What does HackerNews think of aperture?
⚡️HTTP 402 Lightning Service Authentication Token Reverse Proxy ⚡️
> High Cost and Resource Requirements
For training and local inferencing use, quantization may help. Problem becomes local via quantization vs. remote full tensor use. Solution may involve distributed inferencing. Techniques like model distillation can help create smaller, more efficient models for inferencing.
> Data Privacy
For training, some private datasets may be needed. For local inferencing use, determining what needs to be inferenced locally vs. what needs to be run remotely may be useful. Problem becomes privacy scope mapped onto a marketplace to mitigate high cost and resource requirements. Techniques like model explainability (versioning) and robustness testing can help build trust in AI systems.
Complying with data privacy regulations and ensuring that AI systems adhere to legal and ethical standards can be a challenge, especially in international contexts.
> Incentives
Instead of assuming the solution when considering the problem, we assume there is an incentive to either simply train a model or use one. Problem becomes financial rewards, data access agreements, or even altruistic motivations.
> Stale Data and Reproducibility
Both the code and datasets for training the model need to be updated. Inferencing needs RAG, so the augmented reference data needs to be updated as well. Anything updated might need some type of revision control, especially if that data (or code) results in poor output. Labeling data and knowledge transfer are other problems that needs revision control.
> Interoperability
We can assume a marketplace for a ML train/inference platform is needed. We have HuggingFace, for example. The problem here is likely based on the tendency for datasets to be private, such as in the case of Llama 2. Models contain the "essence" of the dataset, but we still need RAG to ground the responses.
The use of the Lightning Network combined with a proposed 402 response code is an interesting concept for addressing some of these challenges: https://github.com/lightninglabs/aperture
It could provide a decentralized and efficient way to facilitate payments for dataset access, training, and inferencing to incentivize data sharing and model usage.
While this guy may be shady as they come, and I'm not a big fan personally, the idea of combining Lightning and 402 responses is very interesting, and one I've been thinking about since Joseph and Tadge wrote the Lightning paper.
There is a means to do this today: https://github.com/lightninglabs/aperture
Consider that these entities may become self-managing at some point in the near future. If they need to pay for calls to other models, crypto-payments can be useful because they include identity and payment in one function.
As far as the vendors go, I would agree that some of them don't care about the ethical considerations and they remain focused on grabbing marketshare. That said, it's likely there we be a multitude of models running and trying to get them all talking to each other and paying for the compute required for inferencing is going to be a pain.
While it is true that there have been scams and fraudulent activities in the cryptocurrency space, it is unfair to label all blockchains as scams. Many blockchain projects have legitimate use cases and are backed by reputable organizations.
The Lightning Network is not a cryptocurrency itself but rather a layer-two scaling solution built on top of existing blockchain networks, most notably Bitcoin. It is designed to facilitate faster and cheaper transactions by creating off-chain payment channels between participants, which has risk limited to the initial payment and subsequent updates to the contract.
The Lightning Network can facilitate direct transactions between participants, regardless of the stability or availability of the underlying blockchain network. The Lightning Network operates by creating off-chain payment channels between users, allowing them to transact without needing to wait for on-chain confirmations.
These off-chain transactions are fast and can be conducted privately between participants. The Lightning Network's design enables micropayments and facilitates instant settlement, making it suitable for various use cases, including direct transactions between model hosting companies or any other parties involved in AI services.
I’m not sure if lsat.tech is having issues, looks like the protocol was recently renamed L402: https://docs.lightning.engineering/the-lightning-network/l40...
I said that it was likely paying for compute resources that would be a killer use case. Implementing 402s would be that manifest: https://github.com/lightninglabs/aperture
Now we’re moving deep into the AI markets, this will be a thing, combining both.
That said, con artists will still try to hustle others.
I've consistently posted here about Aperture, Lightning Network's reverse proxy service that implements 402s via Lightning. With support from major browsers, this would provide a means for all different types of services and content providers to easily monetize their sites.
If someone were considering this, here's a means to do it with 402s: https://github.com/lightninglabs/aperture
https://github.com/lightninglabs/aperture
https://developer.mozilla.org/en-US/docs/Web/HTTP/Status/402
Lightning Service Authentication Tokens[1] (LSATs) provide a standard that’s usable today.
With a Lightning Network browser integration like Alby[2] you can access paywalled content behind a proxy like Aperture[3], and it’s a very smooth process already.
I’m curious to know how it could be more widely adopted. Maybe a Flask integration? Other ideas?
1. Don't click on ads.
2. Run an ad blocker or, even better, turn off JavaScript[1] when a page displays excessive ads.
3. Build a new type of Open Source-based search engine which utilizes an alternate revenue model.
4. If you can't do #3, like I am, then pay for one which utilizes an alternate revenue model.
5. Encourage the sites you use to explore using Lightning for payments. Lightning can be connected to requests to the site's content using Aperture[2].
[1] https://chrome.google.com/webstore/detail/disable-javascript... [2] https://github.com/lightninglabs/aperture
Without search, nothing can be found. With it, we still have to crawl things, somehow. Maybe with AI? :)
As for microtransactions, we technically have a way to pay with Bitcoin using Lightning and 402 responses: https://github.com/lightninglabs/aperture
The second part of implementing this is to increase the cost of producing bad content for a particular identity. Using a proof of stake type system based on the identity and content, not the coin itself, might allow decent scaling of moderation.
> work by selecting validators in proportion to their quantity of holdings
In this particular case, the comments on the content (holdings) become the selecting validators. The more high quality validations through consensus of other identities, the cheaper it becomes to comment as a particular identity. If the identity becomes a bad actor later, subsequent consensus from other identities increases the cost of that identity posting new content related (through identity) to their older, higher quality content. In a way, the content an identity produces becomes their "ego" or "account".
But, by using crypto identities, no "account" is needed. A tag may be added to the content in a particular system to indicate which transaction and identity was responsible for paying to add that content. The system remains secure for privacy of participating individuals, other than what they leak through posting their thoughts to a public forum.
Was really thrilled to see this link organically get so far up on HN. Since we initially released the linked website and corresponding blog post [1] a year ago, we've continued to develop Aperture [2], a pure-Go reverse proxy that natively speaks the LSAT protocol. We use Aperture in front of all our Lightning/Bitcoin related services, with the LSAT credential itself serving as richly featured API tokens.
We have a lot of cool stuff planned to bridge the gap here between the server-side/protocol and the web itself, in the form of a new "Lightning Browser Kernel" that'll abstract away a lot of the underlying flow, to make logging in or beginning to stream payments for a computational API as easy as clicking a button. Excited to be able to share more in this direction towards the end of the year.
On the web application side, since then developers like Buck have also started to release wrappers around _existing_ web application frameworks to allow developers to add Lightning-based monetization/metering/authentication directly into their web apps [3]. If you want to see the web app wrapper live, then check out this LSAT playground that lets you interactively produce/verify/satisfy an LSAT right from your browser [4].
Happy to field any questions related to LSAT our our vision of the Lightning Native web!
[1]: https://lightning.engineering/posts/2020-03-30-lsat/
[2]: https://github.com/lightninglabs/aperture
The dream isn't dead, not by a long shot: https://github.com/lightninglabs/aperture