Search This Blog

Monday, April 23, 2012

Google Truth

Sometime I just want to type something like the following into Google: "Really, can't you just tell me if the world is warming? Like what's the probability that the global average temperature is more than 2C higher in the year 2100 than it is now?"

I'd even be willing to type "mirror, mirror on the wall" first. And oh-by-the-way, Google, just who is the fairest of them all these days anyway.  And show me her picture (fully clothed of course - what were you thinking?).

Then Google would respond (about the temperature question): "Given the current state of knowledge, 11.3%".  Or something like that.

All Google has to do to make that happen is to have their search engines actually understand all that information that they have in their servers and to figure out which sources are more reliable than others via the accuracy of previous prediction and reliance on the sites by other sites.

Piece of cake.  I imagine they'll have something like that going within a decade or so.

They'll call it Google Truth.


erp said...

Here's a truth about global warming: the temperature this am here on the east coast of central Florida was 38F.

Bret said...

That was just an example. I want to be able to ask about anything in the universe, no matter how complex and uncertain and have google tell me the answer with probabilities. It may have to ask for clarifications for some questions.

No doubt the cooling in Florida is caused by global warming though. :-)

erp said...

Google has been on board with the narrative expansion for a long time now. Old encyclopedias are going to worth their weight in gold in a the not too far future. It's the only way to learn the truth unvarnished by political correctness.

Bret said...

That's probably true that google (or governments) might purposely distort the available information. It may not be google and it may be decades, but eventually the raw data will be widely available and computers will be smarter than humans as well. at least within the narrow competence area of synthesizing mind boggling amounts of data into coherent understanding.

erp said...

Gosh, I haven't heard that argument in a long time. Unless computers are able to overcome the biases of their makers, it won't much matter how smart they are, they will move the narrative forward even more adroitly until something happens to change it.

erp said...

Sorry about the quadruple comment. Ya think I got Google gods mad by intimating they may be slightly partisan? ;-]

Bret said...

I think it's the other way around. The question is how long will the makers be able to filter the computers input and output in order to instill their biases.

A newspaper editor can easily bias the paper's output because of limited input.

Google's database is estimated to be many hundreds of trillions of entries.

Now imagine many orders of magnitude more information which is what will be collected in the next few decades.

Again, it may not be google that does this. But the information from all blogs, all commentators, all scientific and engineering output, etc. will be collected in one or several extremely large databases.

It will be beyond human capacity to affect the output of information sifting software.

erp said...

I know that anecdotes are just that, but just in my own experience, all the material that I had gathered about Obama from the time he appeared at the Kerry convention until he started his presidential bid, just disappeared. At the time, it didn't occur to me that I needed to take screen shots and copy stuff to my own desktop or better still backup on flash drives.

How much more likely will it be that stuff gets offed before it gets to stick in a future controlled by the freedom haters.

I truly hope you're right and this is merely a dry patch that will be a memory as soon as a fresh rain washes it away.

Hey Skipper said...

In the link, human equivalent is taken a 10E9 MIPS.

That number smells like it was pulled from someone's rectal data bank.

I have no idea what the real number is, but I'm pretty certain there is no way to convert the time it takes to recognize someone from an odd angle in marginal lighting to MIPS.

Bret said...

The methodology for calculating the 10E9 MIPS (which is 10E15 instructions per second) is quite interesting and it goes something like this (this is from memory so it won't be quite right)...

While we don't know how all or even much of the brain works, we do know how parts of it work. One of those parts is the visual cortex that does the pre-processing on images received from the retina. That portion of the brain weighs a certain number of grams (call it X) and the number of MIPS required to do that processing is well-known (call it M). Then, given the brain weighs Y and assuming one can extrapolate the processing density of the visual cortex to the rest of the brain, the total MIPS required for the entire brain is Y/X*M.

While it's certainly not particularly accurate, it's also almost certainly not off by more than two orders of magnitude in either direction. And since two orders of magnitude is about 10 years on the current exponential trajectory, the time when $1000 buys human level processing capacity will only vary by a decade or so.

erp said...

Bret, at this point, I'm having trouble doing my granddaughter's SAT prep math questions (she finds that quite amusing), so I can only follow your logic in the most general way, but I think we humans will develop/evolve the ability to control our brain children and not create modern-day frankensteins who will eventually control us.

Again, just anecdotal, but it seems that kids today are born with the ability to deal with electronic gadgets and complex systems that seems hopelessly complex to us geezers.

I'm much more frightened of people who will voluntarily hand over control to human overseers with EBT cards and promises of eternal bliss.

Hey Skipper said...

From Slate a few days ago:

A lot more than two. Most computational neuroscientists tend to estimate human storage capacity somewhere between 10 terabytes and 100 terabytes, though the full spectrum of guesses ranges from 1 terabyte to 2.5 petabytes. (One terabyte is equal to about 1,000 gigabytes or about 1 million megabytes; a petabyte is about 1,000 terabytes.)

I think the question of capacity is conceptually easier to estimate than speed. In this regard, the variation is more than three orders of magnitude.

Heck, even if 10E9 MIPS is spot on the money, we have no earthly idea what is happening during even one of those MIPS.

I suspect that so long as hardware relies on binary computation, then there will be human mental processes that will forever be out of bounds to machines.