Some highlights from the article are below. Seems like a smart team tackling interesting problems, but as with most projects in this space, raising crazy amounts of money way too early.
1. Raised $45 million (!) in pre-sale financing.
2. 'Song says that her project has solved the scaling problem by separating execution from consensus.
“For each smart contract execution, we randomly select a subset of the computation nodes to form a computation committee, using a proof of stake mechanism. The computation committee executes the smart contract transaction,” Song wrote in an email exchange with TechCrunch. “The consensus committee then verifies the correctness of the computation results from the computation committee. We use different mathematical and cryptographic methods to enable efficient verification of the correctness of the computation results. Once the verification succeeds, the state transition is committed to the distributed ledger by the consensus committee.”
By having the computation committee working in parallel with the consensus committee only needing to verify the correctness of the computation creates an easier path to scalability.'
>we randomly select a subset of the computation nodes to form a computation committee, using a proof of stake mechanism.
who is that "we" who "selects" and what is the source of randomness (i.e. distributed verifiable randomness)? For example, in Satoshi PoW that PoW itself is such a verifiable source of randomness which thus does the "selection".
I have no stake in this whatsoever, I just have some questions that maybe you (or somebody else) can answer, as I'm finding myself agreeing with the parent post.
- What does blockchain have to do with HPC? What advantages does this give me versus current solutions?
- If I was concerned of the privacy of my data, why would I put it on an unproven, public, immutable ledger? We all know what can happen when even minor bugs rear their heads in these types of applications.
- Where does machine learning come into the equation?
- Why would I want to use this rather than AWS/GCE/etc.?
I guess really what I'm asking is, what problem is this solving? Just curious :)
The problem is combining multiple people's data for use in ML, and also establishing provenance of different data sets; existing solutions would reveal privacy completely in these use cases.
Sorry, I might not be specific enough, I am looking for pointers that show that the person created a new research area, as suggested in the previous comment.
If you list pub list, then I have no idea how to find proof of the fact.
Edit: Below are a bit ranty comments on research style. Please ignore it and provide pointers that show the person created a new research area.
And frankly, the changing of the research areas (security -> deep learning -> blockchain) are a typical symbol of smart people who are good at catching trendy research topics, but never achieved the true top-tier research results. To be specific, such people are absolutely smart and intelligent, but they have a pattern to always latch on so-called cutting-edge topics, and quickly gain high reputation. In the end, their fame and what they do usually does not well match each other.
It does help that when you are one of the first to latch on to a trending research area, there is relatively little in the way of competition for attention, funding and publishing priority. Also, in such scenarios domain specific standards and expectations have yet to be set and thus are not something that your research will be held to, at least not until the field begins to mature. In fact your research may form the basis for some of those conventions. This does not necessarily mean that the research was particularly noteworthy, thorough, or inherently outstanding.
Lol Dawn published the very first paper on searchable encryption. She and her students have also pioneered a number of new techniques in software security, new techniques in adversarial machine learning, new techniques in teaching NNs how to "program", etc.
I think this is exciting. I'm curious whether Oasis is still using SGX in their design (the Ekiden paper seems to be super SGX focused but I'm unclear as to whether they have moved away from this).
Dawn is super smart so it's hard to bet against her.
Edit: Oh maybe they're using some ZK stuff now: "Our platform achieves scalability and integrity with its novel architecture and protocol design, without relying on any trusted hardware or central party. "
The quote from the press release in my edit section.
Edit: Specifically the part where they say they aren't using any trusted hardware. If Dawn et al have some sick zero knowledge proof breakthrough, that would be a big deal, but it's hard to know what they're up to :). I wish them all the best, designing trust networks of any kind is super hard.
Dawn is serious. I am super intrigued as to what she might be running to achieve her stated goals (I'd be super curious if she's got a new ZK algo up her sleeve :D).
How is this different from Enigma, https://enigma.co, which originated at MIT a few years back and did an ICO last year? Enigma already has TEE (they have Intel partnership for that) and sMPC is coming up.
> With its unique combination of capabilities, our platform aims to enable true innovations and new applications that could not be built before. It is designed to help users leverage and reclaim control of their data, and enable frictionless collaboration between mutually distrusting parties for greater societal good–all without relying on any central party.
Vague claims to solve the probably-unsolvable - Check.
> Privacy-preserving smart contracts
Heavy Buzzwords - Check
> privacy-preserving machine learning within smart contracts on our platform
Vague mention of ML - Check
My skepticism is at max here. This is awfully light on any kind of even vague detail and is super dense on meaningless buzzwords. They link out to their own previous research - which isn't by itself bad but isn't exactly trust enhancing either. Their staff seems to be all colleagues from previous research work.
In the end cloud computing on the blockchain would make cloud computing overly complex with little to no visible benefit. Compute resources in reality are simple as hell, privacy is always first if that is your policy. But privacy comes from process and tech is just a tiny tool for it.
Frankly, this so far presents not a single hint of a usable product and reads like some people trying to cash out with questionable buzz.
Remember: scientific papers are usually a polar opposite to actual implementations.
Think Kubernetes as presented vs. actually running it.
I couldn’t really bring myself to read their post because of this title. I would argue that the buzz words are actually a detriment to their marketing.
If you’re from Oasis and you are reading this: please think of a better way to sell your product - unless your target market is ignorant business types who live for this stuff.
Another day, another Decentralized AI Privacy Blockchain project raising a ton of money from people who don't truly understand the intricacies of the problems trying to be solved. What's new?
I hate to be so cynical and this teams background does seem solid but I feel the same way when I see this as when the Prince of Nigeria personally reaches out to me for a loan. With all the other failed crypto projects these announcements feel like spam at this point..
The site guidelines ask you not to post shallow dismissals to Hacker News, so could you please not? This comment doesn't clear that bar.
If you're right, then the thing to do is drop the snark and explain some of the "intricacies of the problems trying to be solved" instead of just vaguely alluding to them. Then we could learn something.
My apologies for the hyperbole. My comment was less about this particular project and more about the shaky ground that the entire cryptocurrency industry is currently standing on, seemingly in part due to announcements like this that claim to have solved the scalability problem of decentralization without providing much hard evidence. If there were specific technical claims, I would discuss them, but in this case there are not.
I think it's confusing to the public, who ultimately need to be confident in cryptocurrency as a whole if they're going to use it, when they find them asking, "wasn't this problem already solved by x?" every couple of months.
Some highlights from the article are below. Seems like a smart team tackling interesting problems, but as with most projects in this space, raising crazy amounts of money way too early.
1. Raised $45 million (!) in pre-sale financing.
2. 'Song says that her project has solved the scaling problem by separating execution from consensus.
“For each smart contract execution, we randomly select a subset of the computation nodes to form a computation committee, using a proof of stake mechanism. The computation committee executes the smart contract transaction,” Song wrote in an email exchange with TechCrunch. “The consensus committee then verifies the correctness of the computation results from the computation committee. We use different mathematical and cryptographic methods to enable efficient verification of the correctness of the computation results. Once the verification succeeds, the state transition is committed to the distributed ledger by the consensus committee.”
By having the computation committee working in parallel with the consensus committee only needing to verify the correctness of the computation creates an easier path to scalability.'