A learning experience | AAAI-22 in review
Hey there! As you might have guessed, it's time for my review of the AAAI-22 conference(?) (Association for the Advancement of Artificial Intelligence) I attended recently. It's definitely been a learning experience, so I think I've got my thoughts in order in a way that means I can now write about them here.
Attending a conference has always been on the cards - right from the very beginning of my PhD - but it's only recently that I have had something substantial enough that it would be worth attending one. To this end, I wrote a 2 page paper last year and submitted it to the Doctoral Consortium, which is a satellite event that takes place slightly before the actual AAAI-22 conference. To my surprise I got accepted!
Unfortunately in January AAAI-22 was switched from being an in-person conference to being a virtual conference instead. While I appreciate and understand the reasons why they made that decision (safety must come first, after all), it made some things rather awkward. For example, the registration form didn't mention a timezone, so I had to reach out to the helpdesk to ask about it.
For some reason, the Doctoral Consortium wanted me to give a talk. While I was nervous beforehand, the talk itself seemed to go ok (even though I forgot to create a slide somewhere in the middle) - people seemed to find the subject interesting. They also assigned a virtual mentor to me as well, who was very helpful in checking my slide deck for me.
The other Doctoral Consortium talks were also really interesting. I think the one that stood out to me was "AI-Driven Road Condition Monitoring Across Multiple Nations" by Deeksha Arya, in which the presenter was using CNNs to detect damage to roads - and found that a model trained on data from 1 country didn't work so well in another - and talked about ways in which they were going to combat the issue. The talk on "Creating Interpretable Data-Driven Approaches for Tropical Cyclones Forecasting" by Fan Meng also sounded fascinating, but I didn't get a chance to attend on account of their session being when I was asleep.
As part of the conference, I also submitted a poster. I've actually done a poster session before, so I sort of knew what to expect with this one. After a brief hiccup and rescheduling of the poster session I was part of, I got a 35 minute slot to present my poster, and had some interesting conversations with people.
Technical issues were a constant theme throughout the event. While the Doctoral Consortium went well on Zoom (there was a last minute software change - I'm glad I took the night before to install and check multiple different video conferencing programs, otherwise I wouldn't have made it), the rest of the conference wasn't so lucky. AAAI-22 was held on something called VirtualChair / Gather.town, which as it turned out was not suited to the scale of the conference in question (200 people in each room? yikes). I found myself with the seemingly impossible task of using a website that was so laggy it was barely usable - even on my i7-10750H I bought back in 2020. While the helpdesk were helpful and suggested some things I could try, nothing seemed to help. This severely limited the benefit I could gain from the conference.
At times, there were also a number of communication issues that made the experience a stressful one. Some emails contradicted each other, and others were unclear - so I had to email the organisers at multiple points to request clarification. The wording on some of the forms (especially the registration form) left a lot to be desired. All in all, this led to a very large number of wasted hours figuring things out and going back and forth to resolve confusion.
It also seemed as though everyone appeared to assume that I knew how a big conference like this worked and what each event was about, when this was not the case. For example, after the start of the conference I received an email saying that they hoped I'd been enjoying the plenary sessions, when I didn't know that plenary sessions existed, let alone what they were about. Perhaps in future it would be a good idea to to distribute a beginner's guide to the conference - perhaps by email or something.
For future reference, my current understanding of the different events in a conference is as follows:
- Doctoral Consortium: A series of talks - perhaps over several sessions - in which PhD students submit a 2 page paper in advance and then present their projects.
- Workshop: A themed event in which a bunch of presenters submit longer papers and talk about their work
- Tutorial: In which the organisers deliver content centred around a specific theme with the aim of educating the audience on a particular topic
- Plenary session: While workshops and tutorials may run in parallel, plenary sessions are talks at a time when everyone can attend. They are designed to be general enough that they are applicable to the entire audience.
- Poster session: A bunch of people create a poster about their research, and all of these posters are put up in a room. Then, researchers are designated specific sessions in which they stand by their poster and people come by and chat with them about their research. At other times, researchers are free to browse other researchers' papers.
Even though the benefit from talks, workshops, and other activities at the conference directly has been extremely limited due to technical, communication, and timezoning issues, the experience of attending this conference has been a beneficial one. I've learnt about how a conference is structured, and also had the chance to present my research to a global audience for the first time!
In the future, I hope that I get the chance to attend my first actual conference as I feel I'm much better prepared, and have a better understanding as to what I'm getting myself in for.