Platt Perspective on Business and Technology

Moore’s law, software design lock-in, and the constraints faced when evolving artificial intelligence 5

This is my 5th 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 1-4.)

I began discussing the basic, core issues of this series in its first four installments, in terms of two working examples: one drawn from human-developed technology and its development and evolution, and the other drawn from biological evolution with its vastly longer timeframes. Then at the end of Part 4 I finally turned to the issues of artificial intelligence systems, and to the still illusive but compelling goal of defining and developing a true artificial general intelligence – from where we are now in that ongoing process where no one has even effectively, functionally defined what general intelligence per se is.

I would in fact take what might be considered a contrarian approach to thinking about and understanding that challenge, turning back to my human technology example of the preceding installments of this series, as a much simpler comparative example, as a starting point for what is to follow here.

• Current, as of this writing artificial intelligence systems designers and programmers are steadily learning more and more about simpler, artificial single function intelligence systems, and how to develop and improve on them and their defining algorithms and supportive database subsystems.
• And much of what is learned there from that effort, will likely prove of use when developing actual, working artificial general intelligence systems and capabilities, much as human brains with their (arguably) general intelligence capabilities are functionally and structurally comprised of large and even vast numbers of simpler, single specialty-function components – biologically evolved counterparts in principle at least, to what we are developing and using now in our simpler single algorithm-based artificial intelligence systems.
• The fact that we know that we do not know yet, how to take that next step big leap to artificial general intelligence systems, and that we see and understand how limited our current simpler artificial intelligence systems are, even as we improve them, keeps us from prematurely locking in the wrong development decisions in a premature artificial general intelligence initiative, with their certain-to-emerge consequences that could become baked into any further development effort that might be made, in implementing a “prematurely understood” general intelligence paradigm.
• In my simpler technology example, digital musical note encoding has become fixed in place with MIDI as a defining protocol in place and for large areas of digital music as a whole. And the computer programmers and protocol designers who developed this coding protocol and who promoted it into become first A, and then The digital music standard, did not know the limits to what they knew when they did that. In particular, they did not fully enough know or understand world music: music from non-Western sources and heritages that cannot be parsed into the same notes as organized along the same scales, that would for example be found in the more traditional Western European and American music that they did know.
• At least as setting a largely fixed industry-wide standard is concerned, MIDI’s developers and promoters did act prematurely.
• At least up to now, artificial intelligence systems developers have remained more open minded, and certainly as far as achieving and implementing artificial general intelligence is concerned and as such have not built in the at least categorically similar type of presumptions that have led to the MIDI that we have today.

I wrote in Part 3 of adaptive peak models as are used to represent the evolutionarily competitive relative fitness of differing biologically evolved or technologically developed options. Think of MIDI, as discussed here as a highest point possibility at the peak of a less than highest possible “mountain” in a potentially larger adaptive landscape. Premature decision making and lock-in led to that. So far artificial intelligence systems development, or at least the race to develop true artificial general intelligence has not fallen into that trap.

This does not mean that real, sustained effort should not be made to functionally, operationally define and understand artificial general intelligence, or to attempt to build actual hardware and software-based systems that would implement that. It does mean that real effort should be made to avoid locking in as if axiomatically so, premature technology development assumptions or the short-term solutions that they would lead to, as if they held long-term in value and even permanently so.

I continue this narrative with that as of now benevolent openness and uncertainty in mind, and as a still as-of-now positive virtue. And I do so starting with a set of distinctions as to how smart and connected technology can be divided into four categories for how they connect to the human experience for their input and output functionalities and for their use of big data, as developed by David Rose and as offered in:

• Rose, D. (2014) Enchanted Objects: design, human desire and the internet of things. Scribner.

Rose parses out such technological possibilities into four possible futures as he envisions them, with each predicated on the assumption that one of four basic approaches would be built into the standard implementations of these artifactual capabilities moving forward, which he identifies as:

1. Terminal world, in which most or even essentially all human/artificial intelligence agent interactions take place through the “glass slabs and painted pixels” of smart phone and other device interfaces.
2. Prosthetics, in which a major thrust of this technology development is predicated upon human improvement.
3. Animism, and the emergence of artificial intelligence ubiquity through the development and distribution of seemingly endless numbers of smart robotic nodes.
4. And Enchanted Objects, in which the once routine and mundane of our everyday life becomes imbued with amazing new capabilities.

I see tremendous opportunity for positive development in all of these perhaps individually more stereotypic forms, and expect that all would have their roles, and even in a world that pushes the internet of things to its logical conclusion of connected everywhere animism. To be more specific there, even smart terminals that take the form of highly advanced and evolved smart phones would probably play a role there, as personalized connectivity organizers if nothing else – though they would probably not be awkwardly limiting handheld devices of the type we use today when serving in that expanded capacity for us, on the human side of this still just emerging world.

And this brings me back to the challenges of lock-in. What is eventually, and I add inevitably going to become locked in for the what and how of artificial intelligence and artificial general intelligence, will determine which of the above four, and other possibilities might actually arise and take hold – or rather what combinations of them will and in what ways and in what contexts.

• The points that I raise here are going to prove to be fundamentally important as we proceed into the 21st century, and certainly as genuinely widely capable artificial general intelligence is developed and brought into being,
• And even just as our already actively emerging artificial specialized intelligence agents proliferate and evolve.

I am going to continue this discussion in a next series installment, with further discussion of Rose’s alternative futures and how they might arise and contribute to the future that we actually come to experience. And in anticipation of that narrative to come, I will at least begin a discussion of the likely roles that single function artificial specialized intelligence and artificial general intelligence might come to take and particularly as an emerging internet of things comes to redefine both the online world and our computational and communications systems that support it.

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: