IEET > Rights > HealthLongevity > GlobalDemocracySecurity > Vision > Virtuality > Affiliate Scholar > Minduploading > Dick Pelletier > Futurism
Creating sentient machines with ‘deep learning’ AI technology
Dick Pelletier   Jan 27, 2014   Ethical Technology  

Building machines that process information the same way a brain does has been a dream for over 50 years. Artificial intelligence, fuzzy logic, and neural networks have all experienced some degrees of success, but machines still cannot recognize pictures or understand language as well as humans can.

    Despite the many false starts though, recent advances in making computers work like the human brain have pointed the way to a new era in artificial intelligence. Breakthroughs in deep learning, a process that imitates the brain with digital neural networks that gather information and react to it independently, have prompted the world's high-tech leaders to invest billions in this latest AI technology.

    Google co-founder Sergey Brin said he dreams one day of building the equivalent of HAL from 2001: A Space Odyssey, except that his HAL wouldn't kill people. Deep learning could turn his dream into reality.

    Google already uses deep learning to improve the speech-recognition system in its Android Jelly Bean operating system and Microsoft is deploying it for the Windows Phone and Bing voice search. Experts, including many of the world's top computer scientists believe that this is only the beginning.

    Face book has formed a secretive research group rumored to be working on the possibility of developing software that can identify emotions in text, recognize objects in photos and even predict the likely future behavior of each of its 1 billion users.

    Google's chief engineer, Ray Kurzweil, recently told MIT Technology Review that he envisions a cybernetic friend based on deep learning that listens in on your phone conversation, reads your emails and tracks your every move so that it can tell you things you want to know even before you ask.

    Experts see huge progress ahead in deep learning systems. The recent coming together of hardware fast enough to train the models, a corporate boost to push it forward and teams formed to make it happen more quickly all point to rapid development of this radical new AI technology.

    Naysayers who once scoffed at the idea of a truly intelligent computer making its own decisions, no longer do. Cornell University's Hod Lipson has built self-aware robots that use feedback from their limbs to learn how to walk. Lipson says his robots can learn, understand themselves and even self-replicate.

    Kurzweil believes a conscious machine capable of understanding complex natural language will be developed within 16 years. "I've had a consistent date of 2029 for that vision," he said "and that doesn't just mean logical intelligence. It means emotional intelligence, being funny, getting the joke, being sexy, loving, understanding emotions. This is what separates computers and humans today."

    And this gung-ho science marches on. Blue Brain Project EPFL – researchers using an IBM 'Blue Gene' supercomputer, are reconstructing brains of different species; including the human brain, in silicon. Chief scientist Henry Markram predicts that with Moore's Law fast-forwarding computer technologies, a full-scale human brain simulation of 86 billion neurons will be achieved by 2023.

    This venture could lead to curing brain diseases, such as Alzheimer's and Parkinson's; and one day, it may even give robots and other 'smart' machines simulated human-like emotions and consciousness.

    Brain-Implantable Biomimetic ElectronicsUniversity of Southern California scientists recently developed implantable electronics that they believe might one day replace aging neurons. Foresight Institute consultant John Burch sees more and more technology like this working its way into our bodies.

    By 2050, positive futurists believe we could be replacing all of our brain cells with materials that process thoughts faster than biological brains can. This faster brain would allow us to run multiple simulations in our mind before making decisions, which would reduce mistakes and raise human intelligence levels.

    Burch describes how we would switch to the new brain: a pill would supply materials with instructions for nanobots (projected development – 2030s) to form new neurons and place them near existing cells to be replaced. These changes would be unnoticeable, but in six months, we will be enjoying our new brain. What do you think, readers; are you ready for a new brain?



Dick Pelletier was a weekly columnist who wrote about future science and technologies for numerous publications. He passed away on July 22, 2014.


More computational resources will not make consciousness, because the brain is not a sandwich;  perception is directly connected to action, without a representational and computational layer getting in the way.
    Just use Einstein’s method, and rethink an assumption:  Language is not symbolic.  It is vocal muscle action in the culture, and all memory is in unconscious muscle activation patterns.  Knowledge is in the culture, not in the head. 
    This satisfies Darwin, Wittgenstein, Lakoff and Johnson, and Occam’s Razor.

“...Imitates the brain with digital neural networks that gather information and react to it independently.”

I was under the impression that deep learning was more than just digital neural networks, but now I’m wondering if I was wrong.

“It has been observed that natural systems with these capabilities are controlled by nervous systems consisting of large numbers of neurons interconnected by axons and dendrites. Borrowing from nature, a great deal of work has gone into setting up ``neural networks’’ in computers. In these systems, a collection of simulated ``neurons’’ are created, and connected so that they can pass messages. The learning that takes place is accomplished by adjusting the ``weights’’ of the connections.” (Digital ``Neural Networks’’—- Natural Artificial Intelligence)

...Imitates the brain with digital neural networks that gather information and react to it independently.”
    To me, this sentence is the problem.  The brain does not gather information, and does not react to information.  What the brain does is muscular actions, and successful patterns are stored in unconscious memory.  No knowledge is in the brain, but rather we use muscular word-actions to access the culture, where all knowledge resides.
    Deep learning cannot think, because its perception is unprocessed by selective action, as in humans.
    Humans don’t need feature extraction, since our active perception focuses on the feature and adjusts the muscular perception aspects, and so on in a feedback loop.
    The muscular perception aspects can spill over into language, which is also muscular, which gives the feature verbal aspects.
    When humans talk with another human, the words of the speaker adjust the unconscious vocal muscle patterns of both speaker and listener.  A thinking AI might receive the reactions of thousands of listeners simultaneously, thus accessing the cultural knowledge from many viewpoints, instead of a single individual.
    The complexity of our neural networks is not in information.  It is simply connections between muscle fibers of both perception organs, and body movement muscle fibers, trained by the environment/culture in time, degree, and sequence.  Period.

Although this recent MIT Technology Review article, “Is Google Cornering the Market on Deep Learning?” suggests the technology will benefit companies like Google, Microsoft, and Amazon as they develop new types of products that can understand and learn from the images, text, and video clogging the Web; what if as this author and others suspect, there was an even greater advantage in understanding Deep Learning.

What if a thorough understanding of Deep Learning also unraveled the mysteries of what we imagine “human consciousness” to be. The leading controllers of this technology would certainly control all aspects of business and wealth. Google becomes God?

Comments welcome.

Deep Learning and the various Big Brain projects will add to our knowledge of brain structure, and our store of brain models that fail to model thinking.  We will zero in on consciousness when a sufficiently huge number of failed models finally overcomes the illusion of language, that words represent entities that actually exist.  By then, it will be obvious to everyone, not just Google, that if God were to give us her wisdom, it would be this:  The essence of language is not meaning, but unconscious muscle activation patterns that enable us to communicate with each other, and access our ratcheting culture, where all knowledge resides.  Verbal logic is peer-pressure, therefor superstition.  This wisdom will bring humans back down to earth, will all the other animals, and we will henceforth make decisions based on a set of Highest Guiding Principles.  We will all be God, and Google will help in spreading the HGPs to all cultures and languages of the earth.

Deep learning and neural networks are the same thing. Deep learning uses an unsupervised preprocessing technique to allow the neural network to have more layers (potentially increases complexity/nonlinearity of the approximate function).

YOUR COMMENT Login or Register to post a comment.

Next entry: Three Strikes and We’re Out (of the open Internet)

Previous entry: An Economist/Philosopher, a Journalist/Politician and a Futurist Enter the IEET Bar..