CLAIRE 18 months in
Precisely 18 months ago, CLAIRE was officially launched. It feels much longer than that, since so much has happened since, especially over the last 12 months. In the following, we give you an update of some key achievements and developments, as well as a brief outlook to 2020.
In particular, we cover the following:
- The CLAIRE research network
- ICT-48 Proposals
- International Advisory Board
- Endorsements from governments and national contact points
- Collaboration with the European Space Agency
- CLAIRE offices and headquarters
- Zulip community platform
- Work with the European Commission and Parliament
- Next steps
- Appendix: CLAIRE’s efforts on collaborating with ELLIS on ICT-48 proposals
But before we do so, we would like to thank you all for you ongoing support for our shared vision – a vision that has gained much traction over the past year, and one that’s certainly worth every hour that countless volunteers from the CLAIRE community have put into. CLAIRE is all of us – that is, all of you – and what we can do together. And looking back, that seems to be a lot. Enough for a rather extensive update, but we’ve tried to organise it in a way that makes it easy to find the topics you might be most interested in. Of course, we think it’s all pretty exciting … 😉
Before we get started, a brief reminder that we post regular updates on Twitter (@vision_claire, please follow us there if you have not done so already) and – preferably – Zulip (contact email@example.com if you don’t have access yet, and see instructions below for setting up auto-forwarding of Zulip messages to your e-mail account.)
The CLAIRE research network. Firstly, as you may know, CLAIRE is now the world’s largest AI research network. At 341 member groups and institutions (and counting), ranging from small groups to large institutes, the network jointly covers over twenty thousand AI researchers and staff, with a total annual budget of well over 300 million Euros per year for AI research in Europe, or an estimated 1.5 billion Euros over the next five years. This surpasses, by far, what we had expected when we started building the research network a mere 12 months ago, and is testament to the enormous strength Europe has in terms of research across all areas of AI.
Of course, 300 million per year is less than a third of what we estimate is needed to keep Europe at the forefront of AI research – CERN, for example, has an annual budget of about 1 billion Euros, and that’s “just” particle physics (well, don’t tell them that, since they will remind you that part of this paid for the invention of the WWW, and similar side benefits), and it only covers the additional budget they receive for the large joint projects they are doing together at CERN in addition to the individual budgets of their distributed network of researchers. So, there’s a sizeable gap between what we collectively have and what we’d like to see in terms of public funding of fundamental AI research – not even to speak of AI innovation, which comes on top of that.
ICT-48 Proposals. To address this funding gap mentioned above, at least to some extent, over the past 12 months, we’ve taken some important initial steps. Notably, in the context of ICT-48-2020, a call widely believed to be instrumental for preparing the ground for more significant funding in the near future, CLAIRE has been very active. Specifically, CLAIRE members and supporters were deeply involved in at least 6 of 14 submitted ICT-48-2020 RIA proposals, and had some involvement in several others. CLAIRE members and supporters were also involved in both CSA proposals, with one of those (titled VISION) developed under the leadership of CLAIRE. Although the funding under ICT-48 is very limited, we are confident that its level of participation in this call positions CLAIRE well for future opportunities.
There have been questions from quite a few CLAIRE members and supporters regarding coordination with ELLIS in the context of ICT-48. In a nutshell, we seriously tried to work together with ELLIS on a joint CSA proposal and on a RIA proposal on foundations of trustworthy AI, but for reasons we don’t fully understand, this did not work out. We note, however, that lately, ELLIS appears to have embraced a focus on human-centred, trustworthy AI. We see this as a major step forward, not just for ELLIS, but also in terms of prospects of future collaboration on the topic (as recently as September, CLAIRE’s long-standing commitment to a strong focus on human-centred AI appeared to be perceived by ELLIS as an obstacle to collaboration on an ICT-48 RIA proposal).
