Transcript from the Interview ( Upto Q12 ) whereafter AMA conducted by Cryptosky and his Community on 27/12/2019.
Cryptosky interview
Q1: In short, what is Tau?
Ohad: Tau is a p2p network that is completely defined by its users, effectively and in large scale. For that we have to come up with a decision-making process such that many users will be able to collaboratively decide how Tau will be like on the next block (or "auto update").
Q2: What is the main problem that Tau solves?
Ohad: The problem of large-scale decision-making is commonly treated by voting, like elections for parliament. But there is one big problem with this way: everyone has an equal right to vote, but not everyone has an equal right to propose what to vote over. This is mainly due to practical reasons: how are people going to read through thousands and millions of proposal every day?
This problem affects not only the situation of a p2p network governed by its users, but any decision making process
that involves a large number of people. The traditional solution of voting, offered also in the p2p space e.g. by Tezos,
does not allow the effective propagation of the vast information gained in the process of proposing, discussing, and
voting, and all considerations are reduced to very little information, comparably. This, again, mainly for practical
reasons.
Q3: Which alternative does Tau offer to voting?
Ohad: We solve this problem by trying to reproduce, in large scales, the decision process being taken naturally on small scales.
On small groups, we don't need to vote. We only say what we think, and the opinion map simply arises from what we said, and is clear to everyone without any voting process.
This can be done in large scale as well, as long as computers can "understand" the opinions, and by that create an opinion map and calculate the consensus, even in discussions involving many participants.
We therefore propose a decision mechanism in which users write their opinions in some language that machines can understand, and the system then calculates their agreements and disagreements, and transforms itself according to the users' consensus.
Q4: So, What do you mean by ‘languages that machines can understand’?
Ohad: For the languages to be used over the system, let me quote from the article "The New Tau" (http://www.idni.org/blog/the-new-tau )
"We postulate that there should not and can not be a single universal language. There is no reason for one language to be optimal (or even adequate) for all needs.
We therefore come up with a meta-language that is able to define new languages, but hey this would be back to square one with one universal [meta-]language. We therefore require the meta-language to be able to redefine itself and change, just as it can define other languages. By that we get not only many languages but also a self-amending language, which is an important part in a self-amending system.
It turns out that logics that can define themselves and have nice logical properties like decidability are not very common. We have Universal Turing Machines, but a less expressive and more informative (e.g. decidable) language is not easy to find.
We adopt the logic PFP which its expressiveness is PSPACE-complete as known from Finite Model Theory books, and shown to be able to define itself in Imhof, 1999 "Logics that define their own semantics".
From here we continue to the Internet of Languages. Using the meta-language which we call TML (Tau Meta-Language, can get impression from the ongoing work on github) users define new languages by specifying logical formulas to describe what it means for two documents in different languages to have the same meaning. In other words, to define a new language, one needs to define how it translates a semantics-preserving translation into an existing language.
Semantics in our scope is ontological (objects and relations), and not operational semantics as in programming languages. By that we get an internet of knowledge representation languages that make the choice of language to not matter. A document in one language can be routed (using TML programs) into different languages.
We do not refer to translation as in French to Chinese, as we already stressed that we don't deal with natural languages. Of course, theoretically, it might be the case that one day someone will program over TML something that can understand natural language completely, but we don't count on such an event.
Indeed there are many formalisms of natural language that are quite close to the full language and comfortable for humans to work with (what we refer to "simple enough English that machines can understand"), so we can expect TML to process human-comprehensible languages to some extent. But TML is intended also for machine-only languages. For example, one might want to convert a document into a formatted HTML or into a Wiki, or to convert a program in some high-level language to machine code, or to synthesize code from logic."
The internet of languages will therefore allow many languages to coexist and evolve with time over the system. Users will use those languages in order to make discussions over the network, and enjoy the machine's ability to enhance the discussion, given it understands what has been said.
Q5: What is special about Tau's logical framework?
Ohad: For the ability to soundly and collaboratively reach decisions that affect the whole system, we came up with "the Three Laws of Laws", which are three natural requirements from any language of law (of a country, or of an organization, or the code of a p2p network):
1. We require the question "is X legal" to always be answered in finite time, for any X. In other words, we require this question to be decidable.
2. The language has to allow self reference, as the law typically refers to "the law", especially when it comes to laws of changing the laws. (for more details, see the article "The Art of Self-Reference" http://www.idni.org/blog/art-of-self-ref )
3. Third and minor requirement is the ability to delete laws, for example in case we want to make a new law which contradicts and old law.
The first two requirements can very rarely come together. It turns out that TML satisfies not only the requirements from a meta-language, but also those three laws of laws.
