LIFE 3.0: Being Human in the Age of Artificial Intelligence

Max Tegmark’s latest book, LIFE 3.0: Being Human in the Age of Artificial Intelligence, is an enthralling journey into the future, when the developments in artificial intelligence create a new type of lifeform on Earth.

Tegmark proposes to classify life in three stages. Life 1.0, unintelligent life, is able to change its hardware and improve itself only through the very slow and blind process of natural evolution. Single cell organisms, plants and simple animals are in this category. Life 2.0 is also unable to change its hardware (excepto through evolution, as for Life 1.0) but can change its software, stored in the brains, by using previous experience to learn new behaviors. Higher animals and humans, in particular, belong here. Humans can now, up to a limited point, change their hardware (through prosthetics, cellphones, computers and other devices) so they could also be considered now Life 2.1.

Life 3.0 is the new generation of life, which can change both its software and its hardware. The ability to change the computational support (i.e., the physical basis of computation) results from technological advances, which will only accelerate with the advent of Artificial General Intelligence (AGI). The book is really about the future of a world where AGI enables humanity to create a whole range of new technologies, and expand new forms of life through the cosmos.

The riveting prelude, The Tale of the Omega Team, is the story of the group of people who “created” the first intelligence explosion on planet Earth makes this a “hard-to-put-down” book.  The rest of the book goes through the consequences of this intelligence explosion, a phenomenon the author believes will undoubtedly take place, sooner or later. Chapter 4 focus on the explosion proper, and on how it could happen. Chapter 5, appropriately titled “Aftermath: The Next 10,000 Years” is one of the most interesting ones, and describes a number of long term scenarios that could result from such an event. These scenarios range from a benevolent and enlightened dictatorship (by the AI) to the enslaved God situation, where humanity keeps the AI in chains and uses it as a slave to develop new technologies, inaccessible to unaided humanity’s simpler minds. Always present, in these scenarios, are the risks of a hostile takeover by a human-created AGI, a theme that this book also addresses in depth, following on the ideas proposed by Nick Bostrom, in his book Superintelligence.

Being a cosmologist, Tegmark could not leave out the question of how life can spread through the Cosmos, a topic covered in depth in chapter 6, in a highly speculative fashion. Tegmark’s view is, to say the least, grandiose, envisaging a future where AGI will make it possible to spread life through the reachable universe, climbing the three levels of the Kardashev scale. The final chapters address (in a necessarily more superficial manner) the complex topics of goal setting for AI systems and artificial (or natural) consciousness. These topics somehow felt less well developed and more complete and convincing treatments can be found elsewhere. The book ends with a description of the mission of the Future of Life Institute, and the Asilomar AI Principles.

A book like this cannot leave anyone indifferent, and you will be likely to take one of two opposite sides: the optimistis, with many famous representatives, including Elon Mush, Stuart Russel and Nick Bostrom, who believe AGI can be developed and used to make humanity prosper; or the pessimists , whose more visible member is probably Yuval Noah Harari, who has voiced very serious concerns about technology developments in his book Homo Deus and in this review of Life 3.0.

Advertisements

AlphaZero masters the game of Chess

DeepMind, a company that was acquired by Google, made headlines when the program AlphaGo Zero managed to become the best Go player in the world, without using any human knowledge, a feat reported in this blog less than two months ago.

Now, just a few weeks after that result, DeepMind reports, in an article posted in arXiv.org, that the program AlphaZero obtained a similar result for the game of chess.

Computer programs have been the world’s best players for a long time now, basically since Deep Blue defeated the reigning world champion, Garry Kasparov, in 1997. Deep Blue, as almost all the other top chess programs, was deeply specialized in chess, and played the game using handcrafted position evaluation functions (based on grand-master games) coupled with deep search methods. Deep Blue evaluated more than 200 million positions per second, using a very deep search (between 6 and 8 moves, sometimes more) to identify the best possible move.

Modern computer programs use a similar approach, and have attained super-human levels, with the best programs (Komodo and Stockfish) reaching a Elo Rating higher than 3300. The best human players have Elo Ratings between 2800 and 2900. This difference implies that they have less than a one in ten chance of beating the top chess programs, since a difference of 366 points in Elo Rating (anywhere in the scale) mean a probability of winning of 90%, for the most ranked player.

In contrast, AlphaZero learned the game without using any human generated knowledge, by simply playing against another copy of itself, the same approach used by AlphaGo Zero. As the authors describe, AlphaZero learned to play at super-human level, systematically beating the best existing chess program (Stockfish), and in the process rediscovering centuries of human-generated knowledge, such as common opening moves (Ruy Lopez, Sicilian, French and Reti, among others).

The flexibility of AlphaZero (which also learned to play Go and Shogi) provides convincing evidence that general purpose learners are within the reach of the technology. As a side note, the author of this blog, who was a fairly decent chess player in his youth, reached an Elo Rating of 2000. This means that he has less than a one in ten chance of beating someone with a rating of 2400 who has less than a one in ten chance of beating the world champion who has less than a one in ten chance of beating AlphaZero. Quite humbling…

Image by David Lapetina, available at Wikimedia Commons.

