Thursday, May 15, 2025
Context: In my role as division director of Information and Intelligent Systems (IIS) at NSF, I’m sending out a short message to the IIS mailing list on the Second Tuesday Every Month (STEM). This is the installment for May 2025.
There have been a lot of changes at NSF since my last message, some of which intersect with some of the earlier rumors I mentioned last month. These have been reported in the press and were also announced internally and I’ll just share a subset of things going on.
Greg Hager, the CISE assistant director (AD), left his role at NSF last week: As a rotator like me, he is able to return to his home institution. We are grateful for the time he spent at NSF and his outstanding contributions in support of the broader research community. Ellen Zegura, the division director for the systems and networking division (CNS) is now the acting AD.
Funding for some existing projects has been terminated: NSF has established new internal processes for how to allocate funding for projects in a way that is consistent with stated administration guidelines. To bring the funding portfolio into compliance, we had to stop funding some existing grants and update the procedure for funding future grants. We are now transitioning from the first of these activities to the second, which should increase the rate at which newly recommended projects can be approved for funding. We still hope to get through our funding backlog before the close of the 2025 budget cycle in early August. I’m happy to say that we approved a bunch of CAREERS this week. Those proposals next move within NSF to the Division of Grants and Agreements for final approval.
NSF's budget may decrease substantially in 2026: Negotiations are underway, but the president's proposal includes a sharp cut in NSF's overall budget. However, the proposed budget also includes guidance to hold funding for AI steady. We don't know what the final decision will be and how NSF will operationalize it. (Indeed, we're still waiting for final numbers for our 2025 budget.)
I want to take a moment to reflect on one implication of a possible budget decrease. NSF will spare CAREER awards as best as possible, but lower numbers of funded projects is a likely outcome. Whether or not this happens, the computing field should seriously consider rethinking how it treats CAREERs. A lot of departments use CAREER awards as a stand-in for something like “is this faculty member worthy of tenure?” If funding rates drop, it’s important to realize that it’s not like the quality of junior faculty is dropping at the same rate. Indeed, from my reading of the job market, the world is producing fantastic scientists across all areas of computing, in larger numbers than ever. Based on informal surveys within NSF, the CISE research community is unique in its heavy use of CAREER awards in tenure decisions. I want to urge the community to be open to broader signals of junior faculty achievement as we move into this (possibly) leaner period of NSF funding. Personally, I’m a fan of the ideas of my colleague Stu Geman, who said (in an article with his brother): “Suppose, instead, every hiring and promotion decision were mainly based on reviewing a small number of publications chosen by the candidate.” I think it's valuable advice even if funding remains steady, so I hope you consider it.
Last month, I shared a puzzle I called "misled" that asked (given some context about my broken old Radio Shack calculator), what will the calculator produce as the answer for 230+354+629+982?
I called the puzzle "misled" because the fact that it looks like addition is misleading. In fact, you should focus on the "LED" part of the calculator display (mis-LED). I received more replies to last month's message than any message I've sent at NSF. Maybe 4 or 5 people sent the intended answer, which is 115. The people sending correct answers included a mathematician, a doctor, and a 9th grader whose dad is friends with a biomedical engineer. About a half dozen computer scientists sent me the "solutions" found by various chatbots. Of these, some were insane ravings and at least two were spot on (!). LLMs continue to surprise and amaze... The people who coaxed chatbots to solve the problem all used insight to refocus the prompt instead of just providing the text I used to describe the puzzle. Researchers continue to surprise and amaze...
The shortest complete explanation I have is saying the solution is the "digitwise segment xor". Basically, you should think of the numbers not as numbers but as the set of LED segments that are lit up if that digit were displayed on a calculator. Then, the "addition" operator is actually the segment-wise exclusive or of these sets. For multiple digit numbers, the xor was performed on each digit position separately.
I constructed the puzzle so that that answer would be 115 or, since the LEDs aren't actually representing numbers... it spells out IIS!
One last thing. Last Saturday marked NSF's 75th anniversary. Although some stimulating and festive celebrations were planned, it was overshadowed, of course, by the large and rapid changes the agency is having to carry out. But, in my own quiet way, I want to thank each and every one of you for being an important part of NSF's distinguished history. You matter and I'm sorry we weren't able to do more to celebrate all of the people who have made NSF so special.
Until next time,
Michael