International Advisory Board. On the occasion of the opening our headquarters last week, we were thrilled to announce our newly formed international advisory board, whose members are (in alphabetical order):
- Noriko Arai, Director of the Research Center for Community Knowledge at the National Institute of Informatics in Japan;
- Frédérick Bordry, Director for Accelerators and Technology at CERN in Switzerland;
- Alan Mackworth from the University of British Columbia in Canada, former president of AAAI;
- Francesca Rossi, AI Ethics Global Leader and Distinguished Research Staff Member at IBM Research in the USA;
- Robert-Jan Smits, President of the Executive Board of Eindhoven University of Technology in the Netherlands, former director-general of research and innovation at the European Commission and one of the main architects of Horizons 2020;
- Manuela Veloso, Head of AI Research at J.P. Morgan and Professor at Carnegie Mellon University in the USA, former president of AAAI;
- Wolfgang Wahlster from Saarland University in Germany and former CEO and Scientific Director of the German Research Center for Artificial Intelligence (DFKI); and
- Toby Walsh, Scientia Professor at the University of New South Wales in Australia and Guest Professor at the Technical University of Berlin in Germany.
Jointly, the 8-person board brings a broad range of experience from academia, industry and major European institutions, including the European Commission and CERN to CLAIRE. We expect that advice and guidance from this high-calibre group of experts will play a crucial role in helping us with the next steps towards achieving our bold and ambitious vision for European excellence in AI.
Endorsements from governments and national contact points. In November 2018, we received a strong letter of support from the Italian government. Over the course of 2019, the governments of eight further EU countries officially endorsed our vision: Belgium and Czechia (in February), Slovakia (in April), Luxembourg (in May), Spain (in July), Finland (in August), Greece (in September) and The Netherlands (in October). We know of two further letters that are tantalisingly close to completion, and we expect to receive these within the next few weeks. Because part of our vision will require strong financial backing from national governments, these letters are very important stepping stones, and it is thus very encouraging to see increasing momentum in this regard.
To further strengthen the operation of CLAIRE on the national level, we are now putting into place national contact points in all countries in which there is major support for CLAIRE. These national contact points form a natural interface between CLAIRE, the AI ecosystems in the individual countries, the national governments and the national AI associations (which also endorse the national contact points and, of course, are one of our main connections to EurAI).
Collaboration with the European Space Agency. Over the last 12 months, we have built strong links with the European Space Agency (ESA) – one of Europe’s major success stories and an organisation with a keen interest in AI research and innovation. Following a very successful theme development workshop on AI and space in early March, we’ve enjoyed enthusiastic support from the highest level of ESA, as is evident from a message sent to us in June by Johann-Dietrich Wörner, Director General of ESA European Space Agency, in which he wrote: “CLAIRE has planted the flag for Europe’s ambitions on AI, and we congratulate you. […] I am sure our collaboration with CLAIRE will help us realise Europe’s ambitions for space technologies and on Earth, and for the advancement of AI in all our Member States.” In the autumn, there have been two further theme development workshops related to climate science, which form the first steps of an emerging CLAIRE focus on AI and climate. CLAIRE and ESA have officially formed a Special Interest Group on AI and Space, whose steering committee met in early December to map out activities over the next 12 months – stay tuned, there will be many opportunities to get involved (we will announce these on our Zulip community platform – see also below)!
CLAIRE offices and headquarters. As per previous updates, CLAIRE has opened administrative offices in Saarbrücken (DE), Prague (CZ) and Rome (IT), as well as headquarters in The Hague (NL). These offices increase CLAIRE’s capacity to organise events (such as the CLAIRE symposium in February), to maintain and extend our website and to operate our Zulip community platform. The offices also play an important role in engaging with the AI communities in the respective countries. We are very grateful to the institutions that fund these offices. The offices have already proven to be very valuable to us, and as our organisation takes further shape, we expect to constantly adapt the way they operate to support CLAIRE’s mission. We expect several additional offices to be opened in the near future, to strengthen CLAIRE’s geographical presence across Europe and to further define the focus for the work done by each of the offices (see “Next steps” below), for the benefit of the CLAIRE community (i.e., all of you).
Zulip community platform. Since June, we’ve been operating our community platform, based on Zulip, an open-source alternative to Slack. We chose Zulip primarily because we wanted a flexible, open-source and cost-effective solution that can be hosted in Europe. Initial uptake has been good, but we would very much like to see this used a lot more, by you, in a bottom-up fashion, for announcements and discussions of all sorts. We have set up streams for a variety of topics and purposes, such as news, job postings, grant opportunities and topics of relevance to CLAIRE supporters and members in specific countries. We’ll be happy to set up additional streams upon request (just post to #zulip help) We understand that some of you are reluctant to deal with yet another communications tool or platform, although we still encourage you to try out Zulip, which is light-weight and easy to use. Still, if you prefer e-mail, all you have to do after you have your account (for this, please contact firstname.lastname@example.org) is to log in once and set up e-mail notifications (Settings menu under the cogwheel icon on the top right / Notifications / Email / check all four boxes).