Tau is therefore meant to be specialized to a very specific logical framework, that which satisfies those requirements. Such a framework is known in the fields of Descriptive Complexity and Finite Model Theory as FO[PFP].
Q6: How will collaborative decision making take form in a decentralized network?
Ohad: The setting of p2p network limits the collaborative decision-making abilities, mainly due to the inability to uniquely identify a user. By that, without additional care taken, an attacker may influence the network to behave in undesired ways, as the Chinese proverb says: "If the wrong man uses the right tools, then the right tools work in the wrong way".
This is one reason for the plan of having a semi-centralized discussion platform first, where users will discuss how they want the fully decentralized network to be. For more information on the limits of decision making in an electronic decentralized environment, see the article "Consensus and Options" http://www.idni.org/blog/consensus-and-options
Q7: What are the main features of Agoras?
Ohad: Agoras will be a currency which will not only be powered by the Tau technology and by that governed by its users, but will also feature:
1. Knowledge market: pure knowledge-cash transactions, which are described in the article "From Agoras to TML" http://www.idni.org/blog/agoras-to-tml
2. Advanced derivatives market featuring risk-free interest without printing new coins, which are described in the article "Consensus and Options" http://www.idni.org/blog/consensus-and-options
3. Computational resources market, where the idea is described on Zennet's materials (http://zennet.sc ), but of course things will be adapted to fit the Tau framework.
Q8: When will Tauchain be launched as the main net? There were plans before to finish the development of TML and Alpha Tau in 2019. How is the current development going?
Ohad: We almost finished TML and the Alpha: TML is ready enough to support the initial alpha, and already fulfills all axioms mentioned in the introduction, but there's more to be done in order to make TML fulfill its end goal, being a tool for writing translators between formal languages, as part of the design of the Internet of Languages.
The Alpha is currently under development and we're close to release something very initial. Since the last months our development team grew very much, from one person to six people.
We therefore expect the development to proceed faster. Also, almost all the hard parts were already done.
For the main net, it is very hard to give estimates. It might be that things will go so fast and successful, and we'll finish everything in one year.
Q9: What is the difference between Tauchain and Ethereum? Why do you think Turing completion is inherently unsafe?
Ohad: Ethereum is a currency supporting smart contracts, and is very different from Tau's goals as described in the previous questions.
Turing complete frameworks are inherently unsafe because it is inherently impossible to answer all questions about the behavior of programs. For example, we would like to ask whether the program will ever do something that we don't want it to do, or, to ask whether it performs the calculation that we expect it to perform.
In a Turing complete framework, there is no guarantee that such questions will be answerable for all programs. By that, as long as we don't restrict ourselves to languages which are not Turing complete, we lose the ability to know what the program will do, for any case that users might want to query about.
Q10: One feature of Tauchain is formal verification, so the language design is non-Turing complete. What is the design difference between this formal verification and formal verification based on the coq class implementation? Which computer theory is TauChain's formal verification developed developed based on?
Ohad: Coq and similar provers are very general and support a very wide variety of logics. Our scope is for logics that satisfy the above laws of laws. By that we can, for example, have a fully automatic prover, unlike Coq and similar which require user intervention. The theory behind Tau's logic was answered at Q5.
Q11: There are some people thinking that TauChain has similar characteristics to Tezos. Both are self-amendment blockchains, and both smart contract languages have formal verification function. Can you explain the differences between TauChain and Tezos?
Ohad: As described in the laws of laws, we require rules over the network to be written in a decidable language. Tezos doesn't support decidability but is Turing complete. Further, Tezos social-choice mechanism is based on voting, and voting cannot fairly scale, as we showed on the previous answers.
Q12:Tauchain itself does not require tokens. Agrs is just the token of the application Agoras. Will the design on this economic model be detrimental to Tauchain? Why does the Tau team hold only 3% of the total token supply, this seems very little compared to other projects
Indeed Agoras depends on Tau, but Tau doesn't technically depend on Agoras. However, knowledge over Tauchain can evolve much better in the presence of a knowledge market. Also incentives to keep the network running (like miners) are important and will require some form of value.
Ohad: The reason we keep only 3%, which we consider to be even too much, is because we understand how big this whole thing will become. If we truly believe it, we have no reason to keep more tokens.
Cryptosky community Q's
Q: When will Tauchain release a complete version of whitepaper. Many programs comes first with white papers, but Tauchain has not released a systematic whitepaper. It might be difficult for others and people in blockchain industry to assess and evaluate Tauchain.
Ohad: The whitepaper is being written on these very days, by a friend of mine which is a professor of physics and has a lot of experience in writing scientific papers. I guess that it'll be ready in about a month or two
Q: What are cyc, souffle, and xsb? How does TML compare with them in terms of functionality and speed?
