Great initial roadmap. Thank you for pioneering this process and helping rally the EOS community. Chintai and EOS42 look forward to participating and helping form the framework to yield best outcomes for the ecosystems prosperity.
10 Likes
That is not a lot of time for 10 strangers to get to know each other and reach a consensus. Why only 1 hour if the group is willing to spend more time to reach a consensus?
Also, I understand the goal of scheduling all meetings to happen concurrently to be making it more difficult for someone to act as multiple sockpuppets across many different groups at the same time (Sybil protection). But I think that could be sufficiently addressed for a shorter half hour session to just have everyone quickly introduce themselves and allow them to tie names to faces. After that point, the group already knows what each other member in that group looks and sounds like so if they continue the deliberations in a more convenient time slot in the day it doesn’t compromise Sybil protection by much (although there is a slightly greater potential for deep fake shenanigans in that case).
2 Likes
I don’t think that deep fake can be used in live zoom.
1 Like
The technology already exists to do pretty compelling versions for both faces and voices if not in real time. We have also already seen consumer available apps doing real time face replacement with mediocre quality. I don’t have any links to share off hand, but I bet there are already tools researchers have available to do a pretty compelling version of this stuff live as well assuming you had enough GPU power. I’ll search around to see if I can find any examples (and if anyone else has example available please do share, perhaps on a separate thread). Anyway, even if compelling real-time versions don’t exist today I’m confident they will in the near future.
EDIT:
Here is what I found after some quick searching.
Here is an example what could be done with faking people’s voices over a year ago. And here is some code implementing the technique discussed in the referenced paper which includes a video in the README that shows the TTS happening fast enough. I don’t see the process of adding speech recognition aspect to the pipeline to adding too much more latency to make faking someone’s voice impractical.
There is further evidence showing that doing real-time processing of voice audio based on ML models is already possible to some extent today. For example, Google is using ML techniques to compress a voice signal to a very low bitrate and then regenerate them on the other end. And clearly that is a process that is very latency sensitive if they plan to use it for live audio conversation tools like Duo. This isn’t exactly the same as changing one’s voice to another voice live, but the techniques are similar enough that I’m confident that is within reach soon (and for all I know, it may already have been done).
On the video side, there are plenty of real time examples I could give that are not great but still pretty impressive. Here is just one example. I’m confident these methods will get better and more convincing over time. Keep in mind that if you tolerate a little added latency then that gives the computer more time to do a higher quality version. One could easily justify the added latency for responses by pretending they have a worse internet connection than they really do. Furthermore, if one intentionally makes an effort to make more subtle and slower facial movements (shouldn’t be too difficult when the main point of the meeting is to have a conversation) then it could perhaps come across as more realistic given the current computing limitations.
2 Likes
As long as their deep faked appearance is consistent and reputation forming does it matter I wonder?
Edit: more of a worry could be bots that pass the turing test on zoom, first building relationships on telegram, then get voted in. How far away would that be?
If the process has enough participants there should be an IRL component at some stage maybe
1 Like
I didn’t mean to hijack this topic to talk about deep fakes and other Sybil attack methods that can be used against EdenOS. So I created a thread specifically about that: Sybil attacks against EdenOS (particularly with assistance of technology).
1 Like
Latency doesn’t help except for the first frame only reducing frame rates helps increase cpu time per frame. Latency may allow greater parallelism if you can have interlaced computation.
1 Like
The beginning of an anti-fragile future for EOS.
3 Likes
Can’t wait to get started, bring it on 
1 Like
Dan,
I do not like approach of EOS 5k for many accounts as it’s difficult to do smth decent for this money on another hand there is high chance that the person will just give up and will not finish the product. Why not to go another way, similar to ordinary VC funds? For example there will be setup 3 onchain funds (or any number of funds), the community based on your approach will choose managers of these funds (if the fund perform bad community will select other managers for this fund), developers will pitch these funds and managers will disburse (using multisig) funding in small amounts to developers according to the road map.
Thanks inn advance for your reply!
3 Likes
This sounds great. I’m excited to see how the selection process plays out and what its success would mean for wider society. I think it’s a case of moving ahead with it and making corrections along the way if required.
Is there any more information you can share about the EOS Alliance and its members? Brock is a huge win, but would be interested to know who else is being considered, and what the process is for decided who the members should be. What are the goals and responsibilities of the EOS Alliance following the creation of EdenOS in the long-term?
1 Like
Very supportive - nicely designed governance process. And a big note of thanks and recognition to you Dan, for locking up that amount of working capital (loan) for a year or so.
My questions (still playing catchup here) are around flow of funds, and authorisation thereof. If I understand correctly the EOS Alliance proposes to the BP’s that it can utilize eosio.ramfee for funding. What was the original intent for eosio.ramfee and why is that controlled by the BP’s? I was under the impression that eosio.bpay & eosio.vpay were the BP specific accounts? I realise this question exposes me as a complete newbie, so please be understanding.
Going forward is eosio.saving still envisioned and will this fall under the mandate of EOS Alliance, the BP’s etc? Is it still meant to fund worker proposals?
From Q3-22, you mention a budget of 0.25% of EOS supply. I’m assuming this is ultimately from inflation but from which account does that come from? Again, it would seem that the BP’s are entirely in charge of whether or not funding is allocated, but of course as EdenOS grows it can vote in BP’s that are aligned with developing the EdenOS community. Right?
Lastly, all of this is very self-referential. My interest lies in sucking in fiat currency i.e. USD → EOS. I am cracking on with some ideas in this regard, but in the meantime this all looks very exciting.
2 Likes
Smart contracts already exist on EOS. Why are you not using smart contract on EOS BlockChain, is there anything wrong here?
The roadmap is too long for a small task, using EOS smart contract within 1 week to complete the voting function. Are you wasting time?
1 Like
smart contracts are only 5% of the challenge.
1 Like
It was a community fund to be managed by elected producers.
1 Like
More information will be coming out in the next couple of weeks
3 Likes
These people are given budgets which the should be using to fund existing teams and apps, not just starting something new. They could give to anchor or dfuse or wombats or ad campaign.
1 Like
What if you don’t spend the $2 million, will you still receive as reimbursement 400,000 EOS? And how can we account for your expenditure?
1 Like
Hmm, nice roadmap. Making me think about how I would allocate 5k EOS. I would contact a handful of successful app groups, and reach out to this group too for guidance.
For example, when Dan mentioned the API need for decentralization.
Bottom line, I think I may be able to confidently allocate 5000 EOS.
Interesting. My first reaction was that I wouldn’t know how to appropriate those funds. But after thinking about it, I think I could find a really good fit. I’m familiar with like 100 different efforts over the last few years.
One issue: the total amount of money across all members does not seem like much. But it is significant. And we will learn a ton in one year.
I definitely like it.
How can interested users sign up to receive an invitation?
3 Likes