Work with the European Commission and Parliament. This, of course, has been a major focus of our activities. 10 supporters of CLAIRE are members of the Commission’s High-Level Expert Group on AI (HLEG-AI) and have been deeply involved in preparing two key documents: the ethics guidelines for trustworthy AI (published in April) and the policy and investment recommendations (published in June). For the former, there was a formal process for community input, which many of you have actively participated in. For the latter, unfortunately, no such process was offered by the Commission, so our input was given through individual HLEG-AI members.
On 15 November, EC EVP Margrete Vestager was given a briefing on CLAIRE by Holger and Morten, on her request, and this was followed up with additional information. This was our first chance to pitch the CLAIRE vision at the highest level of the Commission, and according to Commission participants at the table, our presentation was very well received. Going forward, Robert-Jan Smits, member of our newly formed international advisory board and former director general of the Commission, will provide us with expert advice on working even more closely with the Commission.
The Commission has worked on establishing a Commission-internal proposal for a public-private partnership (PPP) in AI. A PPP is a particular instrument the commission uses for mobilising innovation in industry and for securing substantial co-financing from industry. CLAIRE decided to engage with this process for three reasons: The first is that we believe that AI is an important part of an AI PPP (the commission had engaged with two existing PPPs that do not represent the core AI community to develop the proposal), the second that we believe a strong participation from research is important for successful innovation, and the third that the target size of the programme is substantial. CLAIRE developed and published a document with ten recommendations for an AI PPP (please see our website), and was asked in the twelfth hour to assist with internal preparations of a PPP-related document through our core team member, Morten Irgens. CLAIRE expects to be engaged if the Commission decides to move this forward.
In early December, we started reaching out to members of the European Parliament, to increase awareness of and support for CLAIRE. This complements work done by some of our supporters at the national level aimed at connecting CLAIRE to the national AI strategies.
Next steps. If there’s one thing we’ve learned over the last few months it is the degree to which we cannot predict what’s going to happen next with CLAIRE. Of course, this does not mean that we shouldn’t make plans. We expect additional offices to be opened in Oslo (NO), Zürich (CH), Brussels (BE), Paris (FR) and Cork (IE), with dedicated staff for EU liaison, for work with mainstream and social media, for outreach and citizen engagement, for networking and event support, and for technical support (Zulip, website, …).
A major focus will be on building an industry network to complement our large and strong non-profit research network. Considering that 871 of our 3300 individual supporters (i.e., 26%) work in companies, ranging from small startups to multinational corporations, we feel there is substantial interest in the CLAIRE vision and organisations from industry. As of last week, we have an informal advisory group (IAG) on industry (see https://claire-ai.org/iags/#IND) that will take the lead, supported by staff in the Saarbrücken office and our HQ in building the industry network, which will complement our research network and create opportunities for cooperation, projects and strategic work on key applications of AI (e.g., in through theme development workshops).
To allow CLAIRE to operate as a legal entity (which is important in the contact of industry contributions, donations, operating offices, etc.), we are currently setting up a non-profit international organisation under Belgian law (the same legal form used by EurAI and many EU-related entities). In order to get something into place quickly, the extended core team decided on a light-weight temporary construct that will ultimately be modified into a longer-term structure, based on input from the CLAIRE community (more on this to follow soon). In the meanwhile, for technical reasons, we will soon request the current members of our research network to officially join the new association.
We are actively working on endorsements from additional national governments, with some very close to being officially issued. Help with this, from anyone with good political connections, is always very welcome. These endorsements are very effective in raising our profile nationally, at the EU level and beyond; they also give us some influence on national AI strategies.
We are planning several community meetings, including a large one at ECAI next summer, as well as additional theme development workshops (details will be communicated on Zulip). In addition, we encourage meetings at the national level, coordinated with the national AI associations as far as possible, and we will be happy to give administrative support for these.
Finally, we will focus on intensifying our already very strong connections with other key players – notably, EurAI, AI4EU, the HumanE AI consortium, ELLIS, the BDVA and euRobotics – as well as with the networks that will be established by the successful ICT-48 RIA consortia. Our goal in all of this is to help overcome fragmentation, to coordinate between different groups in order to achieve maximum benefit for the European AI community as well as maximum impact for “AI made in Europe”.