Portuguese Edition of The Digital Mind

IST Press, the publisher of Instituto Superior Técnico, just published the Portuguese edition of The Digital Mind, originally published by MIT Press.

The Portuguese edition, translated by Jorge Pereirinha Pires, follow the same organization and has been reviewed by a number of sources. The back-cover reviews are by Pedro Domingos, Srinivas Devadas, Pedro Guedes de Oliveira and Francisco Veloso.

A pre-publication was made by the Público newspaper, under the title Até que mundos digitais nos levará o efeito da Rainha Vermelha, making the first chapter of the book publicly available.

There are also some publicly available reviews and pieces about this edition, including an episode of a podcast and a review in the radio.

The last invention of humanity

Irving John Good was a British mathematician who worked with Alan Turing in the famous Hut 8 of Bletchley Park, contributing to the war effort by decrypting the messages coded by the German enigma machines. After that, he became a professor at Virginia Tech and, later in life, he was a consultant for the cult movie 2001: A Space Odyssey, by Stanley Kubrick.

Irving John Good (born Isadore Jacob Gudak to a Polish jewish family) is credited with coining the term intelligence explosion, to refer to the possibility that a super-intelligent system may, one day, be able to design an even more intelligent successor. In his own words:

Let an ultraintelligent machine be defined as a machine that can far surpass all the intellectual activities of any man however clever. Since the design of machines is one of these intellectual activities, an ultraintelligent machine could design even better machines; there would then unquestionably be an ‘intelligence explosion,’ and the intelligence of man would be left far behind. Thus the first ultraintelligent machine is the last invention that man need ever make, provided that the machine is docile enough to tell us how to keep it under control.

We are still very far from being able to design an artificially intelligent (AI)  system that is smart enough to design and code even better AI systems. Our current efforts address very narrow fields, and obtain systems that do not have the general intelligence required to create the phenomenon I. J. Good was referring to. However, in some very restrict domains, we can see at work mechanisms that resemble the that very same phenomenon.

Go is a board game, very difficult to master because of the huge number of possible games and high number of possible moves at each position. Given the complexity of the game, branch and bound approaches could not be used, until recently, to derive good playing strategies. Until only a few years ago, it was believed that it would take decades to create a program that would master the game of Go, at a level comparable with the best human players.

In January 2016, DeepMind, an AI startup (which was at that time acquired by Google by a sum reported to exceed 500M dollars), reported in an article in Nature that they had managed to master the complex game of Go by using deep neural networks and a tree search engine. The system, called AlphaGo, was trained on databases of human games and eventually managed to soundly beat the best human players, becoming the best player in the world, as reported in this blog.

A couple of weeks ago, in October of 2017, DeepMind reported, in a second article in Nature, that they programmed a system, which became even more proficient at the game, that mastered the game without using any human knowledge. AlphaGo Zero did not use any human games to acquire knowledge about the game. Instead, it played millions of games (close to 30 millions, in fact, played over a period of 40 days) against another version of itself, eventually acquiring knowledge about tactics and strategies that have been slowly created by the human race for more than two millennia. By simply playing against itself, the system went from a child level (random moves) to a novice level to a world champion level. AlphaGo Zero steamrolled the original AlphaGo by 100 to 0,  showing that it is possible to obtain super-human strength without using any human generated knowledge.

In a way, the computer improved itself, by simply playing against itself until it reached perfection. Irving John Good, who died in 2009, would have liked to see this invention of mankind. Which will not be the last, yet…

Picture credits: Go board, picture taken by Hoge Rielen, available at Wikimedia Commons.

 

AIs running wild at Facebook? Not yet, not even close!

Much was written about two Artificial Intelligence systems developing their own language. Headlines like “Facebook shuts down down AI after it invents its own creepy language” and “Facebook engineers panic, pull plug on AI after bots develop their own language” were all over the place, seeming to imply that we were just at the verge of a significant incident in AI research.

As it happens, nothing significant really happened, and these headlines are only due to the inordinate appetite of the media for catastrophic news. Most AI systems currently under development have narrow application domains, and do not have the capabilities to develop their own general strategies, languages, or motivations.

To be fair, many AI systems do develop their own language. Whenever a neural network is trained to perform pattern recognition, for instance, a specific internal representation is chosen by the network to internally encode specific features of the pattern under analysis. When everything goes smoothly, these internal representations correspond to important concepts in the patterns under analysis (a wheel of car, say, or an eye) and are combined by the neural network to provide the output of interest. In fact, creating these internal representations, which, in a way, correspond to concepts in a language, is exactly one of the most interesting features of neural networks, and of deep neural networks, in particular.

