Nuclear security, AI and the future

The western world is taken over by a storm and surge which is so overwhelming that everybody or who’s who of any significance are talking about it. This is the Artificial intelligence (AI), a revolution in the making. The research on this concept started in early 1950, but later few decades were lost to the tacit oblivion, a sort of Artificial intelligence winters, in which no significant progress was made till 1980s and 1990s when the private entrepreneur and academia started taking interest without state sponsorship.

Came 21st century and everything changed. Sofia (first female robot working on AI) is given the nationality of Saudi Arabia, trump orchestrated a new concept of everything, Russia is awaiting a resurgence, China is eager to lead in the complete spectrum of technological advancement, India is transforming its middle class by emerging as a cyber power and Pakistan being part of the equation with all of the above countries is trying to achieve a strategic equilibrium. The new equation of balance, counter balance and strategic outreach is based on the emerging technologies based on AI, Machine thinking and deep thinking.

Hirari’s book the homodeus, a special issue of economist published in March 2018 and very recent mind boggling article by henry Kissinger on AI is just a tiny part of literature which is burgeoning with every passing day. The RAND Corporation of USA and SIPRI of Sweden are spearheading the international efforts in the realm of academia to create an awareness thereby straightening few facts. What is actually the concept of AI and how it works, if one understands this, half of the commotion, the fear and the apprehensions are over? Rest of the half will remain, as the next five years are important to further clear the fog.

AI is simple concept with far reaching consequences. It is based on neural networks, designed, replicating human brain. This network functions through layers, first layer identity edges, next see the corners, angles, shapes, structures and so on. The process culminates at retuning. This generates a learning algorithm exposed to terabytes of data, hundreds of thousands of images and years’ worth of speech. The neural network learns through experience. These are good at recognizing pattern but they cannot reason. Learning through experience without identifying logic is learning like a child.

The most promising thing of AI is that it learns through experience like a child, which is not logic. Logic has interpretations and comes at later stages of cognizant development. Strategy, tactics, philosophy & policy are all logical-interpretations. Logic had most of the time left humanity in sheer confusion. AI learning is sequenced through experience, will be countered with more experience and that is how deep and machine learning will be acquired. AI is more than a technology. AI is the new electricity as it is said and as electricity electrified everything, AI is soon going to cognitize everything. Till 20th century, world was governed by hierarchies of different orders. Even enlightenment and renaissance supported the liberal and enlightening hierarchies. Late 20th century and early 21st century saw the emergence of networks. The globalization ride the tide of networks.

Now we are seeing a mashroom growth of incubation centres and incubators, the entrepreneur startups, most of the time outside the state and academia’s control and supervision. CRISPR can now edit the genes of any living thing. In couple of years it will be available as a household machine. The vertical farms and cellular Agriculture is transforming the concept of food. The 3D printers or additive manufacturing (AM) can produce almost anything in a go. This all is happening through AI in small incubators manned by lean teams of researchers spread out globally. A Cambrian Explosion is in the making. AI is shifting from its incremental to the breakout phase when the chink in the armour will be expanded significantly. It is a omni-use technology being wielded by private sector and technical academia (General academia was till aloof till Alpha Go appeared, who defeated human in the game Go, far complexed than chess). Deep learning is taking place with a technique called reinforcement learning, it can be a supervised or unsupervised experience. Humanity is passing through, a neuro evolution and like the debate of Kenneth waltz and Scott Sagan’s nuclear pessimist and optimists, there are AI pessimist and AI optimists.

A question arises that if AI is a omni use technology, it is crosscutting to a level that it will be everywhere inside every known machine and human activity, than how will it affect the nuclear security of a country, specially a country like Pakistan whose history of acquiring technology is in great semblance to the concept of AI, through experience. Before dissecting this frog on the table, there is an interesting Phenomenon, almost all the present day leading AI scientists of the world are of Chinese origin with nationalities of host countries. Li Deng works for Microsoft, Fi Fi lee heads AI at Stanford, Andrew ng (US trained) heads Baido in China and Peta lee works for micro soft. It is not intentional, rather coincidental, as when in early nineties, Chinese were reaching western universities to learn computing, AI was a subject coming out of its winter doldrums. All of these AI scientists remained students of people like Geoffrey Hanton and Judea Pearl who were neural net pioneers. Apple’s siri, googl’s google now, Amazon’s Alexa, facebook’s chat bot, IBM’s Watson and Microsoft cortana were made possible through these people and their teams.

