3 ms·

You hit at the fundamental difference between my approach and others, which may make mine better or worse (certainly worse if nothing ever comes of it, and 'rev

by Jd 18y ago

You hit at the fundamental difference between my approach and others, which may make mine better or worse (certainly worse if nothing ever comes of it, and 'revolutionary' if it does and it is a fundamentally new way of looking at things with some utility). Here is an example of why I would start with modeling the human side (essentially sociological) instead of simply modeling the data. Suppose what we want to model and improve is an academic journal. What will our data be? Submissions to the journal? Articles published by the journal? Should corrected articles re-submitted also be included? What we have is essentially data coming to a node (a human editor) and then sorted by viability, then sent back to the originators when a certain criteron is reached (an article that needs corrections in order to be published), the item is returned in better form, and then various items are collated and released as a journal. The essential work here is the human side, and I know of no possible way to capture the necessary qualitative aspect merely by analyzing the data stream. For instance, is there anything we can say simply from analyzing Twitter feeds without attaching labels assigning some worth? We can talk only about quantity, but not quality. And if we talk about quality (e.g. we create a point system and posts that Pg or Chad Fowler reads are worth twice as many points in our system) then we are introducing a human element, that may in some sense be arbitrary, or even if it is not arbitrary it does not follow necessarily from the our purely mathematical modeling structure. Admittedly in my case study, the academic journal, we also haven't discussed anything about the type of journal and whether or not there is quality of content. That is because I desire to provide a platform to allow different groups of people to assign their own value to things they find important, even if I might not find it especially so myself (i.e. post-modern literary criticism).

physcab 18y ago
I understand the issue you are trying to get at, but I am still unclear as to the actual deliverable. It seems like your idea is a mashup of a human-powered element (think Amazon's Mechanical Turk, ChaCha, or Mahalo Answers) and a computer generated part (think Google or Artificial Neural Networks/Machine Learning). With the computer generated part, you can assign weights or penalties to establish some "worth" (e.g in my Twitter example, you can assign more points to Robert Scoble's tweets than mine). You can also try to simulate context. (e.g If a person said they "bought fish" did they actually buy fish or did they misspell "Phish"?) So here is my next question. Which specific problem are you trying to address? Being as clear as possible will help you move to the next step. No one on Hacker News is looking to steal ideas...I think we all have some "revolutionary" project we are currently working on. :)
- Jd 18y ago
  Clay Shirky describes that online communities have a natural life cycle in which they start off small and vibrant and then dissipate as they grow larger. The perpetual problem is to achieve a persistent high signal to noise ratio. I suggest that a platform which allows persons to independently form communities around common interests (when then could generate their own news feed or journal) will allow new group formation and allow each group to control what is or is not an acceptable signal to noise ratio in their group. For instance, 'Nuclear Physicist News' might not resemble 'Hacker News,' might not resemble 'Recent findings in Biotechnology.' But all of those groups should be able to exist, fail, and be formed again on a single platform. What makes the platform part important is that my single identity carries through all of the different groups I participate in, allowing me to participate in (and read) more groups, more easily. In some ways this might be better described as a protocol than a platform and it could theoretically exist in either form. Part of the reason I want to communicate with persons more technically gifted than I is about whether or not this would exist better distributed over many servers with only a single unified protocol (..ML) or like facebook/your-favorite-social-networking-site-here. Here's a restatement of the problem: social networking sites do not currently facilitate high level discussion.
  - owkaye 18y ago
    I suggest that a platform which allows persons to independently form communities around common interests (when then could generate their own news feed or journal) will allow new group formation and allow each group to control what is or is not an acceptable signal to noise ratio in their group. It almost seems like you've never heard of Yahoo Groups.
  - physcab 18y ago
    Ok so you want to develop some "high quality" social network then. Why not just create one on Ning? Ning allows you to create your own social network. And so what if you create this "high quality" social network. What is the added-value that someone cannot get through personal messaging, forums, e-mail, and current social networks. You also talked about integrating your own identity. How is this different than MyspaceID, Facebook Connect, Google Friend Connect, and Open ID? These services allow you to be interconnected so that your reputation and presence among other social networks is constant. Lastly, I believe we're going to see a transformation of social networks. I agree with that part of your initial statement. But this transformation will arise from a gap in a practical need. You have not fully convinced me that there is a gap. So just to give you some brief, unsolicited advice, I would say that you should look for problems that you encounter in your daily life, and create some tool to fix them. If its important enough, then your userbase will follow. I would also research all the different types of social networks and how they currently function, especially from a technical point of view. You tend to throw a lot of technical jargon around. I want to give you the benefit of the doubt, but you have to be clear about your terminology. For example, when you mention "signal-to-noise" ratio, I think gaussian noise and signal processing. :)

