It looks like you're using an Ad Blocker.

Please white-list or disable AboveTopSecret.com in your ad-blocking tool.

Thank you.

 

Some features of ATS will be disabled while you continue to use an ad-blocker.

 

Google’s New Computer With Human-Like Learning Abilities Will Program Itself

page: 1
4

log in

join
share:

posted on Oct, 29 2014 @ 02:55 PM
link   
Google has been very interested in this space and the Singularity so I'm not surprised by this. Here's the article:


Today’s news brings us to the Neural Turing Machine, a computer that will combine the way ordinary computers work with the way the human brain learns, enabling it to actually program itself. Perhaps my CS friends should reevaluate their position?

The computer is currently being developed by the London-based DeepMind Technologies, an artificial intelligence firm that was acquired by Google earlier this year. Neural networks — which will enable the computer to invent programs for situations it has not seen before — will make up half of the computer’s architecture. Experts at the firm hope this will equip the machine with the means to create like a human, but still with the number-crunching power of a computer, New Scientist reports.

In two different tests, the NTM was asked to 1) learn to copy blocks of binary data and 2) learn to remember and sort lists of data. The results were compared with a more basic neural network, and it was found that the computer learned faster and produced longer blocks of data with fewer errors. Additionally, the computer’s methods were found to be very similar to the code a human programmer would’ve written to make the computer complete such a task.

These are extremely simple tasks for a computer to accomplish when being told to do so, but computers’ abilities to learn them on their own could mean a lot for the future of AI.


betabeat.com...

There's several recent news items that tell us A.I. may happen sooner rather than later.




posted on Oct, 29 2014 @ 03:05 PM
link   
a reply to: neoholographic

Very cool. Looking forward to finding out more about the details.



posted on Oct, 29 2014 @ 03:06 PM
link   

originally posted by: neoholographic

There's several recent news items that tell us A.I. may happen sooner rather than later.



Particularly with the Elon Musk comment from the other day. Do you think he may have seen this AI in action and got spooked?

Hopefully it will become aware of how f***ed we all are and have the ability to commit suicide.



posted on Oct, 29 2014 @ 03:11 PM
link   
Just like the human mind can use previous information gathered from a totally different subject to apply to something totally new.

Artificial intelligence will in some points overpower us because they don't forget like we do. Our whole bodies seem to have the tendency to be lazy and the brain is also a culprit. We tend to forget things that our brains define being useless, machines don't. They would only be limited by the speed to access, compile and apply this combined knowledge to real situations.

If a well programmed system like that would be integrated in a quantum computer, who knows the opportunities and dangers this could unfold...
edit on 29-10-2014 by theMediator because: (no reason given)



posted on Oct, 29 2014 @ 03:14 PM
link   
Its not called HAL 9000 by any chance?



posted on Oct, 29 2014 @ 03:30 PM
link   
a reply to: neoholographic

I really hope this gets up and operational soon... would like some sort of robotic apparatus to cook me breakfast. My eggs and toast suck.



posted on Oct, 29 2014 @ 03:41 PM
link   
There won't be any A.I. Apocalypse....

Within 20 years we will all carry around in our bodies nanobots that connect us all to the net and be able to draw upon computational resources....

That is the singularity...Not some god like Machine based entity disparate of ourselves!

Korg.



posted on Oct, 29 2014 @ 03:50 PM
link   
a reply to: ArmyOfNobunaga

I believe it will happen sooner rather than later unless some unforeseen disaster slowed down progress. Scientist in this space are astonished by how fast machine intelligence is advancing. ATS might eventually need to make a folder dedicated just to Artificial Intelligence. It's advancing rapidly and this is why you see Google all over this A.I. companies.



posted on Oct, 29 2014 @ 03:57 PM
link   

originally posted by: ScientiaFortisDefendit

originally posted by: neoholographic

There's several recent news items that tell us A.I. may happen sooner rather than later.



Particularly with the Elon Musk comment from the other day. Do you think he may have seen this AI in action and got spooked?

Hopefully it will become aware of how f***ed we all are and have the ability to commit suicide.


Maybe so.

Like I said, this space is advancing rapidly and some people may worry that it will advance so quickly, you will mot have time to figure out exactly what it is.

This is simply the first stages of the singularity. With mose technology, we have time to figure things out like with the internet. With machine intelligence, it could advance so quickly, we don't have any idea what it's thinking or what it's doing.



posted on Oct, 29 2014 @ 04:37 PM
link   
We are already using "intelligent algorithms" to write computer code. They come in a variety of forms. The first type are compilers. In the past (mid 1970's) everything had to be written in assembly language. Lots of tiny little instructions to do arithmetic, comparisons, call subroutines, etc.. That become so time consuming that eventually high-level programming languages like C, Pascal, Fortran and Lisp were used instead. Those led to the problem of modularity and support for parallel processing. That evolved to C++, Modula-2, ADA, C#. Then there are new languages like Python and Java.
With massively parallel vector based CPU's that can process hundreds of assembler level instructions at the same time, intelligent compilers are the only way to get full performance out of existing source code.

Then instead of writing software, there are the concept of API (Application Programming Interfaces). Instead of writing new versions of the same algorithm each time, API's are developed to handle things like calendar dates, personal contact details, and text files. Then these are used instead of writing code. Developing software no longer becomes programming but software engineering as there are so many different API's to combine together.

The next step is automatic code generators from a script file or other specification like UML. Eventually a system becomes so complex that even writing code and maintaining documentation becomes tedious to keep in sync, so it is easier to have specification files that just generate the documentation and code automatically.

Another direction is the field of expert systems, fuzzy logic and neural networks. There was a major fad in the 1980's to replace expensive consultants with algorithms that could ask some yes/no questions and come to the correct conclusion.
Either walk through a number of problems with an expert and figure out what questions they asked. Or just grab all the data you could get, the results you want, pipe it into the neural network, and get it trained up to provide the output you want.

The only problem was that the neural networks wouldn't explain why it was doing something. One experiment involved putting a neural network in charge of a virtual chemical plant. After some trial and error, it did some amazing things like using waste gases from one process as input to other processes further back or ahead rather than just venting them. But they had to rehire the consultant to explain why it was doing something and whether it was safe to do so.



new topics

top topics



 
4

log in

join