Platt Perspective on Business and Technology

Moore’s law, software design lock-in, and the constraints faced when evolving artificial intelligence 4

Posted in business and convergent technologies, reexamining the fundamentals by Timothy Platt on November 25, 2018

This is my 4th posting to a short series on the growth potential and constraints inherent in innovation, as realized as a practical matter (see Reexamining the Fundamentals 2, Section VIII for Parts 1-3.)

I have been discussing two case in point examples of innovation, and of innovation lock-in in this series: one human technological in nature and one drawn from biological evolution as a rich source of example-based insight. My technological example has been how musical notes and their digital encoding have become locked in for acceptable and acknowledged form, through what has become an essentially universal acceptance of and adherence to the MIDI musical note encoding standard. And my biological example is provided by the even more widely adhered to pentose shunt of biochemistry, or the pentose phosphate pathway as it is also called.

When A technological solution becomes The technological solution for a problem or task that it would address, and with no realistically competitive alternatives available to chose from for what it seeks to do, it can become an unquestioned and even an unconsidered axiomatic given: inherent limitations and all. For the MIDI sound encoding example, I cite more avant garde Western composers who add sounds and sound patterns into their works that while precisely specified in their scores, do not fit traditional musical note forms as laid out according to a traditional musical scale. And perhaps more to the point, I cite non-Western composers and musicians who perform on instruments that produce sounds that do not cleanly, accurately fit into a MIDI-defined pattern, rendering original performances into what can become more cartoon-like caricatures. Actual live performance music does tend to keep the possibilities of sound encoding alternatives a real possibility. But more persistent, repeatedly playable recordings of those original performances can and do lose some of their essence in their digital transformation when so note-by-note encoded.

It can be essentially impossible to conceive of a realistic, more highly adaptive alternative to a biologically ancient “evolved technological” solution, and certainly in precise functional detail. And that potential blind spot limitation definitely applies when the universally accepted norm in question is one like the pentose shut: where no single mutationally arrived at alternative to what is currently in place is likely to be as viable as it is and where no two-hit double mutation combination that would offer a possible alternative to it is likely to fare any better.

I wrote of topologically formatted “maps” of natural selection fitness in Part 3, and of topographically represented landscapes that range from high fitness peaks on down to low fitness and even lethally deleterious mutational valleys there. I offer that conceptual modeling approach to visualizing and thinking about adaptive fitness, as applying to both biologically evolved and human technologically developed solutions to problems faced.

Citing this model as elaborated upon in Part 3, when all alternatives to some given positively adaptive peak, lead to adaptive submergence and a loss of functional viability, it becomes hard to not take such a biological solution or technical fix as a best possible solution overall. Considering my biological example in that context, it is, however, likely that there were alternative, if perhaps more complex biochemical pathways competing for primacy in competing evolutionary lines when the pentose shunt first arose and when it was initially evolving. That solution simply won out and with time became “optimized” within the scope and limitations of the particular adaptive peak that it evolved into – and regardless of the at least theoretically higher adaptive peak values that some now-unobtainable alternative might have reached.

• Are there realistic alternatives to this pathway in basic eukaryotic cell-based life that have been so thoroughly precluded as possibilities by universal adherence to this particular biochemical solution, that we cannot even know of their existence, let alone their possible details?

Turning back to the MIDI example where at least alternative functional needs are knowable and where analog recordings are also possible, I would argue that both the strengths and the weaknesses of that solution stem from a commonly shared underlying source. The people who initially devised and standardized MIDI digital encoding, did so while deeply aware of the types of music that always and only include in them, sounds and sound patterns that can be captured and with fidelity as streams of specifically formulated and defined musical notes – of the type that their system digitally encodes. But they also did this with at most a partial awareness of musical forms and patterns, and of musical sounds that do not fit the predefined note patterns that they built into MIDI. This does not mean their system cannot capture such music at all; it does mean that their digital recordings will differ from the original live performances that they seek to capture in persistent digitally replicable form.

And with that, I turn to consider artificial general intelligence. And I begin addressing that by noting that both one of the defining strengths and one of the defining weaknesses of that far-reaching development effort can be captured in a simple to express point of observation:

• We do not know what “general intelligence” actually means, beyond assuming that it is generally flexible in its functional reach and that it would probably contain within it, properties that we commonly at least categorically refer to by names as “sentience” and “sapience.”
• And we tend to think of human intelligence as a benchmark standard for “general intelligence” as a whole, and even when acknowledging in the abstract, at the very least that “general intelligence” might take forms that the human brain and mind would not and even cannot specify.

This opens up a whole range of issues and discussion points, and I will at least begin to address them in my next installment to this series. Meanwhile, you can find this and related material at Ubiquitous Computing and Communications – everywhere all the time 3 and also see Page 1 and Page 2 of that directory. And I also include this in my Reexamining the Fundamentals 2 directory as topics Section VIII. And also see its Page 1.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.

%d bloggers like this: