2
$\begingroup$

To reach full autonomy in any fully automated device it must finish its task in such a way that human control is unnecessary. We know when the automation is excellent when there are no manual controls and we call it repair and bring in a specialist if something goes wrong.

Four examples of full automation in existence are.

  • Appliances
  • Home and mobile computer connectivity
  • Mail sorters
  • Hundred million dollar military drones

These four are specifically intelligent in varying degrees.

  • Process control under household conditions
  • Adapting to new hardware and networking
  • Reading addresses written with poor penmanship and odd fonts
  • Adaptively avoiding detection to reach a reconnaissance vantage point

These four are good enough for their markets.

Examples of not being good enough, as indicated by their lack of any substantial market penetration, are these four.

  • Autonomous vacuum cleaners
  • Autonomous cars (without a driver's seat)
  • Unpiloted private or passenger aircraft
  • Narrowly targeted medical nanites

The question, "How good is good enough?" is this one:

What is the challenge for researchers and engineers to provide ENOUGH intelligence into these kinds of autonomous vehicles to make them better than current methods in the minds of policy makers and consumers?

Stepping back to look with a scientific eye at what is acceptable, consider how unsatisfactory the existing equivalents of the above four are.

  • Manual vacuuming misses anywhere from 10% to 90% of the dust depending on the surface, blows microbes into the user's lungs, and produces additional health risk when disposal is required.
  • Human beings drive cars regularly, but they are driving what is technically a piece of heavy equipment in pedestrian situations when tired, drunk, high, while text messaging, or while simply loosing focus.
  • The human resources required to deploy, guide, and land vehicles that have no other obstacles than topographical features and other aircraft is significant and leave open not only human failure but hijacking.
  • Chemotherapy, antibiotics, and other pharmacological interventions often only delay the progress of disease and sometimes produce other negative outcomes of varying scope from symptoms worse than what is being treated to death.

Many things that are manual are like that. They need to be automated. Artificial intelligence, especially miniaturized and low cost artificial intelligence, is critical to achieving anything like excellence.

What makes something intelligent enough. What specific research and engineering efforts can bring the items that aren't good enough into the realm of consumer demand and supported by policy?

$\endgroup$
3
$\begingroup$

The ideal answer should be:

  • When the algorithm is shown statistically to outperform humans in a given task.

However, there does seem to be an emotional component when related to life-or-death scenarios.

I have no doubt that robot-vaccums are vastly inferior to good-old-fashioned cleaning, but as a celebrated writer once noted "you can either have clean floors or you can use a mop." (i.e. if you want floors that are actually clean, you need to use elbow grease and scrub them;) But people rarely die from shoddily cleaned floors and the majority seems to value convenience over all other factors.

By contrast, the internet goes crazy when there's a single autonomous vehicle fatality, even as a significant subset of humans are driving more erratically due to mobile phones and on-board computer distractions. (I liken it to fear of flying--statistically safer than driving but humans aren't always strictly rational.)

Beyond the safety issue, humans are often irrationally selfish--commute in any major city during rush hour and you will notice pervasive grid-locking. While gridlock is mostly just an inconvenience, humans often make the same irrationally selfish decisions accelerating into yellow lights, increasing risk of an accident. Similarly, you often encounter rogue drivers speeding in highway traffic, making unsafe lane changes, regardless of whether it actually saves them any time.

  • My sense with self-driving cars is that they will have to exceed humans by a large margin due to our emotional nature, regardless of what the statistics indicate.

AND, there will be an entrenched, obstinate population of aging drivers who resist the change, even as their own driving becomes more erratic and dangerous. File under: "You'll get my driver's license when you pry it from my cold, dead hands".

$\endgroup$
  • 2
    $\begingroup$ @DouglasDaseeco Humans seem to regard driving as a strictly partisan activity, as opposed to what it actually is, which is communal activity. My sense is that autonomous vehicles will bring a default benefit in this regard because they're designed to follow the rules and avoid harm. (Re: Donne, "Thou had made me, shall thy work decay?" seems apt for human drivers as they age, where algorithms would presumably get stronger and more effective.) $\endgroup$ – DukeZhou Jul 20 '18 at 17:27

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.