TV Un-Festival: Ian Clarke

Degree in computer science and AI from Edinburgh, and designed a project to exchange info in countries where there is censorship. This became Freenet. Non-profit corp in USA, since 1999. 2 Million downloads of software. One of the things that happened with Freenet, even though it was designed as freedom of speech idea – was at same time as Napster – it was perceived as indestructible Napster as it was designed so that it couldn’t be shut down.

Has been thinking, how do you enforce copyright onlne, and the answer is that you can’t. Enforcement of c is about preventing people sharing information when they don’t have permission to. So if you can’t enforce copyright, what is the alternative?

Founded Revver in 2005, to help foster an environment of creativity online, want an ecosystem where creators can be paid for their work. Interested in online video, this was at the time when broadband was starting to make this possible, devised a way to attach unobtrusive ads to end of videos, release under CC no-derivs, attrib, licence, and share the revenue on a 50-50 basis with creators, or 40-20-20 if there was an affiliate, so produce financial incentive to share video.

A lot of the people you may have heard of use or have used Revver, e.g. LonelyGirl15 for about a year, ZeFrank. Almost every well known video blog has used Revver, except Rocketboom.

Whilst at Revver, got curious of how to figure out what people are interested in and show it to them. Not a new problem. Started to look at collaborative filters, basically a system which looks at your behaviour, perhaps what you buy on Amazon, and then recommends things to you that you might like. Amazon, NetFlicks. Problem is that they either work or scale – but they don’t do both. if they can recommend stuff well, they don’t scale, or they dumb down the recommendation so it can scale. Built a filter called Daedalus for Revver, and licensed it to Reddit.

Whilst working on this collaborative filter, noticed that collaborative filters need a lot of data before they can figure out what they are interested in. So Reddit needs people to use the website for several hours continuously. Real opportunity in online news space, with n otable exception of Reddit, no one was really doing personalised news, and those that were were using collaborative filters that are problematic, and the quality of user submitted news is extremely low. If you’re familiar by Digg, you’ll know what I’m talking about.

This was the genesis of Thoof. Alternative to collaborative filtering, figures out your interests more quickly, based on your behaviour, your browser, your approx. geographic location. There are generalisations you can make about mac vs PC users, or Firefox vs. IE, or based on geography. Built tech to recommend stuff to you, but if you see something on the website that can be improved, you can change it and fix it, although there is a voting step – you propose a change and if it survives the voting process it can be applied to the story. Raised a million dollars in seed foundation, launched in June, and traffic growing at 25% per week.

Using Freenet is like using a web browser, but slower – learnt that one of the key problems is that even when there is information available, that doesn’t mean that people will find it. It’s not just about accessing information that they know they want, but about finding information that will interest them once they know about it.

Go to URL in web browser, it’s easy to find out who’s hosting what. With Freenet, info is distributed through the network in a decentralised way, so unless an author chooses to reveal their ID, you have no way to know who they are. But threat model with Freenet at the time was that no one would know what people are doing with Freenet, but that’s not enough. What if you can be punished for just using the software, irrespective of what you are doing with it, e.g. China, so set about redesigning Freenet to use a darknet methodology, so that you could just connect to those people you know personally, so no one knows you are connected to Freenet, but through those people you become a part of a global network. Been working on this for two years, but working pretty well so far – lot of people don’t already have friends that are using Freenet so there is a way you can connect to strangers as opposed to friends if you choose to do that. Freenetproject.org

Q: Are there access points into Freenet, like SMS?

Not an SMS gateway to the best of my knowledge, are web gateways, but using a gateway is … you’re throwing away a lot of the benefit. To get the security, you have to be running Freenet on your computer. It’s not going to run on a typical mobile phone.

Q: You said it was friends of friends, if you try to keep cosiness amongst your contacts, how do you deal with infiltrators?

The only people who can cause problems for you are the people you have immediately connected to. So if your friend is stupid and connects to a government agent, that agent has no way to tell you are part of the network. Many Freenet users don’t care, because they live in the US or UK where they aren’t going to be jailed for this, and will connect to anyone. But we try to place it in the hands of the individual as to how much security they want. There’s a trade off between convenient and connected.

Q: What happens with people abuse the tool.

Any tool can be abused. But the freedom to communicate – if one person wants information and someone else wants to have it, that freedom is essential in a democracy. Our leaders are chosen by us, and in order to make effective decisions we need free information. So totalitarian countries spend a lot of money controlling their people’s ability to communicate. Any tool can be misused but the benefits outweigh the potential abused.

Q: How do people know who to trust?

When people are anonymous how do you know when to trust them? That’s a question that the internet in general gets, blogs etc. But at Freenet we address that problem with the concept of a Nym, an anonymous identity so anything you publish is signed by the nym, so you can link together discussions and content to the same person. Even an anonymous identity can build up trust. Similar to the WWW, a blogger can build up trust. This problem is not completely solved – what might be intersting to experiment with, and we may do it with Thoof in the future, and Thoof’s approach is to fix in a peer reviewed way which works well. What we’re likely to do in the future is to create a ‘web of trust’ so can build up trust based on performance. So if you propose a change and it’s rejected, that decreases your trust level, but if you propose a lot of changes and they all get voted through you’d get more trust. So maybe then stories you right are promoted more quickly, but the mechanisms within the site will get rid of it really quickly.

Q: Where are Thoof stories from?

You can submit anything that has an URL. YouTube, BBC article, anything. What goes on Thoof itself is a title, description and tags. Intention is that the title and description will be impartial and unbiased description of what’s being linked to, but the thing being linked to can be expressing an opinion. Can edit review, title, and even URL.

Technorati Tags: