ΑΙhub.org
AIhub coffee corner: The future of conferences
The AIhub coffee corner captures the musings of AI experts over a 30-minute conversation. This month we discuss conferences and whether they will ever be the same again now we’ve had a taste of the virtual.
Joining the discussion this week are: Kamalika Chaudhuri (University of California, San Diego), Tom Dietterich (Oregon State University), Sabine Hauert (University of Bristol), Carles Sierra (CSIC).
Carles Sierra: I think we will see more and more of these virtual conferences. We could probably work as well as we did when meeting physically, although some aspects will be different.
Sabine Hauert: Last week I gave three talks, at three separate conferences, and never had to leave the home, which is really great for work-life balance and childcare responsibilities.
Sabine: I have been attending ICRA (one of the major robotics conferences), and I was at a workshop yesterday which was great. The thing I missed most was the applause, and the chit-chat that happens on the side-line of these workshops. I think a lot of that we could do virtually but we just haven’t got into the habit of it. These workshops are still intimidating, especially for junior people. I did have some “backroom” discussions where, after my talk, someone privately messaged me, which was great. But that wasn’t the norm. People should use more of the little “clap” icons, and the other things that make it feel more human, so you don’t feel like you’re just talking to a screen. The issue is part software, part us – we need to start being less intimidated by online speaking. I think we’ll get better at it – I’m quite hopeful.
Carles: There are other advantages to virtual conferences. It is a bit more democratic. I organised an event where I should have physically been in Brussels, but I obviously couldn’t, so it went online. The event was broadcast through YouTube, and we had a very large number of attendees. Most of these would not have been able to attend if it was held physically in Brussels. Also, in terms of economics, everything is much cheaper. Bringing all of those people to a physical location is so much more expensive. I think the system will accept it, little-by-little. In the same way that World War I made the 19th Century move into the 20th Century – from the local to the global world, I think that COVID is getting us from the physical to the virtual world, the 21st Century. I think we are going to change a lot of things that we normally did physically to virtual. Now, teleworking; all the companies are thinking “it’s much cheaper to have all my employees at home working than having them in an office, that I have to pay the rent for, I have to pay the electricity for. The workers don’t have to come to their job by car”. They realise that many workers produce more at home than when they are in the office. Capitalism will reinvent itself, and exploit us probably, of course!
Tom Dietterich: In some Twitter conversations, people that I know in Africa are very disappointed because the chance to travel to an international conference is so valuable to them, both in terms of the culture and experience side and the technical and scientific side. Also, you can’t really participate in these meetings unless you have a good internet connection.
Sabine: You do miss meeting people too. You go to the same conference for 10 years and you don’t necessarily go there for the papers, you go to meet up with friends. That aspect is really hard to recreate.
Tom: Yes, it’s harder to recruit students or start up collaborations. I imagine you are all outside of the sessions half of the time talking to people. My diary gets full of appointments during these meetings. Necessity is the mother of invention, so maybe we’ll find ways to do this virtually.
Tom: Charles Sutton wrote up a nice blog post called “ICLR 2020: my first virtual conference” in which he reflects on how it went. I had an experience similar to Carles in that I was scheduled to give a talk at a machine learning summer school run by BigML. It was going to be an in-person event in Seville with 200 participants, and it turned into a virtual thing with 1500 participants. In terms of the audience you reach – it really lowers the bar to participation.
Carles: I think this is a massive democratic way of providing access to knowledge and information. I think it’s unstoppable. I understand the argument that it is more difficult for people in places with poor internet connection, but it will be far cheaper to improve their internet connections than to afford the millions that moving people requires. I think this is the way we’re going to go. Also, teleworking will become a massive asset of many companies.
Tom: We’ve also seen the opening up of global seminars. There’s a reinforcement learning theory seminar that is worldwide now. They meet once a week and have someone give a talk. These are the best people in the world talking in a weekly seminar. A lot of the top groups are also opening up their internal seminars and putting them online [See, for example, Harvard ML Theory]. The access to information is overwhelming.
Kamalika Chaudhuri: I’m speaking at one of these kinds of seminars in Montreal. They have this ML optimisation seminar, which they have now opened up and they have invited a bunch of people. But, this is just one example, there are many more.
Carles: The future could be: publishing in open access journals like we discussed last time, then having seminars to broadcast it worldwide. The purpose of conferences is to make the latest results available to researchers, and people had to travel to get those results. Now, why do you need to travel? You just need to listen to the main professors and the young researchers in the field and do it from home.
Sabine: Could there just be a yearly meet-up somewhere? We do all these virtual conferences then we all hang out in Hawaii for a week and there are no talks. We would already have seen all the talks throughout the year and this meeting would cover the social networking side.
Tom: With NeurIPS last year, they organised a bunch of meet-ups around the world for people who couldn’t travel to the main conference. So, another possible route is to have a number of simultaneous meet-ups, maybe linked with video, or unlinked so they can be in their own time zones.
Carles: Also, think of the political problems, such as visa problems; all this is avoided with virtual meetings, as long as you have an internet connection.
Sabine: To what extent are business models going to have to change for the organisations that do conferences? Is this going to be very difficult for them? Will they just need to reinvent themselves into online conference providers? They could maybe organise more specialised conferences on different topics throughout the year.
Carles: A lot of these conferences are a cross between a publisher and a travel agency; they would publish papers then hold a meeting somewhere. The publishing part can continue in a natural way. Look at IJCAI this year – they are publishing all the papers in July then plan to hold a physical conference at a later date. We need to keep some social contact; this is especially important for young people who are building their careers and collaborations. It’s very difficult to build a new relationship in the purely virtual world. We need to have physical meetings but we need to redesign them, and serve a different purpose than just paper presentation.
Sabine: The way that ICRA have organised their event is that it’s held over a month, rather than a week. They’ve created a Slack channel that everyone can join. There has been a lot of chat from people introducing themselves and saying what they work on.
Tom: I wonder if the conferences will just become continuous. Then it basically becomes arXiv plus a video.
Carles: Now we have a fantastic method of continuously giving our results to the scientific community, almost on a real-time basis. We can finish our work, upload to arXiv, we can record a video and make it available to everyone. Conferences will need to be completely reshaped.
tags: coffee corner
AIhub is supported by:
