Your concerns can and will be part of the discussions taking place on Tau in the future :).. well if it works, which I think the project is really onto something profound.
Think of it as a communications enhancer by way of human-machine-human communication. Instead of statistical-models in deep machine learning algorithms (which is indeed the pervasive thinking / practice in mainstream AI models), Tau is based on a logical framework as everything expressed on it are as definite as computer programs. So it doesn't deal with natural language per se, as Tau doesn't assume opinions. I think the core will take sometime to develop and would be a case of ethical and philosophical discussions. After that, it can start supporting sub-branches with different rulesets.
It could open up app development to more people than ever before since the Beta phase will be about code synthesis by way of discussion as well, so I guess it's a good thing. Check out this post: https://steemit.com/tauchain/@trafalgar/the-power-of-tau-scaling-the-creation-of-knowledge
Also, Tau will be a social choice platform. It's partly based on https://en.wikipedia.org/wiki/Social_Choice_and_Individual_Values
OK, @trafalgar's post gave me more details and clarity about Tau's possible real-world applications.
Recognizing the problem and stating it as "discussions are not scalable" seems pretty good to me.
I'm still not clear on how people's inputs will be converted into computer code.
Another thing in relation to the social governance and scaling of discussions is that to me both of these have to look at data as the starting point of a discussion (at least most of the time, when we're not on a meta level of discussion). So when there is discussion around vaccination or GMOs or defecation postures or anything else, the data of every related experiment has to have already been entered into the blockchain, along with the metadata (who designed and conducted the experiment, which devices collected the data and how, etc.). So then the data from all the experiments is immutable and it can be aggregated, so clear patterns will start to emerge - or not, in which case we need to perform more and different kinds of experiments. When there is data as the starting point, discussions can be more informed and generally geared towards (making progress on) solving actual problems. Do you see discussions (on any scale, even between two people) often leading to any solutions otherwise?
Social Choice and Individual Values
Kenneth Arrow's monograph Social Choice and Individual Values (1951, 2nd ed., 1963) and a theorem within it created modern social choice theory, a rigorous melding of social ethics and voting theory with an economic flavor. Somewhat formally, the "social choice" in the title refers to Arrow's representation of how social values from the set of individual orderings would be implemented under the constitution. Less formally, each social choice corresponds to the feasible set of laws passed by a "vote" (the set of orderings) under the constitution even if not every individual voted in favor of all the laws.
The work culminated in what Arrow called the "General Possibility Theorem," better known thereafter as Arrow's (impossibility) theorem.
This platform has helped human beings on this earth, including me, can buy books and help fellow human beings. Success for all of us