy than I can bring to bear.

Homework:

Synthesis of many different types of organic small molecules using one automated process Junqi Li, Steven G. Ballmer, Eric P. Gillis, Seiko Fujii, Michael J. Schmidt, Andrea M. E. Palazzolo, Jonathan W. Lehmann, Greg F. Morehouse, Martin D. Burke

EDIT: Somebody did it: “As proof of concept, we have demonstrated its capacity by optimizing the lycopene biosynthetic pathway. This fully-automated robotic platform, BioAutomata, evaluates less than 1% of possible variants while outperforming random screening by 77%. “–Mohammad HamediRad et al. Towards a fully automated algorithm driven platform for biosystems design, Nature Communications (2019). DOI: 10.1038/s41467-019-13189-z

______

¹ We’re getting there, fellas.  Keep up the good work.

² Please try to keep up.

³ If ever.