Artificial Intelligence: Duh? What?

December 13, 2014

I have been following the “AI will kill us”, the landscape of machine intelligence craziness, and “Artificial Intelligence Isn’t a Threat—Yet.”

The most recent big thinking on this subject appears in the Wall Street Journal, an organization in need of any type of intelligence: Machine, managerial, fiscal, online, and sci-fi.

Harsh? Hmm. The Wall Street Journal has been running full page ads for Factiva. If you are not familiar with this for fee service, think 1981. The system gathers “high value” content and makes it available to humans clever enough to guess the keywords that unlock, not answers, but a list of documents presumably germane to the keyword query. There are wrappers that make Factiva more fetching. But NGIA systems (what I call next generation information access systems) use the Factiva methods perfected 40 years ago as a utility.

These are Cheetos. nutritious, right? Will your smart kitchen let you eat these when it knows you are 30 pounds overweight, have consumed a quart of alcohol infused beverages, and ate a Snickers for lunch? Duh? What?

NGIA systems are sort of intelligent. The most interesting systems recurse through the previous indexes as the content processing system ingests data from users happily clicking, real time content streaming to the collection service, and threshold adjustments made either by savvy 18 year olds or some numerical recipes documented by Google’s Dr. Norvig in his standard text Artificial Intelligence.

So should be looking forward to the outputs of a predictive system pumping directly into an autonomous unmanned aerial vehicle? Will a nifty laser weapon find and do whatever the nifty gizmo does to a target? Will the money machine figure out why I need $300 for concrete repairs and decline to give it to me because the ATM “knows” the King of Concrete could not lay down in a feather bed. Forget real concrete.

The Wall Street Journal write up offers up this titbit:

A more immediate concern is that a machine doesn’t have to be super intelligent to do a lot of damage, if it is sufficiently empowered. Stock market flash crashes are one example: Hundreds of millions of dollars have been lost in minutes as a result of minor, difficult-to-completely-eliminate bugs. The clear and present danger, if not the greatest long-term danger, is that mediocre computer programs can cause significant damage if left unchecked. What will happen, for example, when nearly perfect—but still imperfect—software controls not just stock trades but driverless cars? It’s one thing for a software bug to trash your grocery list; it’s another for it to crash your car.

Yikes, killed by a smart blender or crushed by a wayward self driving Volvo, or drugged to mummification by a misconfigured automatic opiate dispenser at your local doc in the box.

Like it or not, memory, CPUs, and programming tools are improving. With each little twitch upward on the performance scale, programmers will try stuff to see what happens. Some of these innovations will mash me into the Swedish plastic grill. Others will allow Uber-inspired outfits and Amazon-wannabes to create solutions that are capable of generating revenue for the lucky innovator.

The reality is that the AI revolution has already taken place. Remember Dr. Norvig’s standard text. It was first published in 1994, and it was a summation of methods some of which were known by Bayes and LaPlace. Don’t recall the dates? Well, older than an experimental Ford zipping around a test track in Michigan.

Here’s my view of where we are with smart software:

  1. Already in operation and largely unnoticed by the glassy-eyed mobile device users. Because the interaction is behind the scenes, no one cares and few take the trouble to learn what is actually going on when one clicks on a mobile device’s map looking for pizza
  2. Improving rapidly. The “real” journalists report on each step of the earthworm in terms of what’s bright and shiny. The pace of innovation in firms using Norvig’s and other wizards’ cookbook of numerical recipes are refining how methods interact.
  3. Discovering new stuff. As more programmers exploit readily available computing resources, little breakthroughs occur. These go unnoticed and then diffuse among the folks who are part of the in crowd. A good example is the temporal method embodied in Foresite. Ah, don’t recognize the name? Well, get Googling.

Thus, AI is here and getting better. Like most things deeply technical, those who understand will become more influential. We know how the US economy rewards those who have “it” and punishes those who don’t have “it”, don’t we.

Worry about AI is an academic exercise. Just kick back and enjoy the Netflix recommendations. When getting more Cheetos, think about NGIA systems and their implications for happy humans. Remember. Humans are in control.

Or…are they?

Stephen E Arnold, December 13, 2014


Comments are closed.

  • Archives

  • Recent Posts

  • Meta