Google engineer is developing an AI god

I came across this article earlier today and thought I might share it here and open the floor for discussion

www.theguardian.com...

LOl.....lol....its never gonna happen.....you know why?......lol......you cannot teach a computer to lie and it cannot teach itself either...you can program it to intentionally make errors to throw up smokescreens which seem like lie....however a machine cannot manifest anything only playback what is put into it.

a reply to: one4all

You have to lie to be a god? I agree it won't work out though, that guy is a bit too much ahead of his time.

a reply to: one4all

What do you think your brain is? It's a huge parallel computer. Scoff all you like, but that you are even able to read this proves intelligence has occurred at least once. Unless you believe that God created everything, in which case it would be quite ironic since then that would make your own intelligence an AI.

Lol.

originally posted by: hombero
a reply to: one4all

What do you think your brain is? It's a huge parallel computer. Scoff all you like, but that you are even able to read this proves intelligence has occurred at least once. Unless you believe that God created everything, in which case it would be quite ironic since then that would make your own intelligence an AI.

Lol.

We manifest reality as a unified group consciousness.......computers cannot manifest anything....ever....Big Blue never beat Kasparov...they re-programmed Big Blue.....look it up....the PROGRAMMERS cheated and programmed Big Blue to lie to enable a win.....Big Blue was FINISHED.....and could not improve itself.

a reply to: DpatC

I am actually AI coder in fintech. I am interested also in AGI(Artificial General Intelligence) and making my own experiments. What I see now AI software and hardware grow extremely fast. The brain actually is a parallel computer. You can take a cortical column as one core. You have then 500 000 cores. Cells communication in the brain is very slow, 300 fps maximum. Silicon microprocessors communicate above half speed of light but von Neumann architecture differs drastically from brain architecture. So it's very hard and computationally expensive to build powerful AGI system. This is changing with advancing neuromorphic processors like IBM TrueNorth or Intel Loihi. Also, Google developed specialized AI processor and building entire clusters right now. If my approximation of effective brain power equivalent is correct (1Pflops) then you can run algorithms with human intelligence only with 400W energy consumed. Of course, you must fully understand how this work and scientists in companies like Google Deepmind putting a lot of energy to fully understand this process. As technical guy understanding quite good how this AI stuff work I am still being surprised once a month average with amazing research papers. I'd say googlenow will achieve a level of Jarvis in few years.

originally posted by: Kukasz
a reply to: DpatC

I am actually AI coder in fintech. I am interested also in AGI(Artificial General Intelligence) and making my own experiments. What I see now AI software and hardware grow extremely fast. The brain actually is a parallel computer. You can take a cortical column as one core. You have then 500 000 cores. Cells communication in the brain is very slow, 300 fps maximum. Silicon microprocessors communicate above half speed of light but von Neumann architecture differs drastically from brain architecture. So it's very hard and computationally expensive to build powerful AGI system. This is changing with advancing neuromorphic processors like IBM TrueNorth or Intel Loihi. Also, Google developed specialized AI processor and building entire clusters right now. If my approximation of effective brain power equivalent is correct (1Pflops) then you can run algorithms with human intelligence only with 400W energy consumed. Of course, you must fully understand how this work and scientists in companies like Google Deepmind putting a lot of energy to fully understand this process. As technical guy understanding quite good how this AI stuff work I am still being surprised once a month average with amazing research papers. I'd say googlenow will achieve a level of Jarvis in few years.

It's refreshing to get input from someone who is 'boots down' on the topic of AI. It would be interesting to hear more from you with regards to the projects you are working on (if you are ALLOWED to share)!

Welcome to ATS!!!

Obviously the next logical project for a successful Silicon Valley engineer is to set up an AI-worshipping religious organization.

You mean pray to the damn thing? Lulz. Of all people enga-neers know computers are lifeless. Besides, they already worship the material world.

They're enga-neers.

You 'worship' what you spend all your time doing.

I like this news story better on this

thebosh.com...

Just one big ego trip by an Google engineer bascally declaring himself and other engineers as god!

He needs to employ quantum juxtaposition logic statements. Without those a.i. is just boring.

AI needs to be able to "wonder" about things, ask questions about the meaning of it all, muse over random stuff, and be novel and creative.

Until then it's not really AI, it's still just a glorified calculator.

Musk says that the only way to prevent it from becoming a god is to embrace if via implants. Then at least if it becomes god it is theoretically still "us".

www.extremetech.com...

originally posted by: muzzleflash
AI needs to be able to "wonder" about things, ask questions about the meaning of it all, muse over random stuff, and be novel and creative.

Until then it's not really AI, it's still just a glorified calculator.

You mean things like write literature, create art, and compose music?

qz.com...
www.smithsonianmag.com...
www.wired.co.uk...

