If we are ever going to crack the biological barrier (IE, make real intelligence), we can’t have this kind of inefficiency in our machines:
“a Google search uses just about the same amount of energy that your body burns in ten seconds.”
Ten seconds!!! Man, that’s slow. It takes us, what, about two orders of magnitudes less, if not three, to conjure up a thought, right? Kevin?
I think you’re comparing the wrong thing. It would take a human years/decades/centuries of energy to perform the search without help.
Matt has it right, comparing the time would help
Also, compare the amount of energy that Google uses to the amount of energy a human would use to find the same information
@matt @reg4c
I believe that it’s not about being better or more efficient than a person, I believe it’s about not being efficient enough for a large scale AI.
The Google response is to a piece of well crafted PR fluff (or BS depending on how you want to assess the quality of the article…) that appeared in the London Times over the weekend. Now the article (imho) smells very badly indeed, but to get to your point, John, Google is referencing the energy consumption of the entire human body, not the brain. You burn a fair amount of energy per second just existing (DNA replication, cellular repair etc etc). Not sure how much energy is consumed with a “thought” and I’m not sure you can even define the quantities in a rational manner. Your starter q for that one would be, how do you calculate the energy expended during your life to learn what you need to understand how to search and then to understand the answer. Oh, and what’s the unit of thought? I could go on…
But just think of all the energy Google saves. Trips to libraries & book stores to find the answers to random questions. Trips to return bad purchases because we couldn’t read reviews. Waste time driving because an establishment was closed, moved, or we got lost because we didn’t have direction!
Artificial Intelligence CAN just be information retrieval through one or more sensory pathways with an ability to stratify, associate and weigh each piece of information against every other piece of information for decision on some action within a limited number of possible actions.
We have those now. A simple type would be a thermostat with just one sensory input. More complex would be anti-lock brakes and airline collision-avoidance systems.
SO what we’re really talking about I think is a more independent intelligence, either partially or totally.
And for this you have to deal with multiple motivations and goals continually being weighed against both an established and ever-expanding data base and live-streaming new sensory data through whatever systems employed suitable for the environment in which our golem is operating.
Bottom line…
Yeah, either much faster, smaller and more efficient…
Or its some version of Kevin Kelly’s “New Kind of Mind” which can be worldwide but brings up interesting identity issues worthy of thinking about.
See article and my comment this morning at
http://www.kk.org/thetechnium/archives/2009/01/as_if.php
This post made me laugh. Nice.
But still, it takes maybe two seconds to move the hand from the keyboard to the mouse and get the cursor into the Google toolbar, and then type out the thought (and hit backspace several times and retype, if you are like me) and then 0.15 seconds for Google to run the search.
Are we at 10 seconds yet?
The difference is that we can feed more and more energy cheaply into the machines while human energy intake is limited.
“efficiency” is far less important than cost.