Ohad: Souffle - is Datalog implemenrtation. Datalog is very similar to TML, but still different.
XSB - is a Prolog implementation, which is less similar to TML than Datalog.
Cyc is a +30 years old project that formalizes what they call "common sense" into logic. it is aimed to teach computers some facts about the world. this in contrast to Tau which is not about absolute truths but about opinions and discussions
Q：Does tauchain have to be implemented in a decentralized way, a friend just asked?
Ohad: No, but Tau does not limit the users from using centralized services over the network
Q: is tml suitable to write other blockchain program?
Ohad: yes, in fact tml can implement virtually any program that comes to practice
Q: Would rewriting btc with tml be possible?
Ohad: yes, in fact just this morning we discussed the ability to compile any C/CPP code into TML,
it will not be Turing Complete because it will assume finite memory
Q: what’s the relationship between tml and blockchain? Is there any consensus structure in this blockchain project?
Ohad: there is no consensus structure that is perfect or close to perfect, and each one of them comes with a tradeoff. this makes it a matter to be decided by the community and not by the tau developers
this is one of the reasons to start the discussion platform in a centralized way, and there users can give opinions and decide the consensus structure of the decentralized tau
Q: people say "the pi calculas is the best theory for blockchain sharding. can Tau implement pi calculas like sharding technology in the future, or does tauChain not need to use pi calculas sharding?
Ohad: for pi calculus, we are interested in a very specific logic, because this is the only logic that satisfy the 3 laws of laws, no one ever discovered those requirements from the language of law, and of course no one ever found a logic that fits them, and it turns out that there is essentially only one logic.
Q: so if the community wants, tauchain can implement pi calculas like consensus?
Ohad: pi calculus can be implemented in tml under the bounded memory assumption yes
Q: As we know the current blockchain project normally has the consensus structure at least. Maybe I can assume this is not a blockchain project, or you restructured the “blockchain ”
Ohad: indeed blockchain is a very small part of what tau is about
Q: Is there no need to develop blockchain with tml?
Ohad: there is, because we want the system to define itself. it can be in any language defined over the internet of languages, not necessarily tml
Q: Does tauchain or tml have any relation with AI
Ohad: yes, tau/tml has a lot of relation to what called "logic-based AI"
Q: how does the community decide the consensus of tauchain?by voting?
Ohad: by simply expressing their opinion, without any order
Q: Could you give me sample of TML?
Ohad: TML is just Datalog with negation. for now it's not enhanced with additional features that make it easy to write, but this is what i work on right now, in the support of second order logic
Q: how do we prevent bots sending many opinions to manipulate the consensus?
Ohad: to prevent a bot is a very real problem, and i wrote about it in length in the article "consensus and options"
http://www.idni.org/blog/consensus-and-options
Q: also want to ask you how do you program an AI over non turing complete languages? bitcoin isn't turing complete and therefore has little programmability
Ohad: in fact all machine learning algorithms, including neural network, can be implemented already in datalog which is a strict subset of tml
this is because in the definition of a "learner", at the scope of "PAC Learning", we require the learner to run in polynomial time. hence it must be expressible in datalog
Q: What will tau be/look like? Something like Knowledge Graph or in a different form?
Ohad: yes basically tau will store knowledge graphs
Q: can people in a group write a dapp on tauchain with languages defined by tml?
Ohad: yes definitely, tau is all about collaborative programming
Q: How will we eventually present this towards the normal users?
Ohad: we see a lot of importance in making it usable for normal users, and we'll hire top ui/ux developers and researchers for this task.
Q: So anything dapp based on tauchain can upgrade with large scale discussion?
Ohad: yes, people will discuss what a program should do, and its code will be auto-generated. this is the "beta" part on our roadmap.
After all, people need to discuss what tau itself should do, and then auto-translate it to code.
Q: so tauchain is a world computer?
Ohad: tau is a world logical solver yes
Q: what do we mean by discussion? Not in a natural language right? But using code?
Ohad: right but the languages will evolve and improve and become easier over time, because the support of the internet of languages, to be more accurate, not code, but logical formulas, like in first order logic,
just like datalog, only allowing negation.
Q：can we have a look at what kind of language these logical formulas currently look like? Is there a demo etc?
Ohad: yes there's the tml playground，but i'll need to check with Tomas where the latest version of it is, I'm not sure it's working , try tml.klapka.cz
Q: Does that mean to create a dapp over tau, say like wechat, you just have to use a close to natural language , and it can even update itself over time?
Ohad: yes, except that the initial languages will probably not be so close to natural languages, but they will evolve with time.