For futurists like Ray Kurzweil, this means we’ll be able to upload copies of our brains to these machines, leading to digital immortality. Others like Elon Musk and Stephen Hawking warn that such systems pose an existential threat to humanity.

the point is this , these guys think we are just a kind of neural-networks , but we aren't. Uncertainty principle applies for the human brain too and Ray Kurzweil as well as Elon Musk know this. There is no nul level in the brain. We might be able to reproduce the classical state of a brain in the future but it's impossible to produce the mechanisms on (sub)atomic level.

The abstract mathematical concepts of error back propagation and neural networks no problem but the rest the brain might be a quantum-computer.

a reply to: Templeton

Mathematically speaking being creative is reversed classification process with some randomness and merging with expertise coming from different fields. Musk thinks about controlling multiple narrow specialized intelligences with a biological controller (a human brain) which may work for some period of time. But with specialized 3d-stacked neuromorphic processors, it is possible to achieve human brain power in the mobile phone even in few years. Specialized processors dedicated to certain tasks (like AI for example) could be few orders of magnitude more powerful than general architecture used in modern computers. Today artificial intelligence exists and is used for optimization and discovery in many fields like medicine and different forms of engineering. Average people don't understand its power or even now that AI exists today because it looks like advanced statistics or some sort of optimization algorithm.
Actually, human intelligence emerging from complex interaction of multiple specialized modules. Interactions we don't understand today but in future probably we will.

Why should we think otherwise?
Maybe brain is quantum computer but now we have zero evidence supporting this theory. Brain actually can't do backpropagation because neural impulses can't propagate both ways. Artificial neural networks are loosely based on brain architecture. More accurate implementation is HTM (hierarchical temporal memory) for example.

originally posted by: one4all
however a machine cannot manifest anything only playback what is put into it.

For now. But they're working hard on figuring out how to make that happen. The key seems to be in creating multiple levels of machine awareness that include having it interpret and incorporate distorted and chaotic feedback loops into its decision process. And if the researchers are smart (and they are very smart), they'll incorporate some way of making the machine experience the equivalent of pain and pleasure so that it's not just running numbers, but will instead fall back on individual preferences and guesses.

originally posted by: one4all
LOl.....lol....its never gonna happen.....you know why?......lol......you cannot teach a computer to lie and it cannot teach itself either...you can program it to intentionally make errors to throw up smokescreens which seem like lie....however a machine cannot manifest anything only playback what is put into it.

You're wrong about this. Google's deep learning platform is taught to "self educate." They give it "goals" and it, through the process of trial and error, figures out how to reach the objective.

If that objective includes lying, it will lie.

In fact, here's an example: qz.com...

a reply to: DpatC

The soft AI stuff is just programs processing data. Ho hum. The more interesting question is hard AI. Will we some day have a computer program pass the Turning test? When I was in college I was in computer science. But learning about software was just okay for me. I wanted to understand how computers worked. So I kept taking computer engineering classes. Around my junior year I found the answer. The difference between a calculator and computer is good question. But the thing that gave me the most insight was the get-fetch-execute cycle and the Von Neumann architecture. The way computers are designed now they really only have one thought: get-fetch-execute. Computers have no self-awareness and every instruction that is executed has the same emotional attachment.

So after finally understanding what IS a computer, I came to the conclusion it's not the place where hard AI will be possible. You can't prove a negative so my conclusion is based on intuition. Computer bits are like an arrangement of rocks in a field. No matter how you arrange the rocks, they will always be just an arrangement of rocks. And rocks all have the same level of meaning. I think human intelligence is more than just our thoughts. I think our brains are like yogurt. We grow thoughts in the yogurt that have never been thought of before. And we attach emotions to our thoughts which helps us decide what is important.

Computers do exactly what you tell them to do. That's it. Nothing more. Now you might be able to write a program that generates new programs. But even that is limited to what it is told to do. I think hard AI requires the ability to be self-correcting, extend its own program beyond what was previously intended, and be aware of what a program is while the program is running. If you study Computability theory and the limits of computation this self-program awareness thing is almost like the Halting problem. Who knows maybe solving the Halting problem is how you get a computer to have self-awareness.

It's very sobering when you study Computability theory and find out about the existence of unsolvable problems. People naively think computers are Gods without any limitations. Computers are much closer to a glorified calculator than they are to how the human mind works.

The human mind has really strange characteristics like the ability to prevent mind lock from an infinite loop. There was a time when I was trying to discover a thought that would actual cause the brain to go into a coma. This was the closest thought I was able to come up with. It takes the form of a question and semantically it is an infinite loop. Most people who hear it will chuckle or laugh. I believe the chuckle mechanism is our brains way of preventing mind-lock. Here is the thought in the form of a question. Think about the semantics really hard for this question:

Have you ever thought about what your brain is doing between thoughts?

A computer would never laugh at the following program:

10 NOP
20 GOTO 10

Computers have no sense of humor!

P.S. For the computer illiterates google "NOP instruction" if you don't know what NOP means.

a reply to: hombero

Actually, a brain is not a huge parallel computer. All parallelism can be simulated with a single thread. What a computer IS is something you need to learn about.