Our course runs on open-weight models hosted by a university consortium. That is a choice, and it deserves an argument rather than an assumption.
It also sits inside a live public fight — about energy, about data centres, and about who gets to control the tools that increasingly mediate scientific work. Three readings, with three very different sets of interests behind them. Read them as such.
Assigned¶
1. The open-weights letter.
In July 2026, Nvidia CEO Jensen Huang published an industry coalition letter, “Open Weights and American AI Leadership,” arguing that the US should not restrict publicly downloadable model weights. It launched with 25 signatories and grew past 150 within days. Anthropic did not sign.
The letter (mirrored by Microsoft; PDF at
aka.ms/openletter)
2. A skeptical reading of that letter.
Jensen puts his thumb on the scales against open-weights fearmongering, Tobias Mann, The Register, 27 July 2026. Free to read.
Read these two together, in that order. The letter makes a public-interest case for open models. The Register points out that the person who convened it sells the hardware that every model — open or closed — runs on. Both of those things can be true at once. Working out what you actually believe, given that, is the exercise.
3. The scientific case, which is a different case.
Buhler, Pérez & Boettiger, Why scientists should lead the shift away from AI mega data centres, Nature, August 2026.
If you hit a paywall, Berkeley News covers it openly.
The argument here is not about American competitiveness. It is that open models small enough to run on ordinary hardware can use dramatically less energy than their commercial counterparts, and that scientists have specific reasons — reproducibility, autonomy, cost — to prefer tools they control.
What this is doing in a data science course¶
A fair amount of campus energy right now goes into opposing data centres, and the concern is a real one. It is also frequently attached to an assumption worth examining: that AI means a hyperscale data centre somewhere, and that using these tools at all is participation in that.
That assumption is not correct, and this course is a working counterexample. The models you will use are open-weight, hosted by a research consortium, small enough that several of them would run on your laptop. Nothing you do this semester requires a commercial data centre.
That does not make the concern go away. It sharpens it into better questions:
Efficiency arguments cut both ways. Processors have become orders of magnitude more energy-efficient over decades, and total electricity spent on computing has gone up. Why would more efficient models reduce total energy use rather than expand what people attempt? (This is the same argument you met in the module introduction about why faster coding has never meant less coding.)
If open models are cheap enough to run anywhere, who benefits, and who loses the ability to charge for access?
The letter’s signatories argue openness improves safety through transparency. Anthropic, which did not sign, argues roughly the opposite. What would count as evidence either way?
To discuss in class¶
Come prepared to say which of the three readings you found most persuasive, and — this is the harder half — what the author of that piece stood to gain.