The previous LOTRW post found me eager to attend the (then-upcoming) American Meteorological Society’s Annual Meeting, which took place from January 28-February 1 in Baltimore – and sold on the importance of conferences and meetings to the progress of science and its application for societal benefit. Closed in this vein: “Bottom line? Meetings are an engine for innovation – not a vacation from it.”
Proved to be an understatement. AMS 2024 meteorologist-attendees got to witness Schumpeter’s gale[1] up close and personal. (More generally known as creative destruction, Schumpeter’s gale is an economic process in which new innovations replace and make obsolete older innovations.) Was blown away.
Let’s dig in – but first, a caveat or two. My experience of the AMS 2024 Annual Meeting was different from prior years – more personal, and more limited.
It was also more Lagrangian than Eulerian. For two decades, it had been my good fortune to work at AMS. In those years, my Meeting-participation meant staffing Council and committee meetings, arranging and participating in policy-related side events, doing a collection of meeting-related jobs that the organization needed done. The Meetings – at least that small sliver of the meetings – came to me, as it were. The experience was Eulerian; a perspective confined to a single “fixed” point, as the Meeting week flowed past.
This time around, at least for a few days, my experience was again confined to a single point, but that point was more like a tagged particle, at times moving relatively freely through the Meeting, at times carried along, following the crowd. I could choose where I wanted to go, and/or respond spontaneously to opportunities as they were presented. My experience was Lagrangian.
Which brings up the second caveat. My 2024 perspective and experiences were therefore greatly limited. This had always been true, but as a staffer I’d always been privy for months prior to the larger aspirations, preparations, and issues driving the conferences, sessions, and side meetings. Not so much this time around. And like a blind man sampling only a single feature of the elephant, my experience most likely differed substantially from that of others (including yours, if you were there).
Okay, you’ve been warned! But this is what I found and felt…
First, the atmosphere. In the Convention Center Sunday afternoon and evening, and through the week, the vibe was lively: 7000+ scientists; engineers; international attendees; domestic and foreign government-, industry-, and academic leaders; sales reps; and 1300 student registrants for the student conference. A bigger (and younger?) crowd than ever. By my subjective estimate somewhere between 25-33% of the participants were early-career (a demographic that bodes well for the meeting, the AMS, the field and the larger world). The ebb and flow of participants, the swarms at the posters and in the exhibit hall? Tremendous vitality, freshness. Promise. Sparkling reports of innovation newly completed and inspiring visions of innovation to come filled the air. You had to love it.
The AMS Annual Meeting always offers a buffet of topics. Here’s my experience at one end of the buffet table. (The other end? I made it there as well. But that account will have to wait until the next post.)
Start with Artificial Intelligence. In prior years while I’d been spending all my time at AMS Annual at the side meetings, I’d always been missing a lot of the science and technology advance. This would be my chance to catch up. And what better place to start than at AI?
This got off to a sobering start. The program informed me this was the 23rd Conference on Artificial Intelligence for Environmental Science.
23rd?
23 years of AI sessions? Where have I been? Reminds me of Thomas Jefferson’s song in the musical Hamilton. While Washington, Hamilton, Madison et al. had been fighting the Revolutionary War, doing the hard work of standing up the young country, Jefferson had been off in Paris, representing US interests (it’s a hard job but someone has to do it). Lin Manuel Miranda captures this nicely. He has Jefferson coming back to New York, all posh and refined, and asking, in rap and song, What’d I Miss?, belatedly concluding that “I guess I basically missed the late 80’s.”
It got worse. My at-the-Conference AI experience itself? A massive fail. Showed up to the several sessions, held in big rooms. Big rooms, but not vast rooms. Meaning – by the time I would get there, slowed by encounters with friends along the way, those meeting-rooms were packed. People standing cheek-by-jowl along the entire four walls. People clustered three deep at the door.
Young people. Realized I wouldn’t have the staying power/stamina to get much out of the sessions. The good-and-bad news? To the good – the sessions were all recorded! Since then, I’ve made a start of viewing them from the comfort of home. The bad news? Given the pace of advance of AI, the two-three-week delay means my understanding is already hopelessly out of date.
