My attendance to the JENAS symposium in Madrid this week provided me with the opportunity to meet some of the senior colleagues who will influence the future development of technologies for fundamental research in the coming decade and more. Over coffee-break discussions, poster sessions, and social dinner I exploited the situation by stressing a few points which I have come to consider absolutely crucial for our field. 

Of course I am moved not only by caring for the progress of humanity but also by the fact that I would like the research plan I have put together in collaboration with a few colleagues to succeed... Ultimately, the two things are very well aligned though!

Besides engaging stakeholders with discussions of future developments in the field, I also exploited the symposium to talk to the full audience. Below I report two comments I made after talks by Pietro Vischia (Univ. cath. De Louvain) and Ian Shipsey (Oxford University). The first talk was a description of the research program of the MODE collaboration, of which I am the founder and scientific coordinator -that talk was indeed our best chance to highlight the need to pay more attention on today's capabilities of artificial intelligence tools when we design our future experiments. The second talk was an overview of the detector developments that are planned to address the needs of future large experiments at the high-energy frontier of particle physics, by my friend Ian Shipsey.

Here is the comment I made after Pietro Vischia's talk (whose slides -and probably also recordings?- are/will be available at the web site of the symposium):

"I would like to make a comment on the research plan you have just heard about in this presentation.

As stressed by Pietro, MODE a spontaneous initiative of physicists who have realized that it is possible today, although maddeningly complicated, to use AI tools for the end-to-end optimization of the scientific instruments we use for fundamental physics research. This has led us to join forces with experts from the computer science and mathematics community, and draft a long-term plan that can only bring dividends in the time scale of ten years or more.

As this audience includes many of the decision-makers in our community, I take this opportunity to point out that Jenas (and IRIS-HEP in the US) are our only sources of funding at the moment. We wish the community would support this direction of research more consistently, because as Pietro pointed out there are then huge potential gains and innovations that may become possible. AI is disruptive, but we need to build our own interfaces to make it work, and this is very hard!

The difficulties we are facing are due to some inertia in our field, and maybe bad marketing on our side. So I have a chance here to stress that when we say that we want to create AI-driven software instruments that allow the optimization of the design of detectors, we are not planning to send to retirement our detector builders! We mean to put them in the center of a decision process that is better informed and assisted by machines, to the benefit of our communities."

Then, on the following day I made this comment after Ian Shipsey's talk:

"Thank you so much, Ian, for this insightful view of future detector developments awaiting us. I would like to make a comment connected to the detailed planning of improvements on the various devices for the applications that are on our wish list.

While the table you showed at some point of your talk includes desirable goals - anybody would concur that it is desirable to "improve resolution for a gas device", e.g.- it misses, I believe, what I have been calling as of late the "elephant in the room". That elephant is artificial intelligence. We are on the steep slope of rapid developments in our capability of extracting inference from the large and complex data that our detectors yield. Since we are looking forward to 20, 30 years from now, it is absolutely clear to me that we cannot omit to consider what improved methods for the extraction of information will be available at commissioning time.

This is to say that we need to insert in the design procedures of our instruments the improved capabilities of future pattern recognition. In 20 years we will not be reconstructing charged tracks with a Kalman filter, which is a performant tool by today's standards, but rather, with AI-powered instruments. If we do not account for those improved capabilities, we will end up constructing instruments that are not aligned with them; that misalignment, I guarantee it, will produce sub-optimal results.
The goal above -a realignment of our design process to future capabilities- is among the pointed questions that MODE has started to attack. But we are a very small group, and so I feel the need to raise the awareness of this community on the fact that that those challenges cannot be neglected when defining a plan of R&D studies for future instruments."

All in all, I think the messages we broadcast with Pietro's talk and my attempts were well received by the JENAS community, which includes astroparticle physicists, nuclear and neutrino physicists, and accelerator-based research scientists. But I am well aware of how hard it will be to have an impact with the research plan of MODE: there is, in fact, a good reason why we are the only ones who have taken on the challenge of end-to-end detector optimization: it is a maddeningly difficult task!