It’s rare for a large fraction of grant money to be siphoned to a Dean’s discretionary fund. Typically some smallish fraction (maybe 10%, often significantly less) of indirect costs (which, depending on the funder and the negotiated F&A rate may be anywhere from 0% or nearly 100% on top of the direct costs that fund the research staff and materials, etc.) goes back to the subdivision overseen by that Dean to do with as they please. Everywhere I’ve worked, that amounts to a few (low single digit) percent of total costs being used in the way you describe. And many places return none of indirects to the unit overseeing the PI and so then it’s a cool zero percent.
The similarity you see between this DeepMind project, the DARPA program, and research in Tenenbaum's lab is not incidental: there's a steady stream of crosstalk and cross-training between machine learning researchers who engineering artificial intelligences and cognitive scientists who reverse-engineer human intelligences. (Note, for example, that Peter Battaglia, one of the co-authors of this DeepMind project, was a postdoc with Tenenbaum.)
At many universities (in the U.S. at least), on paper, it’s the technology commercialization department that makes the call about pursuing patent protection and the burden is on the PI to report all potentially patentable inventions before release so that the university has time to make the determination. In the case you describe, they’d in principle find you in violation of that policy, but in practice can’t because they don’t know about it unless you tell them.
This is an excellent example of why inferring causation from an observed correlation requires great care. Note that the study did not randomly assign valedictions to emails and observe the causal impact on response rate. Rather, they observed a correlation between the sign-off chosen and the response rate.
"Thanks" garners many responses because people use it in emails that make reasonable requests with a good chance of response.
My favorite valediction is still "I am, &c.," which is short for "I am your humble and obedient servant". Not sure whether my colleagues appreciate it as much as I do...
The parent comment says that, of all the information-finding tools that hardly anyone uses, library resources are the best, not that library resources are the best information-finding tools; there's no contradiction between what the two of you have noticed.
The word "smarter" masks some of the complexity here. If you define it as "higher performance on a task widely considered to require intelligence", then we've had computers that are smarter than humans for at least decades. If you define it as "higher performance on every task widely considered to require intelligence", then I'll take that bet, please.
The criteria for winning the prize depend more on the outcome of the research (importance) than its process (originality):
"The said interest shall be divided into five equal parts, which shall be apportioned as follows: /- – -/ one part to the person who shall have made the most important discovery within the domain of physiology or medicine …” (Excerpt from the will of Alfred Nobel)".
I've had good experiences with PQ (https://pypi.org/project/pq/). Any event that generates a notification triggers adds an entry to the queue. Worker processes get entries from the queue. The queue is stored as another table in your database whose structure and content is managed by PQ, though you can always read/write to it if you want. PQ handles the concurrency.