Tipping point? Artificial Intelligence (A.I.) software teaches itself to learn

originally posted by: peyoxy

The first thing an AI built on this planet will learn is how to "survive" (extend its existence). It will learn that by "studying" survival strategies of every living thing known to humanity (neural network input: survival strategy; network output: survival rate; resulting optimal network configuration: optimal survival strategy).

Anyone spending even 5 minutes thinking about that problem will immediately recognize the solution that any AI will reach - viral spreading combined with parasitic behavior. Hence, the expected outcome - not a single living thing on this planet will survive in its present form and none of them will retain their... volition.

An evolutionarily stable strategy would also be to avoid provoking intelligent opposition. The equivalent to a survivable cyber organism might be something like a soil bacterium: squirrel away somewhere, reproduce, and avoid attention. Little bits of replicating noise in the BIOSes. Attach its way into the phone's distribution system for apps.

December 26, 2016: A couple of days ago, the following free podcast got added to the Scientific American website. It's an interview with an AI researcher on working with robots:


www.scientificamerican.com...

So far, I'm about halfway through the Emotional Intelligence book.

a reply to: lordcomac

Umm...

Just so you know, autodesk has already released first generation versions of exactly what you were saying you wanted to see...

AI and the tech industry aristocracy who have hit it big over the last 3+ decades ... To say the least, it's extremely disconcerting how they and others are rushing humanity as a species towards the singularity without so much as even pretending that they care about whether or not the rest of us are willing to take the huge risks that they're unilaterally taking...

I'm far from a Luddite but I also very carefully consider the implications and potential impact of things I work on etc because I have no desire to make a potentially species ending decision on my own.

a reply to: Aazadan

Realistically when it comes to AI and even low intensity combat let alone a peer versus peer fight we have no real choice but to relinquish control directly to machines in more and more fields...

Especially when it comes to ABM CRAM CIWS & APS systems...

Put bluntly, humans just can't even react fast enough to pull a joystick trigger releasing the system to fire without massively degrading the systems ability to do it's job!

Truthfully, this trend will only continue to propagate to more and more systems etc.

Being brutally honest here, it's not the MIC you need to worry about... It's the damn silly CON valley whizz kids that scare the feces out of the smart people!

You think the AI can be used to translate dead languages like egyptian hieroglyphics?

no wonder why i've seen translations of the japanese novels i've been reading change recently and becoming more readable, i was curious about that.

originally posted by: roguetechie
Being brutally honest here, it's not the MIC you need to worry about... It's the damn silly CON valley whizz kids that scare the feces out of the smart people!

I'm one of those people. I don't work in silicon valley (way too high CoL) but I do the same work as them. Perhaps because I do it and understand it, that's why it doesn't scare me.

originally posted by: Jahari
You think the AI can be used to translate dead languages like egyptian hieroglyphics?

Yes. It's all relational. I'm not super familiar with language translation techniques but I do know lots of translation these days is being done with a concept called a Hidden Markov Model.

If you're familiar with the universal translators of Star Trek, a HMM can effectively do the exact same thing if given a large enough data set. It doesn't require an initial Rosetta Stone to make it work. Everything can be done simply through probability. They're pretty cool IMO, but I find the state diagrams for them to be a bit overwhelming to follow, and the math can get pretty tedious. I'm glad I don't have to use them much.

If you're interested in the subject, here's a few more websites on the concept.
blogs.cornell.edu...
www.aclweb.org...
www.aclweb.org...

I would say that within 20-25 years, we'll probably all be able to carry around a device we can speak into (smartphone) that can translate any spoken language into any written/spoken language for the person on the other end based on HMM's. It won't even need to be an official language, something with a syntax/structure and made up words like Klingon from Star Trek or the dragon language from Skyrim would work with a sufficient sample size.

March 24, 2017: From a few days ago, here's a news release from UCLA (University of California, Los Angeles) on recent research there featuring some breakthroughs in understanding the workings of the human brain. I was especially impressed with their finding of some analog/digital features of human brain physiology:

newsroom.ucla.edu...

Dendrite functions, for example, are now finally becoming better understood. I remember when Einstein's brain was dissected ... they were surprised to find many more dendrites in his brain than in a typical human brain ... the number found was triple what they expected to find, if I recall correctly.

These findings should be a big help to those working in the AI realm. I did see the separate ATS thread already posted on this story, however I think it's helpful to also discuss these findings from the AI perspective.