27 Feb [Episode 2] Artificial intelligence from hype to reality

AI’s boom-and-bust cycles across its 70-year history show that predicting the moment machines will achieve human-level intelligence is not an easy thing to do. Be that as it may, techno-futurists are not afraid to claim it will happen soon. Ray Kurzweil, who has a track record of accurate predictions, is convinced that by 2029 Artificial Intelligence will pass a valid Turing test. The same Kurzweil believes that we will achieve Singularity by 2045.

In this new episode of our Artificial Intelligence from Hype to Reality video series, Professor Dan Cautis from Georgetown University and QUALITANCE Chief Innovation Officer Mike Parsons will zoom in on the arguments behind the theory of Singularity, what fuels Transhumanism and its radical claims – including the probability of uploading our mind into silicon, in case Singularity is not possible.

[Episode 2] On Transhumanism & Singularity

Besides breaking down the most notable radical claims made by techno-futurists, Professor Cautis will also take us back to how machines started to learn by themselves while running a close-up on some of DeepMind’s major achievements so that we can better understand what deep learning is all about.

Last but not least, you will find out what transhumanism is and how people have come to believe in it. In case you missed Episode 1 – A Brief History of AI, you might want to start there and continue with this one.  In the meantime, check out the highlights of Episode 2:

1. How Machine Learning came in and resurrected AI.
2. Case-in-point: DeepMind and its main achievements.
3. An introduction to Deep Learning.
4. Major breakthrough for AI: machines learning by themselves.
5. What Transhumanism is, how it gained spotlight and who promotes it. 
6. A thorough analysis of 3 radical claims made by techno-futurists.
7. Mind uploading: the backup plan of techno-futurists in case we will not achieve Singularity.  

Recommended readings and notes

DeepMind and deep learning

The past 7 decades, Artificial Intelligence has lived through alternate times of glory and scorn, widely known as AI summers and winters. The summers were filled with optimism and backed by massive fundings, whereas the winters were marked by big cuts, disbelief and pessimism. In case you want to learn more, check out this short review of its advances.  

The second winter of AI (1988-2011), however, was mainly overcome with the help of Machine Learning. In fact, it was DeepMind’s breakthroughs in deep learning that fueled the hype and eventually rescued AI from a true valley of darkness. DeepMind is a British Artificial Intelligence company founded in 2010 that Google acquired in 2014, mainly for its extraordinary research in deep learning.

Based on seven Atari games, DeepMind developed a deep learning methodology, which enabled machines to play against themselves. Since they were programmed to learn from the games, the machines simply got better and better.

Most likely you’ve already heard of their Artificial Intelligence dubbed AlphaGo A.I. defeating Korean champion Lee Sedol in the complex Chinese strategy game of GO in 2016. In the following year, Google developed AlphaGo Zero, a new version of the original AI (aka AlphaGo A.I.). This new A.I acquired a superhuman-level Go-playing skill and beat the original AlphaGo A.I. 100 to 0. Actually, 2017 got even better for DeepMind!

Computationalism vs Connectionism

DeepMind marked a major shift from computationalism to deep learning, from applying a clear set of logic rules to acquiring massive amounts of data through connectionism.

If you’re wondering how it was possible for machines to become so skillful, you need to think of neurons. Think of machines as having neuron-like structures, i.e. connected in the same way neurons are connected in the brain. Don’t forget that, before DeepMind, machines relied mostly on computationalism.

Computationalism, which you can find at the heart of classical AI applications such as expert systems, used deductive logic. In spite of being successful in many applications, the rigid use of logical rules did not help machines learn so much, so fast. Because the system only worked with 2 variables – “true” or “false”, it had no place for uncertainties. The use of Machine Learning, however, brought a whole new perspective. Instead of applying rigid, logical rules to symbols, the new approach makes use of neural networks.

In fact, the large number of connections between neurons was the inspiration behind the term “connectionism.” A connectionist approach comes with many advantages. You can find them in the countless successful applications of neural networks using deep learning structures with numerous layers of neurons.

Nonetheless, the lack of transparency at the level of connections is affecting the algorithms. Even though the machines are getting amazing results, it’s still not clear how they achieve them. The logical systems used in computationalism had transparent rules, which allowed for a rational explanation of the results. In connectionism, the lack of transparency does not provide any explanation, but the progress is visible.

The fact that machines started to learn by themselves gave people hope that we can indeed achieve Artificial General Intelligence, which according to transhumanists is a precondition to accomplishing Singularity.

Transhumanism & Singularity

According to Britannica, transhumanism is a “social and philosophical movement devoted to promoting the research and development of robust human-enhancement technologies. Such technologies would augment or increase human sensory reception, emotive ability, or cognitive capacity as well as radically improve human health and extend human life spans. The modifications resulting from the addition of biological or physical technologies would be more or less permanent and integrated into the human body.”

This concept might sound new, but the ideology of this very large and active movement dates back to the beginning of times. For instance, Greek mythology often refers to humans trying everything in their power to outsmart the gods and overcome their otherwise limited nature. In fact, if you think of it, Icarus and Daedalus are among the first transhumanists.

The concept of transhumanism has been reinforced by every industrial revolution. In a way, history has been repeating itself and now we are challenging the gods again – by creating a kind of intelligence that outpasses the capacities of the human mind. According to Ray Kurzweil, one of the most prolific writers on futurism, Singularity is “the moment we will multiply our effective intelligence a billion fold by merging with the intelligence we have created.”

As a matter of fact, Ray Kurzweil advanced the idea that Singularity is possible back in 1999, when he published “The Age of Spiritual Machines” and introduced the Law of Accelerating Returns. For Kurzweil, Singularity is not a story of science fiction; it is enabled by the technological advances which accelerate exponentially. Hence, he felt this is reason enough for us to witness Singularity in 2045.

Transhumanism has been attracting a lot of people with different backgrounds and interests. Hans Moravec, Nick Bostrom and Callum Chase are only some of the most notable transhumanists and promoters of Singularity. For a deeper understanding of their radical claims, the following books on transhumanism and Singularity might come in handy:

• Hans Moravec’s “Mind Children”
• Nick Bostrom’s “Superintelligence”
• Callum Chase’s “Artificial Intelligence and the Two Singularities”
• Max More and Natasha Vita’s “The Transhumanist Reader”
• Ray Kurzweil’s “The Singularity Is Near: When Humans Transcend Biology”

Last but not least, techno-futurists are known for worshipping the mind. One could say that they even despise the body, because it deteriorates and gets old in time. In their opinion, the mind is everything. Hence, we should preserve it, especially in case Singularity is not possible. One way to do that is through whole brain emulation or mind uploading. To find out the technical process of uploading our mind into silicon, we recommend that you watch Professor Cautis explain it in the video above starting 40:40.

The first two episodes of our series revealed the major achievements of AI and the enthusiasm they sparked among the promoters of transhumanism. Professor Cautis also introduced us to the main assumptions behind the Singularity theory. Stand by for Episode 3, as we’ll be watching him challenge all these assumptions from a scientific, technological and philosophical point of view. Equally important, our next episode will bring forth a critical review of Kurzweil’s claim that “Singularity is near.”