Cryptoamnesia, futures and risks

  

please put like if you want more news  

Community, field of interests includes artificial intelligence (AI) and futurological study of existential risk (X-risk), has recently shown an increasing interest in blockchain and cryptoanalysis systems. Ralf Merkle, the inventor of widely-known cryptographic technology that underlies lightweight Ethereum client, is actively interested in the management model of DAO. Co-founder of Skype Jaan Tallinn suggested blockchain technology as a possible mechanism for the solution of global coordination problems. The creators of prediction markets, which have long realised their potential as mechanisms of control, closely watching the Augur. Why is there such interest? Maybe just computer geeks, before actively interested in computer toy And suddenly not having it is irrelevant, but also beautiful and interesting toys? Maybe between these topics there is a real connection?

I believe that there is a connection. Both community cryptoamnesia, and researchers from AI/computer management/X-risks from the different parties approach the same problem: is it possible to regulate intelligent and complex systems that have unpredictable and spontaneously occurring properties with a simple and "blunt" instruments, with predetermined and constant parameters.

In the context of AI research, there is the problem of determining the function that monitors the behavior supercasanova agent, and thus does not allow to perform actions that formally satisfies a predetermined behavior, but it does not meet the intentions of the creators (Instantiation Edge – extreme solution). For example, if AI receives the instruction to cure cancer, it can come to the conclusion that the most reliable solution is just to kill everyone. If you are trying to patch this hole, he decides to be cryogenically freeze all instead of killing them directly. And so on (Remember the late V. I. Lenin: "the form is right, and in essence – mockery" approx.transl.).

In the work of Ralph Merkle on democracy DAO the problem is that the definition of the objective function correlates with social and technological progress, which society expects, but does not correlate with existential risk, and can be measured only as long as the rest of the measurement becomes the source of political battles.

In the context of cryptoeconomy problems are remarkably similar. The key question of consensus: how to motivate validators for the continued support and management of General accounting with a simple algorithm, "cast in granite," while the validators themselves are complex economic systems that interact with each other free way, and often having discordant or opposing objectives. The incident with TheDAO was the result of the divergence of the overall design of the software developers had in mind a particular function is split, and the result of the operation of the program code. Augur is working on the application of consensus to the facts of the real world. Maker creates a decentralized control algorithm platform, providing assets, a decentralized crypto currency and reliability of Fiat currency. In all these cases, control algorithms, "stupid", and the processes they need to control "smart". Security AI: an agent with an IQ of 150 is trying to control the agent with an IQ of 6000; cryptoamnesia: agents with an IQ of 5 control agents with an IQ of 150. Of course, this problem of different scales, but similarities should not be ignored.

Those are the challenges that both communities have been studying for many years independently from each other, and in that time has accumulated considerable experience. Solutions begin to use a heuristic partial solutions and risk mitigation strategies. In creating the DAO, some developers are inclined to the hybrid approach, which involves the appointment of mentors with a partial control over the assets of the DAO, but gives curators only limited power. Their capacity is sufficient to prevent the attack on the DAO, but not enough to own a full-scale attack – this approach is similar to research in the field of safe abortion AI.

On the side of futarchy we see interest in using interest rates as the objective function, the combination of futuree and quadratic voting by means of the voluntary freezing of funds, as the control algorithm (POS). This can be attributed to a variety of hybrid forms that convey futurei power, sufficient to prevent attacks of collusion of the majority," a way that was not democracy, and at the same time leave the power to the voting process. These innovations are, at least, deserve to be discussed in the group, considering Futari as democracy-building DAO.

Other, previously overlooked solution – the use of control mechanisms to explicitly slow down the development of undesirable processes: proposed hardwork who can pull funds from the Black DAO, was only possible because the contract TheDAO included a set of rules specifying deferred action. Another new area is formal verification. Is the use of programs that verify their purchases and other programs. The goal is to ensure that the results of the implementation of verifiable programs satisfy the purposes for which they were created.

A formal proof of "fairness" in the General case it is impossible due to the complexity of the cost (the assertion that human values have high Kolmogorov complexity). However, we use different ways to reduce the risk. For example, we can set the rule that any action can be taken after 7 days and the other actions conducted within 48 hours, provided that the curators of this DAO will vote a certain way. In the developer community AI such fuses are used to prevent various error compensation. Such errors can lead to completely unexpected behavior of the system due to the extremely high complexity of the value AI. Of course, research of methods of formal verification are performed by different communities for many years, but we now consider their application in the new environment.

Another concept popular in the field of AI safety which can be very useful in the construction of economic systems based on DAO – theory superrational solutions. This how to resolve situations in which there is "prisoner's dilemma", one solution is the software code that is more favorably drawn with agents executing this code. Example actions available to agents with open source, and available to agents that represent "black boxes" described in a little fiction story Scott Alexander: the two do not trust each other agent can jointly commit to the goal, which is the average of the goals that they set for themselves earlier. Until now, these concepts were the stuff of science fiction, but now future DAO allows you to translate them into practice. Summarizing, we can say that a DAO can be an effective instrument of social institutions capable of ensuring strict adherence to code with the desired properties.

The DAO is only the first of many that will start in this year and next year and there is no doubt that all the following will draw a solid lesson from the fate of the first. They will bring a lot of new: security policy, control algorithms, structure supervisors, system delays with safety mechanisms, and formally verified guarantees of the code. All this will help to sustain cryptoamnesia storm.

Some final words: the biggest lesson for the crypto-community is the idea of decentralization: different groups duplicate the work, so minimizing the probability that the error of one system will pass unnoticed through other systems. The crypto ecosystem is still forming and will become the current experiment on the front lines of software development, computer science, game theory, and philosophy. Regardless of whether it will become the mainstream of social applications in their current form or after a few iterations, the results will be available for anyone able to learn and see.