Tommi A. Vuorenmaa, PhD
It is playing out as expected: artificially intelligent programs take control over human decisions. Recently, the Google-owned AlphaGo-algorithm has not only beat the human Zen-masters in Go — a thousands of years old highly complex Chinese board game — but it is also paving the way to a mass-adoption of newly minted AI principles. By shifting away from the old-fashioned brute-force calculations to a set of smarter solution techniques that imitate the process of human decision making, computers are now a few steps closer to matching humans in creativity.
The Go-game sports a 19x19 grid structure with black and white stones strategically placed by two players. AlphaGo learns the optimal responses to an almost infinite number of theoretical moves — more than there are atoms in the universe, about sexvigintillion (or tredecilliard, 10^81). It is the record of past moves that make AlphaGo smart, continuously learning from its records. Obviously, humans cannot match modern computers in their brute-force calculation power, and now that human creativity may become learned, there is only one major road block to deal with.
Enter blockchain: a network of computers using shrewdly crafted cryptographic algorithms that ensure the immutability of historical records, like a heavy stone ledger. Transparency has to be guaranteed, too, else AI is an untraceable “black box” — privacy ensuring data transparency is democracy. And once the “true data” are recorded in a blockchain — i.e. when the network is large — it also becomes less prone to any attacks, such as (distributed) denial of service attacks, than a normal, central database. Data in blockchains are themselves crypto-algorithm-secured.
By definition, data forms “a premise from which conclusions are drawn,” a view that is becoming slowly better appreciated. As the dystopias of the pop-culture movies “The Matrix” and “Blade Runner” become a more real threat, and the bad omens of Elon Musk, among others, start to gain momentum, data immutability and personal privacy enabling transparency should form the backbone of AI. Should the AI input data fall “in the hands of a few” (“the elite”), the risks of AI will start to dominate the benefits. The data input to all AI-algorithms must be true and testable.
As a simple example, should AlphaGo be trained with “Bug Data” instead of clean “Big Data” or, should the algorithm’s parameters be badly tuned, the erroneous decisions could go unnoticed, and lead to bad results. Naturally, analogously, a human could get nervous or sick, and perform badly. But that human inefficiency would be less affected by big data errors; furthermore, thanks to the limits of human capacity in saving and processing big data, humans are also more robust to ”Hack Data” and malfunction — so long as the incentives of humans are ethically acceptable.
It cannot be guaranteed the data saved to a blockchain are any more ethical than the data saved in human brains. Clearly, we cannot claim they are. But AI-algorithms could be programmed to create ethical records and verify them, and then these data, when saved to a robust blockchain data base, could form verified big data and serve as reliable input to decision making algorithms. The common goal of human beings must surround in building trustworthy and transparent data that are double-checked by AI-algorithms and then utilized as a helpful hand in decision making.
Aekraes Kodex believes that data must always be meticulously recorded, processed, and smartly analysed by well-crafted and tested algorithms. Decisions can be reached via consensus similar to the mechanisms certain modern blockchains already have in place. Aekraes Kodex promotes high ethical standards and open dialogue to guarantee democratic forces to prevail. The way to approach the machine age is to make sure humans will stay in control of the ultimate decisions; to make the AI-algorithms to train themselves and compete continuously like in the game of Go.
This article was written by a human, as is most likely apparent from the long, complex sentences. The author understands, however, that in the future it could well be recreated by an artificially intelligent “writing algorithm” trained using an immutable and transparent database of articles. What will hopefully be left of the true author is emotion. Likewise, a human reader of this blog should find it fulfilling to read long, complex sentences, rather than purely seeking information. Similarly, Aekraes Kodex builds blockchain applications that feel right and fulfil the users’ needs.