by Jd 18y ago

physcab 18y ago

I understand the issue you are trying to get at, but I am still unclear as to the actual deliverable. It seems like your idea is a mashup of a human-powered element (think Amazon's Mechanical Turk, ChaCha, or Mahalo Answers) and a computer generated part (think Google or Artificial Neural Networks/Machine Learning). With the computer generated part, you can assign weights or penalties to establish some "worth" (e.g in my Twitter example, you can assign more points to Robert Scoble's tweets than mine). You can also try to simulate context. (e.g If a person said they "bought fish" did they actually buy fish or did they misspell "Phish"?) So here is my next question. Which specific problem are you trying to address? Being as clear as possible will help you move to the next step. No one on Hacker News is looking to steal ideas...I think we all have some "revolutionary" project we are currently working on. :)

Jd 18y ago
Clay Shirky describes that online communities have a natural life cycle in which they start off small and vibrant and then dissipate as they grow larger. The perpetual problem is to achieve a persistent high signal to noise ratio. I suggest that a platform which allows persons to independently form communities around common interests (when then could generate their own news feed or journal) will allow new group formation and allow each group to control what is or is not an acceptable signal to noise ratio in their group. For instance, 'Nuclear Physicist News' might not resemble 'Hacker News,' might not resemble 'Recent findings in Biotechnology.' But all of those groups should be able to exist, fail, and be formed again on a single platform. What makes the platform part important is that my single identity carries through all of the different groups I participate in, allowing me to participate in (and read) more groups, more easily. In some ways this might be better described as a protocol than a platform and it could theoretically exist in either form. Part of the reason I want to communicate with persons more technically gifted than I is about whether or not this would exist better distributed over many servers with only a single unified protocol (..ML) or like facebook/your-favorite-social-networking-site-here. Here's a restatement of the problem: social networking sites do not currently facilitate high level discussion.
- owkaye 18y ago
  I suggest that a platform which allows persons to independently form communities around common interests (when then could generate their own news feed or journal) will allow new group formation and allow each group to control what is or is not an acceptable signal to noise ratio in their group. It almost seems like you've never heard of Yahoo Groups.
- physcab 18y ago
  Ok so you want to develop some "high quality" social network then. Why not just create one on Ning? Ning allows you to create your own social network. And so what if you create this "high quality" social network. What is the added-value that someone cannot get through personal messaging, forums, e-mail, and current social networks. You also talked about integrating your own identity. How is this different than MyspaceID, Facebook Connect, Google Friend Connect, and Open ID? These services allow you to be interconnected so that your reputation and presence among other social networks is constant. Lastly, I believe we're going to see a transformation of social networks. I agree with that part of your initial statement. But this transformation will arise from a gap in a practical need. You have not fully convinced me that there is a gap. So just to give you some brief, unsolicited advice, I would say that you should look for problems that you encounter in your daily life, and create some tool to fix them. If its important enough, then your userbase will follow. I would also research all the different types of social networks and how they currently function, especially from a technical point of view. You tend to throw a lot of technical jargon around. I want to give you the benefit of the doubt, but you have to be clear about your terminology. For example, when you mention "signal-to-noise" ratio, I think gaussian noise and signal processing. :)