The prompts shown literally invite the LLM to complete the copyrighted text by providing unedited selections and asking the machine to finish those. Even if this is problematic in a small number of cases it is not a use case that undermines the business model of the newspaper since it requires the reader to have access to the original text. Nor will it be easy to demonstrate economic harm since this is not how readers consume news and is very far from how users interact with LLMs. Nor are the archival materials used for training remotely reflective of the "time-sensitive" articles that newspapers sell. And archival materials are easily available elsewhere so where is the case for economic harm?
The courts are going to rule that LLM training is a transformative use case that is protected as fair use under copyright law. They may rule that if an LLM-powered service is explicitly designed to enable copyright violation that is illegal, but there is no way any court is going to look at these examples and see it as anything other than the NYT fishing to try and generate a violation by using the LLM in a way that is very different than the service is intended to be used and which -- even if abused -- doesn't hurt the business model under which the text has been produced.
The most likely outcome is that LLM providers will add some sort of filter on output to prevent machines from regurgitating source documents. But this isn't a court case the NYT can win without gutting fair use protections, and that would be a terrible thing.
Perhaps you are being a bit cynical? There are certainly no problems that can't be solved by giving more money to government initiatives like the MaRS Centre in Ontario. These institutions also create jobs and are businesses of a certain sort, so funding them is supporting business directly.
And who can argue with the results? Several promising entrepreneurs each year take advantage of these critical programs to consult with pro-business advisers who assist in job creation by recommending they apply for 3k tax breaks.
I think perhaps you should read the paper if you want to have a deeper discussion? An understanding of how the mechanism works should make it clear how the solution enables outbound payments to network nodes, which -- in turn -- creates a for-profit incentive to run access points and network infrastructure.
Self-provisioning networks are indeed more strongly resistant to "centralization" than those which are deployed by outside parties. The alternative in the blockchain space is a reliance on outside parties and business models like Infura to provide access nodes and APIs. Unfortunately, any external business model capable of monetizing such infrastructure requires closure around data-and-money-flows, which creates key points where cartelization and monopolization emerges.
Looking at the links you've provided, afaict you seem mostly concerned that the term "sybil-proof" is used to describe a situation in which not using multiple identifies to collude is a dominant strategy instead of an "impossibility according to the laws of physics"? Four points here:
The first is we're dealing with an academic term that is used in a specific context ("no information propagation without self-cloning") and even more specifically in the context of an impossibility proof that has stood for a decade; showing that this impossibility proof is not actually valid is a substantive step forwards and nitpicking terminology is missing the point.
The second is that your definition isn't better. Even networks with trusted third parties cannot prevent sybilling by this definition since it creates a definitional impossibility. While a certificate authority can limit entry, it can never truly know that two distinct identities are not controlled by the same person. All a CA really does is provide a point of closure (monopolization, centralization) which can theoretically identify and tax colluding participants.
The third is that achieving a dominant strategy in which sybilling is disincentivized is a massive step forward. It does not make sense to refer to this as "sybil-resistance" in a field in which mechanisms without this property are considered to have "sybil-resistance".
Finally, and most importantly, one of the consequences of this mechanism that is that all attack vectors that can be carried out using multiple identities are more efficiently carried out with a single identity. So it is not the existence of multiple identities or the collusion between them that is the source of the problem.
Do you want your transaction fee to pay for mining, or do you want it to pay for the servers that run the network? If it pays for mining, who pays for the servers that run the network?
You seem to be concerned about censorship attacks -- attacks that necessarily involve orphaning work in some capacity -- either orphaning blocks produced by honest nodes, or orphaning (refusing to include) tx-embedded routing work being sent from honest users.
The payout lottery does make censorship costly for all attackers who orphan work, but you'd need to specify the exact attack method if you want a discussion of specific work-orphaning attack vectors. Even nodes with a majority of "routing work" do not have the ability to costlessly orphan work produced by other nodes, so it isn't clear what exact attack you have in mind or why you think controlling a bunch of first-hop routing nodes under different identifies somehow makes these costs go away.
You're not describing a sybil attack. You're talking about censorship.
If you want to address that, your best strategy is to incentivize people to run access points that do not censor. This requires payouts to routing nodes in proportion to the value of the transactions they process, which requires the ability to make routing payouts, which requires a sybil-proof routing mechanism.
> There are no excess hops in the attack I am describing.
As above, you're not describing a sybil attack. Adding a bunch of 1st hop nodes to a network doesn't mean that those nodes are conducting a sybil attack.
If you want to leverage them in an attack, you would have to have them send their TXS to a central block producer that gathers the TXS from all of these first-hop nodes and then tries to use all of their collected-work to orphan the blocks/work from honest nodes in the network. If you work through the paper, you'll see that this attack is provably-costly.
You could theoretically avoid the penalty if you gave all of your 1st hop nodes the same keypair, but in that case you don't have multiple identities on the network and you don't have a sybil attack...
> However depending on the architecture that is the only motivation needed and depriving others of rewards; in the long term this could result in centralization of routing nodes (by control) and finally censorship.
Highly recommend reading the original paper (!!) as the "depriving others of reward" concern you have is exactly what is solved through sybil-proofing. So the solution addresses your concern directly -- nodes share because it is profitable for them to do so, and it is profitable for them to do so because others cannot deprive them of their due reward without losing money. If we are both nodes then I'll give you my TX flow because the only way it has value to you is if you figure out a way to get it confirmed faster than I can.
> how does the entire network agree on the minimal routing when there are many routes with the same 'depth'
A good technical analogy is to think about how POW works, but have users attach a difficulty-hash to their TXS instead of a fee. Have nodes "collect" those hashes to meet aggregate difficulty requirements. And now have consensus halve the "measured value" of the tx-embedded hash for each hop in its routing path. Early-hop nodes in this system will have a similar statistical advantage producing blocks.
The POW analogy suffers from attack vectors that are not possible in the sybil-proof mechanism, but it should give you a sense of what is happening. An attacker can create 1-hop routing work using their own money, but in that case they are making blocks with only their own wallets and pay the costs of hashing/burning to release payments entirely out of their own pockets.
Author of the original paper this write-up was based on here. The sybil attack as a theoretical problem is defined in the "Red Balloons" paper ("information propagation without self-cloning") so that might be a good starting point.
The solution formally and mathematically achieves these properties:
- not profitable to add routing hops
- profitable to share with others
- not profitable to share with yourself (!!!)
If you have a single central entity somewhere that isn't a sybil attack. Nothing wrong with being concerned about network centralization, but you're much less likely to have it in sybil-proof systems as above given that nodes suddenly have commercial incentives to share data as opposed to hoarding it.
The attack vector you are describing doesn't work. If an attacker owns 1010 nodes and collects inbound transactions on 1010 nodes, then they have to relay those transactions to a second-hop block producer, which halves the value of those transactions for creating blocks.
And even then they could not orphan the work produced by other nodes profitably, which would be possible in POW and POS. So at best you have an inefficient strategy for maximizing routing work, but no profitable attack based on it and a strong incentive to behave honestly even if you have a majority of work.
It would be difficult to modify POS to have these properties, but the question of how best to accomplish that would be something for developers in those other networks to address.