Can AI work on top of Cardano's CL?


#1

Question:

Can AI be used on Cardano’s CL such that it would dis-intermediate Google/Amazon/Apple? If yes, it could be a huge win for individuals and for Cardano.


#2

Sorry, I don’t really understand that question.

By AI do you mean “Artificial Intelligence”? If so, I don’t see how AI has anything to do Cardano SL.

I also can’t see Cardano SL competing with any of Google. Amazon or Apple’s existing businesses.


#3

I am talking about CL— Cardano’s computation later, which, I assume, is going to be turing complete.

If so you should be able to use it as any other server to run your algos, train your machines and otherwise perfect your artificial intelligence.

If that’s architecturally possible, the CL should be much more powerful than any other single computer, which is why it could compete for computational power and with giggle and other incumbents.

I hope that clarifies it.


#4

CL is not computation layer but contract layer. Yes it will be Turing complete, but I’d expect that using it for general computation purposes would be very inefficient, the power of multiple processors being outweighed by, in general terms, the fact this is not what it’s designed for – or maybe just think of that as “blockchain overhead”.

Edit: Maybe the main issue would be that parallel processing requires different processors to work on different data, whereas in the blockchain context nodes mainly replicate each other’s work.


#5

Thank you. Cardano does refer to CL as computation later on its roadmap—although as you pointed out, they use it for smart contracts only.

Could a VM be created to enable parallel processing without changing system configurations on nodes?

I remember a while ago you could dedicate your computer’s idle time to process complex computations on a set of radio data. The project was to help decipher signals in search for extraterrestrial life—SETI.

I wonder if something to that effect could be done…


#6

Utilising global unused computing power is a fairly old idea, but AI has nothing to do with a secure public ledger, so I’m too not sure if it should have anything to do with a blockchain )


#7

Thanks for sending the links. It looks like DAI is a good fit for distributed computing which CL could be designed to become.

The reason why I asked my question was to see if Cardano could be leveraged to become a platform for wider array of services/products. It relates to blockchain in a way that such a system would require “fuel” to run, which will be ADA.


#8

That’s an interesting thing to think about, of course, but for a far future, imo. And as far as I can imagine it right now - it should be thought about separately from CL. Charles, among other things, pushes very hard for people to understand that blockchain itself is a system for extremely narrow usecase - secure public ledger, that’s it. It does open up very many doors that were closed previously but it’s not a panacea that can solve EVERY possible problem in the world. It’s literally - just a secure ledger. And this is exactly why CH pushes so hard for people to start thinking about off-chain solutions, like custom MPC-protocols, that also require almost no trust, and can be eventually settled on-chain.

So if we think that we would need something like a global AI - we should think of that like an oracle-system. It might be distributed, it might be executed by the same nodes that maintain the Cardano chain, or by others, and it would have something like an API for smart-contracts (or other apps) to call it with tasks and then to collect results eventually, and tasks may be fueled with ADA (nodes are payed for executing CPU-bound AI computations).

Something like this is theoretically possible, and, imo, would be interesting to think about or try. But it would not be, and should not be, directly linked with CL as is.


#9

turing complete

Turing Completeness is a concept from Computer Science which is extremely valuable in a Computer Science theory context but has been significantly oversold in the cryptocurrency world.

In order for a programming language to be Turing Complete it must be possible to write a program in that language that is not provably terminating.

However, there are languages (eg Coq) which are not Turing Complete because every program written in that language must be provably (by the compiler) terminating.

As far as I am concerned, being able to prove that a contract on a blockchain terminates (and hence is not Turing Complete) might be useful. If a contract is provable terminating it may actually be possible to predetermine how much gas the computation is going to consume.


#10

Has anyone considered AI voting for the BIP process? Conceptually, I’m thinking of an program that weighs the pros and cons of proposed changes and votes according to some specified parameters that assume the long term survival of the chain as the baseline. You could weigh this vote as some small or larger % to push back on political grouping that is unwanted for the growth and survival of the chain. An example for other changes would have been voting for a block size increase for Bitcoin and weighing the security, cost, and necessity for the change.

I realize that AI is hyped and that they are in many ways just running recursive algos, using massive amounts of computing power, to brute force a solution in a “learning way”. That said, I feel like the field is making progress and that the ability to interface with various learning networks will offer some new possibilities when the constitution is finally set forth post 2020. This doesn’t have to do with settlement to the chain but rather the BIP process and governance.


#11

I understand the trade-offs. Coq probably will have its own use-cases where knowing/setting the max limit is of paramount importance.

But the Turing complete machines, for better or worse, give the maximum flexibility for developers. There is value in that too.