Platt Perspective on Business and Technology

The fallacy of the singularity and the fallacy of simple linear progression, reconsidered

I have recently found myself thinking back to a posting that I first added to this blog in February, 2010 – just a few months after starting it:

Assumption 6 – The fallacy of the Singularity and the Fallacy of Simple Linear Progression – finding a middle ground.

I offered that posting in my directory: Reexamining the Fundamentals, as an installment in a series of brief notes in which I posed questions and suggested reconsideration of a succession of issues that we can find ourselves taking for granted.

My basic argument in the sixth installment to that series, as cited here, was that aside from astronomical events such as the emergence of black holes from massive supernova explosions, true singularities do not arise and certainly not on Earth or in our range of direct and immediate ongoing experience. And if they do not actually arise in our direct experience, neither does simple, same evolutionary path forward linear progression, except as it might arise in a very time-limited manner. Long-term and certainly unending linear developmental progression as I write of it here, is just as much a simplification of more complex phenomena and just as much of a mirage as are putative singularities in any directly human or human societal contexts.

I specifically cited the book:

• Kurtzweil, Ray. (2005) The Singularity is Near: when humans transcend biology. Penguin Books.

in that earlier posting, for how Kurtzweil predicted an acceleration in the pace of innovation and change, until a true singularity for it would be reached. For purposes of this posting and its narrative, let’s construe singularity there to mean the pace of change and of disruptively novel change accelerating to a point where essentially no one can keep up and no matter how much of a pioneer and earliest, fastest adaptor they would be when positioned on an innovation diffusion theory-based, adaptation curve.

In 2010, we lived in a context where that type of singularity event was all but certain to never arise, and certainly when innovation was being developed and advanced entirely from direct human initiative and by essentially the same people and types of people who would ultimately have to accept and adapt to any given step in this process of change, and buy into it as a part of that. Think there, in terms of businesses that would bring next step innovations to market, having to be financially successful enough from earlier efforts at that, and from their success in the marketplace from that, to be able to afford to design, prototype test and manufacture a next step innovation too. Innovation developers who function as such in a business or enterprise, have to have the resources and the opportunity to develop their next step innovation and the next after it. And this requires that the business that pays for this, be able to afford it and still keep their doors open. That, and certainly according to the logic of my earlier Assumption 6 posting, of necessity breaks down if innovation arrives so quickly to the marketplace that it cannot achieve buy-in for it, and if it is impossible to achieve the necessary consumer support that would drive this innovation cycle.

But even as I wrote that earlier posting, there was at least in principle, a possible way around that anthropocentric, from human to human restraint mechanism on the maximum possible sustainable pace of change. And we might be on the verge of seeing the emergence of the first simple test case proof of principle examples of how that might happen. And with that noted and as a starting point for reconsidering the line of reasoning offered in my 2010 posting, as repeated and expanded upon here, I cite a brief news story that recently appeared in the New York Times:

Building A.I. that Can Build A.I..

This is a news story about artificially intelligent machines that can build other artificially intelligent machines, and it focuses on a more blue-sky research and development project that is taking place at Google, that they call AutoML (where ML stands for machine learning.) See this Google Research Blog posting:

Using Machine Learning to Explore Neural Network Architecture

The goal of this project is to develop machine learning algorithms that can design and build next generation improved machine learning algorithms, using a neural networks approach as that is so effectively oriented towards supporting iterative step-by-step, experience based development and improvement.

The proximal goal of AutoML is to make it possible for less experienced and less expert artificial intelligence (AI) programmers to make significant advances in developing and refining their own AI software, that can tap into the specific task-level knowledge and insight that they and the organizations that they work for, have particular expertise in. And in fact, the most highly skilled and experienced AI programmers who are out there making necessary advances in their field now, are and will continue to be in very limited supply even as need and demand for them continues to rise. There are way more areas of specialized need for the skills that they have, than there are expert professionals to do all of this possible work. But as this takes off and initiatives such as AutoML really begin to succeed, that goal will be superseded by the larger and more widespread goal of greater efficiency and cost-effectiveness in the innovative effort.

Where is AutoML now in its development? It still represents what will come to be seen as a more embryonic, proof of principle stage for what is to come. As of this writing, the only working examples coming out of this initiative that have come to light, revolve around more effectively solving tasks such as very simple visual pattern recognition tests so a machine can, for example drop a ball into a blue bowl when it is randomly positioned in a grouping of bowls of other colors.

But … the principles involved there have potentially open ended application and for essentially any tasks that could conceivably be captured in an algorithm, fuzzy logic based as well as more deterministic as is being explored up to now.

I find myself writing this posting at a point in time of fundamental, pivotal change. And I write it with an eye towards where the fruits of projects such as AutoML will develop – not “might” but “will.” And when machine learning can effectively supplant the need for human-based expertise and experience in the design and development of AI systems, that will effectively remove one half of the system of breaks that I first wrote of here in my above cited 2010 posting.

That would not make innovation singularities possible as a matter of reaching an infinitely fast pace of development, but it will force a reconsideration of what singularity means as I have tentatively defined it here, earlier on in this posting. And that has the potential at least, for significantly impacting on how people who would variously fit along an innovation adaptation curve approach the change taking place around them, and certainly for those who would naturally find themselves to be later, slower adaptors to change.

I initially planned on offering this as a single, one-off posting but writing it has prompted me to want to write a second, at least somewhat related other new posting too. I am at least tentatively considering as a working title for that: “Reconsidering Information Systems Infrastructure,” and my goal for that is to expand out the scope that is usually included there, beyond information storage and transmission systems per se, to include information and knowledge processing as well, and certainly as that has moved into the cloud and into the information systems backbone per se. The issues that I touch upon here, become important there too.

Meanwhile, you can find this and related postings and series at Ubiquitous Computing and Communications – everywhere all the time and its Page 2 continuation. And you can also find a link to this posting, appended to the end of Section I of Reexamining the Fundamentals as a supplemental entry there.

Leave a Reply

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

You are commenting using your 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: