Tokenized zero knowledge machine learning and its applications

Link: Tokenized zero knowledge machine learning and its applications

Notes on AI and system complexitiy

Our IT systems, including software and hardware components are getting so complex that we need AI support for monitoring, maintenance, operation or even for development. 

However that AI support will not make our systems simpler and it does not help people to understand that systems better either. So on a long rung it will result in IT systems that can not be operated based on purely biological intelligence anymore, only with the help of AI, or purely with AI.

Replay attack on the Blockchain

Replay attack can be interpreted in two ways on the blockchain:

1. If there is a hard-fork on the blockchain, and the system is spited into two concurrent platforms, there is a possibility to copy one signed transaction from one chain and put this transaction to the other chain. Usually at a hard-fork there is a mechanism that explicitly avoids a replay attack, like there is a modified transaction semantics, or even just one bit on the forked blockchain, so signed transactions of the old chain will not be valid on the forked one.

2. Even without forking there is the possibility to copy an old transaction and try to replay it again on the blockchain. It is not too effective at an UTXO system, because the system will know that the old transaction output has been already spent. However if it is an account/balance based system, further algorithms must be used. One way to avoid replay attack on an account/balance based system is to implement a counter at each variable that has to be increased at each new transaction. Another way can be to create a nonce for each transaction randomly and automatically, the system has to ensure that the same nonce can not be applied two times. There might be some mixed solution as well, where quasi random nonces are used in an incremental fashion, like:

nonce_next = hash(nonce_prev)

In a multi-hash blockchain system we have most likely account/balance based systems, implying that we have to use one of the nonce or counter based solution. That means that the state of the blockchain is actually not the state of all of the balances, but the tuples of balance and nonce

state_i = <balance_i, nonce_i_j >

Blockchain and evolutionary algorithms

From a conceptual perspective a blockchain can be regarded as a kind of a evolutionary algorithm, which is driven indirectly by market forces. The genesis block might contain something as a set of initial variables, which we can regard as a the initial population. The population is reproduced in each round with the help of different market factors and human interactions. The variables might have new values, they might recombine different values to new ones, and even brand new variables can appear as well, mimicking a little bit something as recombination or mutation. Certainly, it is questionable if clear algorithms can be imagined for representing something with a genotype or a fenotype. One such an example is the cryptokitties, where actually both mutation and recombination are defined in a clearly specified way, however they might be not the only application that can be regarded as blockchain based market evolution. 

Combining conveyor belt with artificial intelligence

Combining a conveyor belt with artificial intelligence is a dream that sounds for the first run pretty much as science fiction. However there a realistic chance of realizing such an algorithms. Let we imagine a configuration space with CU = {cu1,cu2,cu3 ... cuN} different configuration units where each cui configuration unit has a cui = {c1,c2 ... cki} set of different configuration possibilities. General software architectures can be regarded for instance to such systems, they provide many functionalities and which can be configured and put together in many different ways. In this example, we consider the configuration space as independent elements, however in general case they might be dependent from each other. 

A configurable conveyor belt is a subset of subsets of the whole configuration space CB =  {cb1,cb2,cb3 ... cuK}, where each cbi is a subset of a given cui. The cbi elements can be regarded as different steps of the conveyor belt and the elements in each cbi contains the different choices for each step. We can say that the conveyor belt is not configurable if each cbi step contains only one choice. We can say that the conveyor belt is human configurable if it contains 7 plus minus 2 steps and each step contain 7 plus minus 2  configuration choice. A conveyor belt might be hierarchical if each cbi configuration step can be regarded as a whole configuration space with configuration steps and sub-choices.

The major task of a conveyor belt learning algorithm is to analyse different choices of the configuration space and create or estimate a human configurable conveyor belt automatically with the help of statistical analysis or machine learning in order to capture the choice pattern that can be used in most of the situations.  

Domain specific languages can be also regarded in a way as special conveyor belts that are capable of realizing only parts of a theoretical configuration space. Unfortunately there is know known method to integrate with the artificial intelligence or machine learning. Perhaps such a theory could provide the basis for the integration. 

Another open question if conveyor belts can be built up a modular or even modular and hierarchical fashion and if they can be supported by efficient machine learning algorithms. 

Notes on decentralized artificial intelligence

Current trends in artificial intelligence, like Singularity.Net, Neureal or Pandora having a little bit the wrong direction. Overall is the idea, that different artificial intelligence components can be somehow marketed and sold with the help of a marketplace which is realized with the help of a blockchain solution and some tokens. However this idea is not really optimal. Distributed ledger should actively help the integration of the different artificial intelligence component, effectively realizing a hierarchical pattern recognition structure. Tokens should not only represent an economic incentive but they should actively carry information among the different pattern recognition components to effectively function as a data bus of the architecture. Certainly classical architectures like blockchain can not really scalled up to carry such a huge amount of information, however with technologies like Hashgraph functioning solution might be carried out.

Certainly, it is a major question if distributed ledger technology is really required, or possibly there is the chance to put everything to a centralized data bus. 

Directed decentralized evolution

Decentralized systems supported by, like blockchain or distributed ledger provide the way of creating systems in a self improved way controlled by some kind of a directed evolutionary algorithm. If we consider bitcoin and the mining evolution, the bitcoin protocol directly makes incentives for a directed decentralized evolution resulting the appearance of the more and more efficient and advanced mining devices. Another interesting example might be the cryptokitties application, in which the appearance of the different digital lifeforms are controlled by a market mechanisms. 

Extended digital existence

As more and more digital service, especially services based on artificial intelligence appear in the in the support possibilities of an individual, it might make sense to redefine the human existence or the person itself. On a long run a person perhaps not only concentrated on the pure biological existence, but all of the biological devices and online algorithms might also be considered as part of the individual itself. In this sense a human being might be much more strongly related to virtual services, so tight that these services might become actually part of the existence itself. In this sense we might as well speak about an extended digital existence.   

Genetic and evolution algorithms and market

The blockchain application called cryptokitties provides some interesting ideas about genetic and evolution algorithms and the market. In classical evolution and genetic algorithms there is usually a target or fitness function to evaluate a certain population or individual and the goal is to minimize the difference between target value and the actual performance of the population. However the whole concept might be put into a market context. The mutated or combined entities of the population are exchanged or traded between different actors. As the time goes on, high value individuals are traded frequently or for a high price as less important individuals will have low liquidity and low price. Certainly, in this way the target value that is evaluated in each round depends on the subjective evaluation of the individuals taking part in the trade. As these subjective preferences might evolve over the time, it is pretty questionable in which direction does the algorithm converge.

Notes on decentralized artificial intelligence algorithms and platforms

As there are many initiatives to create decentralized artificial intelligence algorithms, most of them work in a way that they weakly integrate the blockchain technology with different blockchain platforms, usually via interfaces and a common tokens that represent some kind of an exchange of the services. Similar projects that successfully made an ICO are for example Neureal, SingularityNET, or Neuromation. However it would be much more interesting to integrate a distributed ledger technology with a machine learning algorithm really on the algorithmic level. It would provide this way a whole integrated algorithm for decentralized artificial intelligence. Certainly it is an open question which AI algorithm can be efficiently integrated with which decentralized technology. 

Notes on modular AI infrastructure and Apps

AI algorithms and platform solutions are coming for sure. The only question is how should be designed an AI application from the architectural point of view. In classical computer systems, usually there is something similar as the operation system that provides a low level software architecture with certain basic services and on top apps or applications that cover specific needs. The question is if it is possible to separate this way an AI application, considering some low level primitives as infrastructure and building specific Apps or applications on top.

Questions on DAI (Decentralized Artificial Intelligence)

Considering the hype around blockchain and the different decentralized technologies and decentralized business models, the question raises slowly if there is a way to create artificial intelligence on a decentralized way. Certainly, it is a question how exactly an artificial intelligence algorithm can be made to decentralized. 

- Should it be somehow similar to the autunomous agents ? 

- Is it possible to capitalize the blockchain or the distributes storage as well ?

- What should be the communication interface around the nodes ?

- Which functionality should be realized by the nodes themself ?

- Is it possible to create a real decentralized algorithm ?

- It it possible to create a model somehow the same way as with decentralized storage or computation ?

- Like with SWARM, or GOLEM ?

Philosophical and practialy considerations of working with artificial intelligence

An ongoing philosophical discussion should be renewed as tool supported by artificial intelligence slowly appear on the market, namely by whom was a certain product / art / service created. The original discussion if for instance a certain painting was painted by the painter or by the brush seems to be a little bit too hypothetical for the first run, however considering paintings that are painted by artificial intelligence algorithm, like by DeepDream, the question seems to be less theoretical. Supposing that I am a painter creating paintings with the help of DeepDream or sculptures with the help of DeepMind, who is the creator of the art ? Me ? The AI algorithm or somehow both of us ? 

The question can be much less philosophical if we consider for instance products that were designed and created with the help of AI algorithms. Who can we call as creator, who should have actually the rights for that product ? Similarly if an online service is provided almost 100% by an AI algorithms, then it is an interesting question who should be responsible for the service quality ? The AI algorithm ? The one who hosts the algorithm ? The one who trained the algorithm ? I think these questions will provide a lot of legal and society discussions on a long run. 

Notes on Turing Test

Turing test :"The Turing test, developed by Alan Turing in 1950, is a test of a machine's ability to exhibit intelligent behaviour equivalent to, or indistinguishable from, that of a human" - Wikipedia.

But how about on sub domains where the Machine intelligence actually beats the human ones, like playing chess ? Does the definition make sense ? Supposing that the machine intelligence is better, is there actually a chance to evaluate from humans? As the definition was perfectly working 50 years ago, it is getting to seem to be a little paradox.

