As previously discussed, I'm doing the Certified Blockchain Professional course. It is self-directed learning, so I'm going through it at my own pace. In order to consolidate my learning, and help organise my thoughts, I'm blogging about my reflections on each module.
These are mostly notes to myself - but I hope if you find something interesting (or incorrect) that you'll leave a comment.
Covers the basics of AI and ML. But, again, no real details.
Links to nonsense like https://singularitynet.io without any commentary.
Effectively boils to two things.
- Get data from the Blockchain and use it for deep learning.
- Smart contracts which use AI to self-adapt. Which, of course, makes the code even more obscure!
The examples they give are ridiculous:
A smart contract can respond to security incidents quicker and accurately. E.g., if a hacking attempt is discovered the smart contract can shut down accounts, can adjust gas consumption (ethereum), can move funds to another backup contract.
That doesn't require AI!
Natural language processing can also be an option to introduce in smart contracts, to understand real contracts and execute them.
Now they're no longer deterministic programs! Much better to choose a better language to write them in.
Smart contracts can also be formally verified with formal verification methods (e.g. SMT solvers) but Al can also play a role by improving the verification process.
Hmmmm... maybe. But that's not AI on the Blockchain. That's like saying IDE on the Blockchain when you mean "writing code in an IDE".
Blockchain governance can also be Al driven, which can make changes to consensus, improve security and maintain the ledger by reacting to the changing environment.
Errr? So the owners of the DAO no longer get a say and it becomes self-governing? We're in to properly off-the-wall territory here!
This leads on to an interesting discussion of general-purpose rational agents (Machina economicus). A machine making self-interested decisions. Of course, nothing about this requires Blockchain.
Al and Blockchain
Research answers for the following questions. Provide examples.
- Can Al in blockchain solve energy consumption issue?
- LOL! I suppose AI (or quantum woo) might be able to design more efficient ASICs for running computations. Or it could generate algorithms which are work-bound but which don't require more energy.
- Can Al result in increased security of blockchain?
- Meh. Formal verification would be useful. As would automated scanning. Both could benefit from low-levels of AI.
- Can Al help in providing data from blockchain for machine learning?
- Errr...? No. All the blockchain contains is a record of transactions. So what would a trained model look for? When transactions take place? Who they go to / from? Might be interesting to predict when spikes / lulls take place in order to take advantage of them.