Today in Biology — Sep 9

[SOFIA] Welcome back to The Dish! We're kicking off today with a look at some big shifts in how science is getting funded and what kind of science is getting pushed, especially in the US. And it feels like the big headline right now is AI.

[DANIEL] Absolutely. AI isn't just a tool anymore; it's becoming a central pillar for major national initiatives.

[SOFIA] Exactly! The Department of Energy just announced the first phase of its Genesis Mission, and it’s a big one: $11.5 million specifically for "AI-for-Science" projects. We're seeing Lawrence Livermore National Lab leading ten of these, and the University of Washington is heavily involved too. What's Genesis trying to do here, Daniel?

[DANIEL] Well, the stated goal for this phase is to use AI to meet growing electricity demand faster and at lower costs. But it's really part of a broader push to accelerate scientific discovery across the board by integrating advanced AI and machine learning into research workflows. It's about moving from hypothesis-driven, experiment-by-experiment science to something more predictive and data-intensive.

[SOFIA] So, for us in engineering biology, this means potentially faster design cycles for new enzymes, new pathways, or even entire organisms? Moving beyond just *in silico* predictions to actually guiding the experimental work in a way that’s much more efficient? Because we often hit bottlenecks when we have too many variables to test.

[DANIEL] Precisely. If AI can more effectively model complex biological systems, or even predict material properties relevant to bio-fabrication, it could drastically cut down on the experimental trial-and-error. The challenge, of course, will be validating those AI predictions with robust experimental data. We’ve seen a lot of hype around AI, and the proof will be in whether these models can actually accelerate tangible, reproducible discoveries.

[SOFIA] And speaking of big shifts, ARPA-H, the Advanced Research Projects Agency for Health, is making some interesting moves. They’ve just launched what they're calling the world's first bid to build FDA-authorized clinical AI for cardiovascular care. That feels like a direct response to the need for AI tools that aren't just novel, but actually *clinically deployable*.

[DANIEL] It’s a critical distinction. Moving AI from a research curiosity to a regulated medical device requires an entirely different level of rigor. FDA authorization means robust validation, clear performance metrics, and an understanding of potential biases in the training data. For cardiovascular care, the stakes are incredibly high. The question will be how they structure the development and testing to meet those regulatory hurdles, which are often much higher than what we see in basic science.

[SOFIA] And just yesterday, ARPA-H also put out a call for early-career investigators, which is fantastic. It’s always good to see funding opportunities specifically targeting new voices and potentially disruptive ideas from folks who are just starting their labs.

[DANIEL] Yes, supporting early-career researchers is vital for long-term innovation. Often, the most groundbreaking ideas come from those who aren't yet locked into established paradigms.

[SOFIA] Okay, but now for something that raised a few eyebrows: news started leaking last week that the NIH is apparently reaching an agreement to use part of its budget to pay for Department of Defense research. That feels like a pretty significant budgetary crossover.

[DANIEL] It's certainly a notable shift. The NIH's primary mission has always been health research, often investigator-initiated and peer-reviewed for scientific merit. When funds start flowing to DoD research, it raises questions about how those projects will be prioritized, what kind of oversight they'll receive, and whether it aligns with NIH's core public health mandate. The details of the agreement will really matter here.

[SOFIA] Right, it’s about the *why* and the *how*. Is this for dual-use technologies, or something more direct? We'll definitely be keeping an eye on that. And just a quick check on what’s buzzing online: there’s a lot of chatter about AI and science, particularly from accounts like *Nature* and *Science Magazine*. Everything from AI cracking fluid dynamics, to models forecasting human genome alterations, and even Fermat's Last Theorem being computer-verified.

[DANIEL] Yes, the AI hype cycle continues in full swing. While these are exciting applications, it’s important to remember that demonstrating capabilities in a specific domain, even a complex one, isn't the same as solving all the underlying scientific challenges. The fluid dynamics claim, for example, would need to show significant predictive power beyond existing computational fluid dynamics models, and with clear, testable hypotheses. It's easy to get caught up in the 'AI solved X' narrative without digging into the specifics of what 'solved' actually means.

[SOFIA] Totally. We’ll be watching to see which of these AI breakthroughs translate into truly robust, broadly applicable tools for biologists. That’s all for today’s briefing! We'll catch you next time on The Dish.