What does it mean to understand the brain? Where are we on the roadmap to this goal? What are the effective routes to progress – detailed modeling, theoretical effort, improvement of imaging and computational technologies? What predictions can we make? What are the consequences of materialization of such predictions – social, ethical? I will address these questions and examine some of the most common criticisms of the exponential growth of information technology including criticisms from hardware (“Moore’s Law will not go on forever”), software (“software is stuck in the mud”), the brain (“the brain is too complicated to understand or replicate”), ontology (“software is not capable of thinking or of consciousness”), and promise versus peril (“biotechnology, nanotechnology, and artificial intelligence are too dangerous”).
There is now a grand project comprising at least a hundred thousand scientists and engineers working in diverse ways to understand the best example we have of an intelligent process: the human brain. It is arguably the most important project in the history of the human-machine civilization. The goal of the project is to understand precisely how the human brain works, and then to use these revealed algorithms as a basis for creating even more intelligent machines.
As we learn the algorithms underlying human intelligence, we will similarly be able to engineer it to vastly extend the powers of our intelligence. Indeed this process is already well under way. There are literally hundreds of tasks and activities that used to be the sole province of human intelligence that can now be conducted by computers usually with greater precision and vastly greater scale.
Was it inevitable that a species would evolve that is capable of creating its own evolutionary process in the form of intelligent technology? I will argue that it was.
According to my models we are only two decades from fully modeling and simulating the human brain. By the time we finish this reverse-engineering project, we will have computers that are millions of times more powerful than the human brain. These computers will be further amplified by being networked into a vast world wide cloud of computing. The algorithms of intelligence will begin to self-iterate towards ever smarter algorithms.
This is how we will address the grand challenges of humanity such as maintaining a healthy environment, providing for the resources for a growing population including energy, food, and water, overcoming disease, vastly extending human longevity, and overcoming poverty. It is only by extending our intelligence with our intelligent technology that we can handle the scale of complexity to address these challenges.
wanted to debate a critic who said look at this picture of cerebral cortex. It is too complex. Turns out the picture was of a simulation of the cortex.
Reverse engineering hearing and speech engineering helped boost development of artificial hearing and speech recognition.
In 2006 it happened for vision as well.
We will not mindlessly copy but gain insights, basic principles, counter intuitive from reverse engineering and then build and amplify upon them.
This level of power will be very cheap by the end of the decade and be boosted by cloud computing.
Exponential growth has happens and continues to happen in information technologies.
The philosophical questions will become issues as we approach the singularity.
Why do some immediately grasp and get Singularity concept and why are some unable to emotionally accept
not related to education, achievement, intelligence, not about amount exposure to the ideas
Kurzweil has no good reason
Has comparisons and references related to an older talk (and debate he had with Gregory Stock) of Gregory Stock (where Greg was saying it could not happen instead of the new talk that it could happen but that human values will not be preserved).
Ray is showing the updated exponential slides updated to 2009.
Reverse engineering the brain and the mind.
Reverse engineering the brain Research into the performance of the brain
Processing is not the hard problem in the brain, it is the abstractions and going from objectivity to subjectivity.
All of these expand the AI tool kit
The processing is not the hard part, it is being able to go to abstractions from objectivity to subjectivity.
The cerebral cortex is the key.
A perfect simulation of the human brain or cortex would not do anything until you taught it things. Although prepping a usable knowledge base could be done in parallel.
human DNA has 800 million bytes which can be losslessly compressed to 50 million bytes. 3% is coding.
So how does that create trillions of connections of the brain
Ideas – hierarchy of symbols and symbol combinations and nested groups.
the cerebral cortex is the only part of the brain to deal with hierarchies. Only mammals have it.
The low level modules of the cerebral cortex are pattern recognizers.
There are other modules at the higher level and it can recognize humor and higher concepts. The differences are not great from the low level ones.
the modules are organized like lists. Thus you memorize the alphabet in one direction. If you go backwards it makes a separate list. LISP language used this.
Compares AI boom bust and the internet dot.com boom bust.
AI business still around but transformed.
We have an understanding of how the list modules in the brain work but are learning still how they are networked together.
Wiring is a weakness of the human brain. but we need to understand how they work and the strengths of the brain system.
Need to follow the principles and develop on a different substrate and then extend and use virtual connections and enable better capabilities and performance.
Amygdela (cells that span the brain and drive emotion)
Consciousness. An interest of Kurzweil for 50 years.
recent papers that suggests microtubules or quantum computing. But no evidence provided for this. There is no objectively measurable phenomena. No direct measurement of subjective experience.
Our whole moral and legal system is based on consciousness. Thus it is not a good idea to just say consciousness is an illusion.
You can have tests which could convince others that something appears to be exhibiting consciousness but you do not know that it does.
There is a leap of faith that everyone has about consciousness and who or what is conscious. Different leaps of faith.
Kurzweil has the assumption that if it passes a valid Turing test then he will assume that they are.
Cerebellum module repeated ten billion times
Smartphones and internet enhance the intelligence of people. We will go beyond that with embedded devices,
Consciousness study is like pealing an onion. The question is whether there will be anything left after you finish pealing.
Ray indicates that the goal should be understand the key principles of brain operation. We do not have to go to the full details of how it actually works to the ion channel level, just enough to get the functional processing and principles right to make systems that work but physically different. Once we are getting the results that we want then we may not need to go to the lower and lower detail work.
Neural nets and genetic algorithms are part of the grab bag of AI techniques. Need to look at improvements that we can make with the actual neural nets of the actual brain