This isn’t entirely hype. To get a feel for the new pace of progress, watch Imme Ebert-Uphoff’s core-science keynote entitled A Research Agenda for the Evaluation of AI-Based Weather Forecasting Models. (The link to her abstract is here; the video can be found here.) The Schumpeterian gale she depicts is one in which improving NWP from physical principles and established technologies is plodding along at a (loosely coordinated) few big centers of effort such as ECMWF and EPIC. Meanwhile, multiple, competitive AI-based (NWP-free) prediction efforts are proliferating worldwide. Wholly new initiatives are emerging monthly. The ECMWF-EPIC work is informed and shaped by forecaster feedback; by contrast, the AI-ML work has enjoyed very little forecaster scrutiny. The expertise/disciplinary backgrounds required of the two approaches are largely disjoint; so are the hardware requirements; so are the updates on progress.
Or watch Amy McGovern’s talk from the next day. There’s a fair amount of overlap here. She hints that over the next 3-10 years we can expect to have personalized forecast models at our fingertips. However, she also warns of the high probability that one of the many un-peer-reviewed AI models now on offer will blow a consequential forecast – resulting in fatalities, economic loss and damage to public trust that will cause years to repair. (Picture something analogous to the occasional autonomous-vehicle crash we’re currently experiencing, but on a bigger screen. Creative destruction is creative – but it can still be destructive.)
Speakers made reference to Trust and trustworthy artificial intelligence: A research agenda for AI in the environmental sciences, a multi-authored paper published last year in Risk Analysis, a journal of the Society for Risk Analysis. The authors lay out a convergent approach to review, evaluate, and synthesize research on the trust and trustworthiness of AI in the environmental sciences and propose a research agenda. The agenda makes good sense; its authors read like a Who’s Who of the field. But came away feeling that the needed evaluation will be inadequately funded, resourced – and therefore all too likely to lag the progress it’s intended to evaluate. Likely too little, too late to keep pace with the “move-fast-and-break-things” culture.
As stated, Schumpeter and others see creative destruction as an ongoing process. That’s true in meteorology; even here AI’s creative disruption has a prequel. Half a century ago, AMS Annual Meetings were sleepy affairs that brought together a few hundred people at most. The exhibits area – located in a hotel lobby or even a hallway, featured card-table-top displays, with retired NWS employees either offering individual consulting services or selling Belfort anemometers and other meteorological instruments. Then-AMS-Executive-Director Ken Spengler and his assistant Evelyn Mazur would look a few months in advance for meeting venues at small hotels that offered last-minute bargain rates. Then, in the 1980’s, Dick Hallgren, while Director of the National Weather Service, formulated and led NWS through what was known as the 10th Modernization and Restructuring. The main features included an upgrade to the weather radars, the introduction of the Automated Weather Information Processing System or AWIPS, and elimination of myriad small NWS Service Offices in favor of a smaller number of larger, more-capable NWS Field Offices, to be collocated with the new radars. The AMS initiated a new conference at the Annual Meeting on information processing (now known as the Conference on Environmental Information Processing Technology and in its 40th year). Participation in this one conference quickly dwarfed all the other annual conferences and symposia comprising the Annual Meeting. When Hallgren took the AMS reins as Executive Director in the late 1980’s he had the vision to see the possible synergies between the technology advance, increasing private-sector roles, and international interest. The big government satellite, radar, and IT buys grew the exhibits into splashy displays of today. Hallgren invited WMO officials and the heads of national meteorological and hydrological services worldwide to collocated annual workshops. These visitors would shop the trade exhibits. The crowds attracted U.S. government officials, business- and academic leaders to the conversation. The connectivity/critical mass allowed the field’s leaders – not just the bench scientists – to accomplish in days what would require weeks back at the office. AMS Annual Meetings doubled in size and then doubled again (requiring, inter alia, that the meeting planners book larger hotel- and conference venues a decade in advance).
By analogy, it’s tempting to forecast that future AMS Annual Meetings may soon be double the size of this year’s[2]. And that doubling? That’s at most only half the story. Note that the influence of AI is not just changing the size of the meetings – and the so-called Weather, Water, and Climate Enterprise. It’s changing its intellectual/disciplinary center of gravity, shifting it from emphasis on geophysical fluid dynamics, chemistry, and radiative transfer to technology. And winds from at least one other Schumpeterian gale are blowing across AMS these days – a growing urgency to apply these advances for societal benefit.
But that is a topic for another day.
[1] Treated Schumpeter’s gale earlier in an LOTRW post from 2015 (slightly different context).
[2]In the face of this trend, some AMS members bemoan what they see as a loss of the intimate feel of the meetings. That may well be, but part of me says this loss is more imagined than real, or at least more under individual control. AMS Annual meetings are a colony of more specialized conferences, symposia, and diverse side meetings. Participants may find themselves spread thin – but that’s a choice. They can retain the small-crowd feel by confining their attention to one or two of the many parallel conversations on offer.
