Platt Perspective on Business and Technology

Reconsidering Information Systems Infrastructure 3

Posted in business and convergent technologies, reexamining the fundamentals by Timothy Platt on April 1, 2018

This is the third posting to a series that I am developing here, with a goal of analyzing and discussing how artificial intelligence, and the emergence of artificial intelligent agents will transform the electronic and online-enabled information management systems that we have and use. See Part 1 and Part 2 of this series, and also see two benchmark postings that I initially wrote just over six years apart but that together provided much of the specific impetus for my writing this series: Assumption 6 – The fallacy of the Singularity and the Fallacy of Simple Linear Progression – finding a middle ground and a late 2017 follow-up to that posting.

I have at least briefly touched upon three conceptual and functionally operational dichotomies in this series up to here:

• Artificial specialized, or single task intelligence and its implementation, which is a rapidly advancing reality already, versus artificial general intelligence that is still a vaguely understood possibility, and certainly as for how it might be arrived at,
• Simple tasks, which as I define as tasks that could be carried out algorithmically by a single specialized artificial intelligence powered agent as that form of AI is defined above, versus general tasks that could not be,
• And in the context of a specific “modest proposal” task goal that I have been developing here in a pharmaceutical industry, new drug development context, I have posited two types of drug discovery problems that I identified as Type 1 and Type 2. Type 1 problems would at least in principle, be amenable to management and resolution by simple, single process-oriented (e.g. single task type) algorithms and by single agents that carry them out, while Type 2 problems would require more complex AI systems: systems that might include several or even many such problem solving agents, each with their own separate guiding AI-driven algorithm, with each of these functional elements directly working on and seeking to solve their own specific part of the overall larger problem at hand. And these agents might be functionally managed in a coordinated manner by other still-specialized, organizationally higher level single algorithm agents that carry command and control responsibilities. The lower level agents that they would work with and manage would work on parts of the overall problem presented to this system for resolution, and agents at this next level up, organizationally, would work on “solving” the problem of enabling their coordinated functioning as a more effective collective effort, managing output-to-input sharing between “their” lower level agents for example. And they would flag and preferentially share such data as it would offer more optimized lower level task solutions, coming from those lower level agents, with other agents in this system. That would functionally serve as a filtering process that would serve to restrict and channel next-step action paths taken by those lower level agents, to keep them collectively focused on what has come to seem the most fruitful overall path forward in solving the original Type 2 problem in place. I stress here that however complex this type of agent network might become, and however many organizational levels of action and management that it might come to include, all agent nodes in this would in and of themselves fit the basic pattern of the artificial specialized, or single task intelligence-driven agent.

I also finished Part 2 of this series as an example of what in the film industry is sometimes called a cliffhanger. I have intimated in Parts 1 and 2 that one possible path to the development of a true artificial general intelligence might at least include the development of systems of simpler, more single function agents, and functionally specialized subsystems of them that fit the type of hierarchically networked structure that I just touched upon above, with suitable feedback systems put in place to enable sharable learning across those entire systems and at all relevant organizational levels. I will further discuss this area of consideration and its network infrastructure implications a bit later in this series, simply noting for now that I have still just touched upon this with what amounts to a teaser preview up to here.

And for purposes of this discussion, I have raised the issues of the second and third dichotomous distinctions of this series as repeated above, with that done in Part 2. But I have not effectively connected the more general second and the more problem category-specific third dichotomy sets together yet, at least in any directly practical sense. I began so connecting them here with my reframed and expanded-for-details third bullet point as just offered above. And I begin addressing a second aspect to that here that I have not even touched upon yet, by raising a point that should be pellucidly obvious.

• Any realistic, practical sorting-as-to-type distinction between simple and general tasks is going to be, and will remain fluid and open, and both for what any given point in time’s current technology can effectively address, and for how that technology is perceived and understood.
• And we can expect that much of what would now and in our current stage of technology development, seem to be more general than simple here and of necessity so, will become simple in nature and manageable by even just single task specialized AI agents and in ways not possible now and not even imagined yet.
• Ongoing innovation, and the steady flow of disruptively novel innovations in particular in this, will all but certainly drive that happening. Though even just the accumulated weight of cumulative smaller evolutionary changes will significantly contribute to that reframing too, and to creating the paradigm shifts that we will progressively organize all of this into as we see apparent tipping points reached in what our technologies have become capable of.

What I have been doing in this series, up to here has been to offer a set of building block elements that would go into creating what will become an overall artificial intelligence information management and communications network, and system of such networks as a whole. I am going to turn to at least briefly discuss the questions and issues of infrastructure in this and how it would arise and how it would be managed and controlled, starting in my next series installment. And as part of that, I will more fully consider at least some basic outline requirements and parameters for functionally defining an artificial general intelligence system.

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: