On Watson and AI -- Responses to the Responses

I've had a few more comments on IBM's Watson, and thought I'd share the essence of the commentary as well as my responses, especially since I clarified a lot of details in those comments.

First, "Priest Valon" commented...

AI theoretics is in my wheelhouse

Decision and analysis in a complex space is not easily modeled, but relatively recent advances have done so.
Essentially, computers are very very fast, very very accurate calculators... but a traditional algorithm, even an adaptive algorithm will never operate at a level of decision making that approaches real world complexity.
The solution is quite simple, at an "atomic" level...  constructs called neural agents.  They operate using a finite algorithm, are smart (in that they are independent instances of a class object) and follow an independent solution path, and then compete with other agents for the correct solution.  The group of highest ranking paths defines the "thought" of the system.  Computational game theory is the foundation of this and is widely used in proprietary systems like stock analysis.

To which I replied...

This is an excellent point, and something...

I wanted to discuss, in a slightly different way, about AI. While I am open to the idea of an eventual, sentient AI (or something very like it) being possible, it's in many ways the fact that we can break down important, complex tasks into a series of steps that we program computers and/or design other machines to accomplish.
In addition to various methods of problem solving -- evolving problem-solving agents, sifting through existing archives of options -- there are all the things we can design a machine to do.
For example, how complex is it to do your taxes, really? If all of the data could be inputted through the receipt of pre-entered data, or scanned from standardized forms, could your computer (or "Muse") do your taxes for you, and within reason pay them, without even involving you? How many tasks have we already broken down sufficiently that a reasonably competent AI could step in and handle the gruntwork itself?
Note, both of the computers I've mentioned which are already doing scientific research are essentially taking tasks that involve a large number of repetitive steps and automating them. In contrast to your example of neural agents, the designers essentially found a method or set of methods that worked -- which in some cases they knew already worked -- and implemented them on a massive scale.
To be blunt, I think Watson's potential ability to "just follow orders" is going to be huge, because so much labor can be broken down into discreet steps already, without requiring terribly complicated decision-making skills, just the capacity to understand what is being asked for.

"Valon" makes s
ome further points about the tricky nature of understanding everything from syntax to motivations in trying, for example, to do someone else's taxes on TurboTax. He also points out that you can train someone with multiple attempts at a task, but only if they have some metrics by which to differentiate success from failure.
To which I replied...

Indeed, which is why Watson's ability to understand the odd phrasing and fuzzy logic of Jeopardy questions is such an important first step. Other issues -- voice recognition, facial recognition, text scanning, handwriting recognition -- are well along already. But if you can consistently understand what humans living in the real world want without having to ask, and especially without their having to laboriously explain it, you can take a huge number of tasks off their shoulders.
Because we've already simplified and standardized those tasks already  -- but someone still has to do them. But soon, that someone may be named Watson.

Oh, and this leads into another issue that of the limited AI executing its program in a highly focused, anti-social way that ends up obliviously killing off the ecosystem around it. While this point is often made about many shortsighted corporations, I'm reminded of the rapidly evolving, voracious micro-corps of Accelerando, who ended up adapting to a fiercely competitive environment by feeding on each other and killing off the more advanced AIs. These fast-thinking, semi-sentient micro-corps kept evolving to fill any available economic niche, with only their ability to seize resources and destroy competitors mattering in their "social system" and sense of "morality."

Eventually, they wiped out every civilization that gave birth to them, and themselves as well.
One thing that gives me hope about new social networking systems like Facebook is that being a bully or a troll is not automatically rewarded over being a good person.
I think people are starting to make more active judgment calls about who they really want to be associated with in real life and online, and on who they want to cut out of their lives.
Which makes us all that much harder to manipulate, or undermine, or thwart, or destroy.

Regarding the potential, further displacement of human jobs by automation...

The fact that so much human labor can be replaced is one reason I keep bringing up human enhancement and human augmentation.

To be blunt, how many human beings do we have who already live up to their full potential? How many are doing truly satisfying, meaningful work? How many are being listened to, and having their creativity and other gifts fully tapped by their present work, and their present lifestyle?
I would say... very few.
Simply changing that one aspect of life would likely be one of the most dramatic augmentations we'll ever see. Not only in terms of ending all the waste of human potential we have now, but in the flowering of civilization made possible by billions of people reaching even their most basic intellectual, artistic and human capacity, right now.

And on the same subject, with the further concern that our society is in no way prepared to handle it...

Well yes... and no...

You have a very good point about a lot of jobs getting automated away and not necessarily replaced.
I would point out two things. One, our present peaking fossil fuel production, climate-change disruptions, harvest shortfalls and general economic issues make a lot of investment to replace workers who can be had increasingly cheaply in so much of the world a bit iffier of a proposition. Not that it can't happen, especially if we can change our energy sources and work through the rest of our problems, but there it is.
The other thing is... I think a lot of people are starting to look more seriously at how they live their lives, and whether they value simply having more and more electronic, motorized and other "stuff" in bigger and bigger piles... Or whether they want a life that is more connected with nature, with their community and with their family and friends. And in which their possessions and the services they receive are more personalized and meaningful.
If the latter... Or if resource constraints force us to change, or both...
Then I suspect we'll be looking at more localized food production -- more local farms, gardens, orchards, aquaponics, urban gardening, community gardening -- more handmade items (especially local ones, or made by oneself or one's friends), and so forth. Simply how people live their lives, and whether they value those around them for moral, aesthetic or basic insurance reasons, has a huge impact on whether we keep people employed, and employed at meaningful jobs, or not.

Another poster argued that I was not offering enough attention to the social-learning "growbot" strategy for developing AI, and that overpopulation and climate change made a major die off in our global population inevitable...

Well, that brings up quite a few issues...

With AI, there  are actually a number of different prevailing theories (how many depends on how seriously you take each one and each permutation of each one).
I mean no offense to the "growbot" perspective or any other. My point with Watson is that we appear to at last have a basic, general AI -- not perfect, not omnipotent, but capable of understanding our requests well enough to soon begin handling a lot of tasks with minimal supervision.
In fact, it's exactly the huge amount of work that has been done by humans with computers and other systems that suggests that a cheaper, more powerful Watson could soon be the controlling interface "just following orders" which thereby revolutionizes a wide array of efforts.
You know how modern computers have operating systems (OS)? Think of this as potentially an operating system's operating system.
If that tremendous amount of work were not already in place -- tax software to do your taxes for you, Droid phones that can make restaurant reservations, Google as well as Watson, and so much more -- then IBM's effort would be far less impressive. As it is, we seem to be looking at a turning point in AI -- not an earthshattering change, but a step that will lead to others.
As for the challenges facing humanity...
Though I wish to stay out of the political discussion of what happened in Cairo, I think the young people of Egypt made a huge statement insofar as their revolution was practically bloodless in spite of those determined to see it crushed.
So far, the young of this world are not playing by the cynical playbook that so many insist is their only option. I don't think they intend to.

To a poster who felt that Watson was not a basic AI, but no more important than speech recognition or an ordinary chess-playing computer...

I have a different perspective...

Just about every Jeopardy question is some kind of a basic joke, or riddle, or otherwise murky way of phrasing the "question" -- which is a non-question in itself.
If you've read through what I've posted on this subject, then you know we already have two computers doing scientific research on their own, and that a host of other functions have already been automated. Some of which I've described, but I've left out some of the most formidable, such as completely automated manufacturing from a set of plans and inputted raw materials.
Watson, as I've said elsewhere, can understand relatively obscure phrasing and determine what it is you're asking for. That's huge, because there's already a host of fully automated functions out there which it could perform, if it -- or its successor -- can handle the most basic function: Understanding what you're asking for and acting on that request.
Effectively, as I've noted, Watson is potentially an operating system of operating systems. Just as your laptop or smartphone provides an OS and a platform within which its various sub-programs work, Watson could be the master program that interprets what you say when you ask a question or give an order and then acts accordingly. Does it matter if Google is better at searching the entire Web if your computer pulls up that information and filters the results for its owner? Does it matter if TurboTax does your taxes if Watson pulls down the data from your digital and hardcopy records and effectively does your taxes for you?
There are steps to be filled in, but like an iMac or iPhone, or even a supercomputer, it's not just the hardware and the operating system, it's the applications. But this is so impressive because so many of those applications already exist, we're just oblivious to them.
Now if you believe that an AI must be a godlike, recursively self-improving, fully sentient intelligence, then I acknowledge that by your definition, Watson is in no way an AI. But that is by your definition, not mine.

A remark that once our backs are to the wall on energy, or food, we should finally see real change in how governments handle these issues...

One thing that gives me hope is that I see a lot of people acting, and not in the frantic, panicked, destructive way so many observers seem to expect from the public. Instead, whatever you may think of recent unrest, it's impossible to note that a major Middle Eastern power just fell in an essentially bloodless revolution organized by smart, non-fanatical, pragmatic young people.
You know, the very people who are supposed to be dis-empowered and radicalized.
Could it be that -- gasp -- our bigoted assumptions could actually be Wrong?