Supply demand analysis on the robo-advisor market

Analysis of the robo-advisor market from the supply-demand perspective can be seen on Figure 1. Robo advisor market is characterized by the demand characteristic of a standard IT good, meaning that as soon as it is profitable to roll out a a robo-advisor there is a possibility to produce as many additional replicas or copies as needed practically for a zero additional price. 

Let we consider S1 as a standard supply curve without robo-advisors, S as the supply curve with robo advisors. Market equilibrium will be pushed off from the original {P1, Q2} point to the  {P2, Q1} new equilibrium manifesting in a P1 - P2 price reduction and Q3 - Q2 general quantity increase on the market. It is important to note however that Q3 - Q1 quantity is not produces by humans anymore, meaning that comparing to the Q3 - Q2 jobs have been automated comparing with the original market equilibrium.

Certainly the model is ideal, it considers only the characteristic of a market segment on which the robo-advisors can produce a high quality service. Certainly segments might remain where the human competence and experience is still needed and the segment can not be served by automated robo-advisor services. 

Figure 1. Supply demand analysis of the robo-advisor market

Notes on Collaboration 3.0 and corporate efficiency 3.0

The next version of collaboration and corporate efficiency will be much less identified by the nowadays way. Collaboration and cooperation with machine intelligence agents and algorithms will play a much bigger role. From a practical point of view, it is probably true that everything that is produced on a task basis on a measured way will be done by algorithms. As a consequence, human efficiency will be rather represented how one can collaborate with machine intelligence and produce something more creative, something more innovative than just the certain set of steps.

Probably the old saying becoming slowly true:

"Efficiency is for robots"

Notes on machine intelligence and alien intelligence

I do not like the word "Artificial Intelligence" it clearly describes our efforts to create something that is copying or mocking the human intelligence. Well and actually AI has a lot of achievements ranging from automated cars to beating the chess wold champion, so we can surely say it is intelligence, however it is not a human one. It is something different, just like as an alien intelligence that is slowly emerging based on our work. I think the word "Machine Intelligence" describes the situation much better: It used to plan to copy the human intelligence, however we just got something fundamentally different.  

Machine assisted business models

Similarly to machine assisted human intelligence, there can be a brainstorming about machine assisted business models, especially if we regard the current trends and improvements of the machine intelligence and communications technologies. In broader sense machine assisted business model that would not exist without some core IT or communication technologies. As a classical name, we usually call these companies as born on the web. Perhaps it is better to focus however on the business models, so perhaps the most exciting cases are companies that business model is based on some core communication technologies but the product or service that the company offers is independent from this technology. 

From a smaller perspective, machine assisted business models are companies where the core business model is based on machine intelligence. It is again more interesting if we speak really about the business model or the value chain and not about the product or the service of a company. Perhaps one example might be financial institutes that are having robo-advisor services as well. In these examples part of the value chain, that is customer interaction, service is based on machine intelligence. A little bit less machine intelligence example is The DAO, where general company management was considered to be replaces by smart-contracts and Blockchain analytics systems. 

There might be a general consideration based on the value-chain model from Porter which activity of a company can be supported by machine intelligence. If we consider general reporting and big-data as well as part of AI then we have already a lot of needs on the management level for such a technologies, however they are not necessarily part of the business model itself. I think we should clearly distinguish the two directions "supported by machine intelligence" versus "based on machine intelligence" and we should consider our efforts to the second one.  

Figure 1. Porter's value chain model.

As Porter's value chain model might be a good starting point for the analysis however it is based on the old-fashioned model of a company. It is an exciting question if the old model can be disrupted considering the new technologies and if we can imagine companies that work somehow absolutely differently. As an example let we imagine something as a fully autonomous cash-flow system that works absolutely without any human involvement. I would certainly consider that as a company though it would be pretty difficult to put into the value chain model from Porter.  

Notes on Blockhcain and Machine Intelligence

It is surprising that although Blokchcain and Machine intelligence are two mainstream technologies, there is not very much common elements or common applications of them. Just as a first wild brainstorming so mixed technologies might be decentralized machine intelligence, not just smart, but really intelligent money. It is of course an open question if such a mixing of these technologies make sense or is it better to consider AI and Blockchain as two separate layers in an application that are implemented independently from each other.  

Notes on machine assisted software services

PowerBI, it is awesome,

PowerApps, it is awesome,

Microsoft Teams, it is awesome,

However, actually the next releases will be much better. The next releases will probably have the possibility to create BI reports, Apps and communication best practices in a fully automated way based on artificial intelligence. 

And the idea is basicaly simple, put relative simple quasi freemium service onto the market, train your AI on the data (yes you can do it without actually seeing the data, without having problem with any data protecton regulations) and in two years you can launch the next release of your software fully automated based on machine intelligence trained on a huge amount of industry best practice data.

Welcome to the dawn of the machine assisted report generation and machines assisted software applications.