You sound like someone who knows your stuff on this and I regret if I was in any way making it sound personal or disrespectful to you personally. I maintain it’s an unfortunate if not offensive phrasing, but I’m in no position to carry rocks around glass houses: I say unfortunately or offensively-phrased things too.
There isn’t really a robust consensus that I’m aware of as to what constitutes a blockchain per se: Wikipedia lists git as one, and I suppose that’s as good a source as any absent such consensus.
git is an (often if not typically in practice degenerate) Merkle Tree, the contents of one atomic (and sometimes de facto immutable) node contain a hash (O(1)-verifiably k-equivalent… you know the drill) of ancestors.
In more pragmatic/colloquial usage I might define a blockchain loosely as a “tamper-resistant, directed, and typically acyclic / bounded-cyclic data structure with an implied machine economics optimization around infrequent but critically important fully-verifiable history subject to heuristically-determined / freely parameterized bounds on branching factor, duration in branched states, and a bounded susceptibility to adversarial interference in a verifiable consensus on the periodic elimination of branching on some semi-predicable cadence”, which is pretty hand-wavy but I think captures the spirit of the general usage. By that definition git is only a blockchain by common convention, there’s nothing preventing or even discouraging arbitrary, unbounded branching other than it doesn’t have a ton of widely valued use cases: most any time you’re fine with a branch that never has any scope to interact with any other via rebase or merge you could just make a copy or maybe a copy and a copy of some metadata/history, though git in practical terms is a good tool for such a copy.
And I think you’re right that as with any over-hyped technology, it gets attached to projects that don’t need it when it’s “hot”, preoccupies both investors and entrepreneurs without better ideas for how to deploy their time and money when it’s “in”, and is therefore constantly oscillating between being a magnet for snake-oil types and being out in the cold.
Throw in a bunch of electricity consumption that’s maybe net driving up carbon emissions and maybe net attaching a financial incentive to electricity so cheap that it basically has to be renewable but it’s kinda too soon to tell, and I think I’m now having trouble seeing how crypto three years ago and “AI” last year are any different along these dimensions.
The difference in my view is that AI is probably higher variance by a lot on social welfare, and not because of some dumbass “paperclip-indifferent AGI” tripe.
Blockchain as applied to finance has the scope to create transparency into financial markets and compel governments to open the books on what is and isn’t legal regarding money, for who, and why. It will never like, totally disintermediate government from money, because money is the #1 national security priority of any functioning government, so inventing money that the government can’t control is more likely to buy you a R9x than a Turing Award (in a macabre way it’s darkly amusing to contemplate the fact that it could buy you both). It also has a positive (in my view) externality of creating broad-spectrum incentives for the public to understand a little better how important digital identity, security, privacy, and autonomy are in 2024 and build at least a little muscle memory around running a slightly or maybe even substantially tighter ship on personal digital footprint. I’ve apologized to two friends this week because I lied to them about something that is now news that recently broke on the Onavo/Meta thing TechCrunch ran and I wanted them to hear it from me. I lied about this because before it hit the press, I felt it would have been detrimental to the national security of the United States to talk about it, but what I really wish is that we wouldn’t end up in situations where anyone faces such dilemmas in private industry.
AI has more obviously useful applications at the consumer level (though it’s largely a solution to itself as a way to get information one could previously get from a search engine before it ruined the indexes of search engines by making arbitrarily persuasive falsehoods too cheap to meter, we’ve had spam for a long time, but spam so good it’s convincing to experts in anything other than a bad mood? That’s new.). The danger with AI is that it winds up being something other than “available weight” and “operator-aligned”, i.e. whoever is the last man standing has arbitrary unaccountable power to convince anyone of anything and prevent that from being accessed by anyone else.
So probably higher stakes.