Hi Everyone,
I have been watching this thread from a far for a while now, coming back in and out, just to see how the project is going along. I remember tua being announced right after ethereum just finished its ICO. I didn't think much of it. Primarily because I didn't know if Ohad was qualified, which changed once i saw his linkedin profile. But, that again made me skeptical because he was just one person, and one person alone cannot complete a project in a "reasonable" time frame to get things moving in the pace crypto usually moves. With that said, I left behind my interests with tau.
Then, it wasn't until Tau appeared on cryptonews and some other ICO sites did I turn my interest back at the project and tried to understand what is being made here.
I can see that its a very different type of project and possibly an "ethereum like" chain of its own since its different to ethereum from a fundamental use basis.
Naturally as all of are aware, this project is a step in the direction of machine learning, and I see how this 'intelligent' autonomous software could perhaps write "smarter" definitive contracts than ethereum as a result of its decide-abilty, if I'm right, which Ohad is trying to build. Using a different type of programming language, namely RDF Languages, because these languages are not turning complete?. At this point I get really lost since I'm not a developer and I tried to read a bit here and there, without any much understanding.
However on the large scale I can see the bigger picture here, and what decideability means for "smarter" contracts when Ohad used the analogy of the Student and Law Professor, in one of his presentations.
The debate was concerned around these languages and the foundations around building this system. Most of the arguments from what I can grasp seems to suggest that no tools either exist or are properly equipped to handle what tau is trying to be. Which ofcourse then is a "highly" risky investment. It also seems like this project is very similar to DeepMind from Google which is already using multiple PC's to learn a solution to problems and these answers are shared in a cloud which other machines could use as truths and solutions to similar problems they might encounter. Kinda like the Skynet we are talking about here.
This then makes me think that any day google could use blockchains and this idea which Ohad has, and to create its own network for its DeepMind. Unless Ohad can beat them first.
Perhaps, Ohad can look into the approaches used by DeepMind and again propose whether or not this project is different, and worth our time and money. Including Ohad's own time for himself.
I would love it if he is indeed the only person who can produce something of this nature, but I don't know that. And so I would like to know. Fundamentally, Google is a company operating for profits.
Blockchains are rather different, and so it is quite a worthy venture indeed. I doubt that DeepMind will be a freely available public AI tool, but it sure is more advanced and headed in the right direction. Using its established work and translating it into a blockchain network shouldn't be that complicated if they want to 'Plug it' into the chain, and use it to decide on smart contract outcomes on its own ethereum/tau like intelligent/smart contract chain.
Ohad, with respect, could you spend a little time to clarify the differences. And similarities. I understand it is a digression from normal activity, but I'm just looking at the 'larger' picture and outlook for the future.
Thanks
thanks, there's a fundamental difference between what tau is about and what a machine learning platform (eg deepmind) is about.
[machine] learning, is the process of being able to solve problems after being trained with pairs of question-answer instances. after training, the learner can hopefully answer questions. this is "learning by example". and this is what neural networks are about.
a whole different approach would be logic-based system, where answers are given from information and rules, either explicit or implicit. moreover, programming in general can be done using rules, while it'll be a very bad choice to do everyday programming using machine learning. but, leaners are still just programs, so a rule-based system can implement learners as well. tau is therefore much more general than deepmind et al.
Show Posts