Once all this was happening world was watching with a casual intent. Now private corporations like google, Amazon, facebook etc are way ahead from the most advanced countries in terms of AI research. Even google has started using tailor made chips called tensors for AI applications. No one can deny this singularity happening all around us which dictates that when super intelligent machines start improving themselves with human involvement, triggering a runaway cycle that leaves human in dust, earthly or cosmic, that time will decide. There can be three scenarios of future in terms of AI predominance, one, human in the loop (every step controlled), second, Human on the loop (Machine works independently but under human control), Human out of the loop (No human interference). Third one is a dangerous scenario, here comes the direct intervention of states to harness this technology so that it remains at the first two steps of its actualization.

Lethal autonomous weapons, the swarming drones, Russian Killer robot-status 6, the continental missiles with standalone AI computers on board, the centrifuges with the deep thinking mechanism, the pilotless bomber and fighters or what not will seriously affect the concept of nuclear security. The academia dealing with nuclear strategy, nuclear policies and the nuclear politics are not buying this idea as being too dangerous to realize. But once AI is going to be part of every machine, how nuclear paraphernalia will not be effected. The trajectory of AI has divided the nuclear experts in three Categories, one, the complacent who think that no nuclear nation will involve AI in the nuclear decision making, second, the alarmist, who say that whether you involve or not the strategic equilibrium of which nuclear weapons are a part will be disturbed up to a level that deterrence in all its form will fail. Third, the subversionists who say that you cannot remain aloof, acquire as much AI for these system as the adversary is doing so that like all technologies a balance of AI, if one may cal,l can be achieved. Factually, AI could severely compromise nuclear strategic stability and will increase the risk of war due to, resurgent Russia, Rising China, successful limited use and regional nuclear rivalries.

AI if is incorporated in delivery systems and C4ISR (command, control, communication computer, intelligence, surveillance and reconnaissance) than humanity will be at a third scenario where man is out of the loop and into somewhere. Historically the bipolarity of cold war encouraged both crisis and arms race stability. There are nine nuclear weapon states and multiple strategic rivalries, indirectly effecting each other.

USA worries about Russia and China, Russia is in competition with both China & USA. India is in strategic competition with Pakistan and China. With the application of AI to the international nuclear architecture, all the three stabilities i.e. first strike stability, crisis stability and arms race stability will be gone. AI challenges all three simultaneously. Not only this, the goals of any nuclear strategy, coercion, deterrence, compellence, assurance and reassurance will also become topsy turvy with no up side up. A nuclear power without central deterrence and extended deterrence will be in a perpetual state of paranoia at the mercy of some Sofia, Alexa or Watson. 3D printing or Additive manufacturing (AM) now poses a major challenge to international export control efforts under NSG, Wassenear arrangement, Australia group or MTCR.

Still if the experts on nuclear strategy think that AI will not affect the nuclear architecture, then only one has to wait for Alpha Go to play the nuclear Jijitsu but at what cost. It is commonly said that there should be laws to counter the LAWS (Lethal automatons weapon system). In fact, Article 36 of 1977 Additional protocol to 1949 Geneva Convention imposes a practical obligation on states to determine whether study, development, acquisition or adoption of new weapons, means or methods of warfare its use would in some or all circumstances be prohibited by international law.

Under Article 36 this is the time that new laws etc. be formulated. Pakistan along with being part of these future enactments should also develop its Institutions of AI excellence like USA’s defence advance research agency and office of security of defence for research & engineering. As Moore’s law is ticking like a time bomb so there is not a second to be wasted in complacency.

The writer is a free-lance who contributes regularly on security related issues. He is also a Ph. D Scholar.

Leave a Reply

Your email address will not be published. Required fields are marked *