We are introducing a monthly digest to keep you up-to-date with the latest happenings in the AI world. You can catch up with any AIhub stories you may have missed, get the low-down on recent conferences, and generally immerse yourself in all things AI.
The big news from AIhub is that we recently launched our first ever focus series: “AI for good: UN sustainable development goals”. Each month we will be concentrating on a different sustainable development goal (SDG) and bringing you a collection of work from people in the field. Our first topic was SDG number 3: good health and well-being. Here are some highlights from the series:
In February we turn our attention to climate action, which is the UN SDG number 13. We will be posting articles on that topic throughout the month so don’t forget to check this page for all the latest content.
On the topic of climate action, the folks at Climate Change AI have introduced a new feature to their newsletter: “Dataset of the Month”. To begin, they have picked ClimateNet: an expert-labelled open dataset and deep learning architecture for enabling high-precision analyses of extreme weather. You can read the paper about the dataset here.
If you are interested in finding a collection of weather and climate datasets for AI research in one place then this new website, Pangeo ML Datasets, could be just the ticket. It includes pre-processed datasets, raw datasets, and some hybrid ML-physics models. Anyone can contribute by adding a new dataset.
January saw the virtual staging of IJCAI-PRICAI 2020. Thousands of delegates descended on GatherTown and were treated to invited talks, workshops, tutorials and panel sessions, and, what’s more, a programme of cultural events such as origami, live Japanese drumming and an interactive art gallery. There were a vast number of talks and posters, and meandering through GatherTown, getting lost, and stumbling across interesting work did, in some way, recreate a real conference environment. You can find papers from the technical sessions here.
There was an interesting panel discussion on the One Hundred Year Study on Artificial Intelligence (AI100). The mission of AI100 is to launch a study every five years to track how artificial intelligence propagates through society and how it shapes our lives. Read our write-up here.
During the conference opening ceremony, a number of prize winners were announced, including the IJCAI-JAIR Best Paper Prize and the AIJ 2020 Classic Paper Award. Find out who won what in our summary post.
Conferences are coming thick and fast at the moment and this month we’re attending AAAI 2021 which started on 2 February. We’ll be covering the conference with a number of articles so stay tuned.
You may have seen our seminar resource where we collate any free, online, AI events that have come to our attention. See January’s post where we list forthcoming events for the months ahead in addition to recent past events. If you are interested in browsing all the events from a specific organisation, we have that covered too: check out our list of seminar providers here.
It was a good month for Natural Language Processing start-up Hugging Face who not only revealed a new interface for their model hub, along with their updated website, but also announced that they are now cash flow positive.
The eagerly anticipated paper On the Dangers of Stochastic Parrots: Can Language Models Be Too Big? 🦜 hit the virtual shelves this month. In this article, which has been accepted for FAccT2021, the authors ask what the possible risks associated with this technology are, and outline paths available for mitigating those risks.
Speaking of large language models, Google’s new 1.6 trillion parameter model was reported this month. Read more in this arXiv article Switch Transformers: Scaling to Trillion Parameter Models with Simple and Efficient Sparsity.
Those of you who are fans of the excellent Radical AI podcast will no doubt be aware that hosts Jess and Dylan have partnered with John Havens and IEEE Standards Association to produce a series of special episodes called Measurementality. They will consider how, and if, we can measure success in AI Ethics. The first episode aired on 28 January.
In October Devi Parikh posted the first of her Humans of AI: Stories, Not Stats video interviews. The idea was to better get to know prominent AI researchers as people. The final interview was published in December, but now Devi (with help from Varshini Subhash and Mukul Khanna) is creating new videos where the answers from each guest to a particular question are collated. This is a neat idea and it’s interesting to hear the variety of responses to the same question. You can find all of the original interviews and the collated answers here.
You may remember the 2020 AI Song Contest. We wrote about it last April. If you are interested in taking part in the 2021 edition, then you can find out how to enter here. The competition is open to anyone (solo or team) and the deadline for entries is 18 May 2021. Time to dust down your keyboards, crack out those MIDI cables and get composing with the help of your favourite algorithm.