Why robots will be smarter than humans by 2029 +/- months

Technological Singularity

Today I came across an article posted on Robohub.org titled, Why robots will not be smarter than humans by 2029. The author, Alan Winfield, points out several reasons why. Displeased with his views I thought I’d comment on his comments.

Dr. Winfield,
After reviewing your comments, it’s clear you do not understand the technological singularity thesis. Yes, in the past few days we’ve seen a spate of headlines boasting about Ray Kurzweil’s 2029 date for the birth of true Artificial Intelligence (AI) (Technological Singularity). As much as I respect your credentials, when it comes to your article, Why robots will not be smarter than humans by 2029, I think you are profoundly wrong, and here are several point-to-point reasons why:

AW: What exactly does as-smart-as-humans mean?
Winfield, you seem to think AI will not have emotions and therefore will not be emotionally intelligent and therefore will not be as-smart-as-humans. Emotions are just subjective body states, just another element to be modeled algorithmically. Plenty of research is being done on the cognitive architecture of emotions, and many AI researchers believe one day AI and robots will have emotions. This is not a farfetched notion – have you not seen the film “her”? In fact, one day, not only will a robot will say “I think therefore I am”, a robot will say, “I feel therefore I emote”. With new paradigm of AI, an embodied paradigm, which, let’s call the new robotic revolution, robots and AI will have emotions and use their emotions and reason, much in the same way we do, and at the end of this article I will provide a brief overview of the element that will make all this possible.

AW: Human intelligence is embodied.
Rolf Pfeifer & Josh Bongard’s book, How the Body Shapes the Way We Think, is outstanding, here you are right. However, instead of using their book as a reference point to prove why human-like intelligence is not possible, I see their book as a great reference on how to make it possible. Winfield, yes you are right, a whole new approach is needed to achieve human-like intelligence and Pfeifer & Bongard’s book does lay out some of these new approaches. Again, later I will explain these new approaches.

AW: As-smart-as-humans probably doesn’t mean as-smart-as newborn babies, or even two-year-old infants.
Winfield, here you are very wrong. As-smart-as-humans does mean as-smart-as a newborn. By this I mean, do not underestimate the intelligence of infants. From a cognitive perspective what they can achieve is far greater than what most robots can do today. When we create our infant AI/robot this will be the cusp of the singularity. In 2029 +/- months we will see the birth of a digital baby with the cognitive functions that equal a flesh-and-blood infant. Winfield, pardon my pun, but yes, developmental robotics is in its infancy, but 15 years in this technological world is a long long time, long enough for a technology in its infancy to develop. This will be explained, soon.

AW: Moore’s Law will not help.
Winfield, wrong again. Yes, building human-equivalent robot intelligence needs far more than just lots of computing power, and every AI enthusiast knows this very well. We know it will take more than Moore’s Law to achieve human-like intelligence, and the new robotic revolution will work in conjunction with Moore’s Law to see this happen.

AW: The hard problem of learning and the even harder problem of consciousness.
Learning is not that hard, we have come a long way from the old days of the single perceptron. We have neural nets that have amazing learning capabilities. The even the harder problem of consciousness is not as much as a mystery as it once was. Perhaps part of the problem is that you think we humans are the only creatures with a conscious mind? But consciousness is a gradient; it’s not a discrete function. Consciousness, like emotions, can be modeled and can be designed. Sorry, a philosophical-zombie-bot is unacceptable. Furthermore, using Noonian Soong as an example of why AI is not possible by 2029 is just bad math. In a virtual world time is very relevant. One of our years can be lived out in days, hours, if not minutes, with a fast enough computer. This is another reason why Moore’s Law is so important and still relevant.

LL: The new robotic revolution
Above, I have mentioned the new robotic revolution many times and mentioned this is a new paradigm of AI and robotics. This paradigm is unfolding right now. It consists of mutable new perspectives on robotics and AI, that is to say, GOFAI is gone forever. This is the dawn of the embodied age of AI, and here are some key elements:

  1. Moore’s Law – is stable as ever and is still a key factor for Kurzweil’s 2029 prediction. The skeptics are very quick to remind us that we are almost at the physical limits of current chip design. But there are new chip architectures that are only at their beginning of their design and development stage.

    Moore’s Law will also apply to the cognitive computing processors designed to process in parallel – this is an important piece of the puzzle. This will allow us to run the new embodied cognitive architectures.

  2. New Cognitive Architectures – projects like the Connectome and other projects taking an embodied perceptive give us a much better understanding of the architecture of the brain and its relationship with the mind and its environment. New models of cognition are being develop based on embodiment and are being modeled in virtual environments.
  3. Virtual Autonomous Learning – the virtual is a world of endless possibilities. In the virtual world, time is just another variable to be manipulated. We know that autonomous learning and development is another key factor. The time scales in virtual environments are not the same as in real environments. Also, virtual agents are being modeled based on human morphology.
  4. Organic Morphology – robots of tomorrow will be built based on the virtual human models, with bones and muscles (as opposed to motors and actuator). This is now possible with 3D printing. We have the capability to print robots with organic-like morphology using different types of materials to simulate the hard parts and the soft parts of the human body. These new robots will have the same affordance as humans, hence the same organic movements.

Embodied cognition the body/mind/environment perspective is theoretical foundation for the New Robotic Revolution and is the future. So, Alan Winfield, I am not sure if you really understand the notion of exponential technology. If you did, perhaps you would not be so pessimistic with your views on the future of AI. Actually, it’s not even about being optimistic or pessimistic; it’s about understanding that technological development is not linear. It’s exponential. That is why robots will be smarter than humans by 2029 +/- months.

Post-amble a good and very relevant comparison is the Human Genome project. This project started in 1990 and was expected to be completed in 15 years. However, after 8 years, only a small percentage of the genome was actually decoded. Then in 1998, Craig Venter’s company, Celera, started to use new technologies that expedited the decoding process. Then, only a few years after, Celera announced that the first draft of the human genome was completed – years ahead of schedule. The main point here is, with regards to achieving true AI by 2029, don’t rule out the possibility of it happening even before the race has begun. And make no mistake, this is a race. Fifteen years is a long time in this technological age. This is just the beginning of the New Robotic Revolution. In fact, even if the naysayers are still say nay in 2028, don’t be surprised if in 2029 +/- months the birth of a virtual autonomous learner is born.

The last confounding variable I will mention that can contribute towards the goal for the technological singularity accruing in 2029 is the media. The fact that AI is getting more and more media attention only means it will get more and more $$$. The fact that big players like Google and IBM are in the race only means more and more money, and more and more talent, and hence more and more synergy. That is what the technological singularity is all about. If there is one thing we can be sure of, with regards to what we humans can collectively achieve, is, when we put enough money and will into something we can achieve the seemly impossible.

– – –

Tags: ,

Leave a Reply