Appendix: CLAIRE’s efforts on collaborating with ELLIS on ICT-48 proposals
Over the last few weeks, we have received quite a few questions regarding CLAIRE’s coordination with ELLIS in ICT-48-2020 proposals. Here is a rather detailed account of this, based on notes and e-mail messages on the process reviewed by CLAIRE extended core team members. We also shared a slightly condensed version of this account with the ELLIS board, in the interest of clearing up some confusion that seems to have arisen on this topic.
Following our announcement of CLAIRE’s ICT-48 strategy, which included an invitation to ELLIS to cooperate on a RIA proposal on foundations of AI and on a CSA proposal, to which ELLIS responded positively, we were notified in early July by the ELLIS board that ELLIS preferred primarily focussing on their own proposal, and to explore, in parallel, the possibility of joining the two proposals into one. We were also informed that no one on the ELLIS side could be found to join us in writing a CSA proposal.
In mid-August, we reached out again to ELLIS, offering cooperation on a joint RIA proposal, a CSA proposal and two other topics. ELLIS then expressed interest in discussing a possible merger of the two RIA proposals on foundations. These discussions were held by Arnold Smeulders and Sami Kaski (as leaders of the ELLIS proposal), Fredrik Heintz (as coordinator of CLAIRE’s TAILOR proposal) and Holger Hoos (as the core team member most deeply involved in CLAIRE’s overall ICT-48 efforts). Both ELLIS representatives declined requests for informal discussions before a more formal meeting, which we felt would have helped to quickly find common ground.
The ELLIS team lead the negotiations, proposing a combination of the two RIA efforts that would lead to fully symmetrical representation of CLAIRE and ELLIS. The CLAIRE team was prepared to accept this, under the condition that a highly visible focus on human-centred, trustworthy AI could be preserved and that the coordinator role would either be jointly taken by CLAIRE and ELLIS, or that it would go to Fredrik Heintz on the CLAIRE side, based on our perception that the CLAIRE proposal was broader and further developed at that point. After discussions on both conditions, which were complicated by the fact that EU rules require a single coordinating institution, and several tweaks, it appeared that the ELLIS side did not want to focus on human-centred, trustworthy AI, and that they were unwilling to consider the proposal to make Linköping University, which has both CLAIRE and ELLIS supporters, the coordinator of a joint proposal. From the CLAIRE perspective, there was still room for further discussions on both topics, but the ELLIS team decided to abandon this line of negotiation and put a new proposal on the table. This so-called “room in your house” proposal was to give ELLIS full control over one work package, with sizeable budget, in the CLAIRE-led TAILOR proposal.
Despite concerns within CLAIRE regarding the outside perception of this solution and the degree to which it would enable meaningful scientific cooperation between CLAIRE and ELLIS (something we wanted to achieve in a joint network), on 14 September, CLAIRE accepted this proposal in full. Without any further discussion, on 19 September, we received a message from the ELLIS negotiation team telling us that, after careful consideration of all information, they were no longer willing to go ahead with their own “room in your house” proposal. Much later, ELLIS explained that the reason was that the “room in the house” proposal was dependent on ELLIS securing rooms in more houses than one, which they had not been able to do. At the time, there was also a clear indication that ELLIS was not interested in further discussions (e.g., regarding the earlier line of negotiation).
Following this, the CLAIRE and ELLIS teams submitted two separate, uncoordinated proposals on foundations of AI. As always intended, CLAIRE also submitted a CSA proposal (VISION) and, in this context, reached out to all ICT-48 RIA consortia we had knowledge of for coordination with the VISION proposal. Letters of support were received from 9 of the 14 RIA consortia that ultimately submitted proposals, including the ELLIS-led ELISE consortium. The CSA proposal also included letters of support from EurAI, the Big Data Value Association, euRobotics, the Open Machine Learning Foundation, and the Italian Ministry of Education, University and Research.
CLAIRE’s TAILOR RIA and VISION CSA proposals both contain mechanisms for involving researchers and groups beyond those involved in the consortia that end up being funded. This was done to ensure that the entire European AI community is involved to the largest possible degree in the networks of centres of excellence to be established under ICT-48, and it includes, of course, all members and supporters of CLAIRE and ELLIS.