Therefore, systems creating their own languages are nothing new, really. What happened with the Facebook agents that made the news was that two systems were being trained using a specific algorithm, a generative adversarial network. When this training method is used, two systems are trained against each other. The idea is that system A tries to make the task of system B more difficult and vice-versa. In this way, both systems evolve towards becoming better at their respective tasks, whatever they are. As this post clearly describes, the two systems were being trained at a specific negotiation task, and they communicated using English words. As the systems evolved, the systems started to use non-conventional combinations of words to exchange their information, leading to the seemingly strange language exchanges that led to the scary headlines, such as this one:

Bob: I can i i everything else

Alice: balls have zero to me to me to me to me to me to me to me to me to

Bob: you i everything else

Alice: balls have a ball to me to me to me to me to me to me to me to me

Strange as this exchange may look, nothing out of the ordinary was really happening. The neural network training algorithms were simply finding concept representations which were used by the agents to communicate their intentions in this specific negotiation task (which involved exchanging balls and other items).

The experience was stopped not because Facebook was afraid that some runaway explosive intelligence process was underway, but because the objective was to have the agents use plain English, and not a made up language.

Image: Picture taken at the Institute for Systems and Robotics of Técnico Lisboa, courtesy of IST.

Stuart Russell and Sam Harris on The Dawn of Artificial Intelligence

In one of the latest episodes of his interesting podcast, Waking Up , Sam Harris discusses with Stuart Russell the future of Artificial Intelligence (AI).

Stuart Russel is one of the foremost world authorities on AI, and author of the most widely used textbook on the subject, Artificial Intelligence, a Modern Approach. Interestingly, most of the (very interesting) conversation focuses not so much on the potential of AI, but on the potential dangers of the technology.

Many AI researchers have dismissed offhand the worries many people have expressed over the possibility of runaway Artificial Intelligence. In fact, most active researchers know very well that most of the time is spent worrying about the convergence of algorithms, the lack of efficiency of training methods, or in difficult searches for the right architecture for some narrow problem. AI researchers spend no time at all worrying about the possibility that the systems they are developing will, suddenly, become too intelligent and a danger to humanity.

On the other hand, famous philosophers, scientists and entrepreneurs, such as Elon Musk, Richard Dawkins, Bill Gates, and Nick Bostrom have been very vocal about the possibility that man-made AI systems may one day run amok and become a danger to humanity.

From this duality one is led to believe that only people who are away from the field really worry about the possibility of dangerous super-intelligences. People inside the field pay little or no attention to that possibility and, in many cases, consider these worries baseless and misinformed.

That is why this podcast, with the participation of Stuart Russell, is interesting and well worth hearing. Russell cannot be accused of being an outsider to the field of AI, and yet his latest interests are focused on the problem of making sure that future AIs will have their objectives closely allied with those of the human race.

The wealth of humans: work and its absence in the twenty-first century

The Wealth of Humans, by Ryan Avent, a senior editor at The Economist, addresses the economic and social challenges imposed on societies by the rapid development of digital technologies.  Although the book includes an analysis of the mechanisms, technologies, and effects that may lead to massive unemployment, brought by the emergence of digital technologies, intelligent systems, and smart robots, the focus is on the economic and social effects of those technologies.

The main point Avent makes is that market mechanisms may be relied upon to create growth and wealth for society, and to improve the average condition of humans, but cannot be relied upon to ensure adequate redistribution of the generated wealth. Left to themselves, the markets will tend to concentrate wealth. This happened in the industrial revolution, but society adapted (unions, welfare, education) to ensure that adequate redistribution mechanisms were put in place.

To Avent, this tendency towards increased income asymmetry, between the top earners and the rest, which is already so clear, will only be made worst by the inevitable glut of labor that will be created by digital technologies and artificial intelligence.

There are many possible redistribution mechanisms, from universal basic income to minimum wage requirements but, as the author points out, none is guaranteed to work well in a society where a large majority of people may become unable to find work. The largest and most important asymmetry that remains is, probably, the asymmetry that exists between developed countries and underdeveloped ones. Although this asymmetry was somewhat reduced by the recent economic development of the BRIC countries, Avent believes that was a one time event that will not reoccur.

Avent points out that the strength of the developed economies is not a direct consequence of the factors that are most commonly thought to be decisive: more capital, adequate infrastructures, and better education. These factors do indeed play a role but what makes the decisive difference is “social capital”, the set of rules shared by members of developed societies that makes them more effective at creating value for themselves and for society. Social capital, the unwritten set of rules that make it possible to create value, in a society, in a country or in a company, cannot be easily copied, sold, or exported.

This social capital (which, interestingly, closely matches the idea of shared beliefs Yuval Harari describes in Sapiens) can be assimilated, by immigrants or new hires, who can learn how to contribute to the creation of wealth, and benefit from it. However, as countries and societies became adverse at receiving immigrants, and companies reduce workforces, social capital becomes more and more concentrated.

In the end, Avent concludes that no public policies, no known economic theories, are guaranteed to fix the problem of inequality, mass unemployment, and lack of redistribution. It comes down to society, as whole, i.e., to each one of us, to decide to be generous and altruistic, in order to make sure that the wealth created by the hidden hand of the market benefits all of mankind.

A must-read if you care about the effects of asymmetries in income distribution on societies.