Shreeharsh Kelkar @shreeharsh.bsky.social · 06/09/2026Zeke Faux' Number Go Up on stablecoin has a whole chapter on this which is quite good. 010
Shreeharsh Kelkar @shreeharsh.bsky.social · 28/08/2026Follow-up post goes back to Turing's question: what does the success of LLMs tell us about the nature of computational machinery and intelligence? computingandsociety.substack.com/p/what-does-... It's hard to say but it does tell us that programs powered by *industrial machinery* can do a lot!computingandsociety.substack.comWhat does the success of LLMs at natural language tell us about computing machinery and intelligence?It validates the sociologist Harry Collins’ view that language competence does NOT require embodiment. And it suggests it is industrial, not artisanal, machinery that allows language competence. 010
Shreeharsh Kelkar @shreeharsh.bsky.social · 28/08/2026Wrote up two posts over the last two weeks on my Substack (1) Is the Turing Test obsolete? Well, it depends on which version of the test you mean: the one based on second-order social knowledge or the one based on language performance. computingandsociety.substack.com/p/have-llms-...computingandsociety.substack.comHave LLMs made the Turing Test irrelevant?Yes, kind of. But the Turing Test is often misunderstood as a test for natural language rather than imitation. Still, LLMs have certainly upended our notion of machinery and its possibilities. 110
Shreeharsh Kelkar @shreeharsh.bsky.social · 06/08/2026@lastpositivist.bsky.social Without endorsing the Vanderbilt stuff, professors do have to think about reconciling academic freedom with political shitposting. Academic freedom was conceived when reaching beyond students or colleagues involved heavy gatekeeping. Social media changed that drastically. 000
Reposted by Shreeharsh KelkarJeffrey Yost @justcode.bsky.social · 29/07/2026It was an absolute joy & honor to work w Con & such amazing chapter authors on Just Code!! THANKS to T&C's reviewer Ricky Wichum! J.C.'s authors on Bsky: @histoftech.bsky.social @mysdick.bsky.social @shreeharsh.bsky.social @dmulvin.bsky.social @janakis.bsky.social muse.jhu.edu/pub/1/articl...muse.jhu.eduProject MUSE -- Verification required! 2114
Shreeharsh Kelkar @shreeharsh.bsky.social · 20/07/2026Nolan's action scenes in particular have always been hard to follow. Go back to the famous truck crashing scene in The Dark Knight and ask yourself if you can understand who is where and what is happening second to second. youtu.be/R1XskC_7GzIyoutu.beBatman the dark knight: tumbler tunnel scene HDYouTube video by Jeff Wong 011
Reposted by Shreeharsh KelkarLee Vinsel @leev.bsky.social · 11/07/2026Great thread as usual from Costa on where data centers fit in concerns about climate. 131
Shreeharsh Kelkar @shreeharsh.bsky.social · 01/05/2026Hi Lilly. Can you please share a couple of these studies on the lack of effectiveness of AI tutors, especially ones that you think are most compelling? I'd be very interested in reading them. 110
Reposted by Shreeharsh KelkarBorners @borners.bsky.social · 08/04/2026My article with @carto-graph.bsky.social on Japanese private railways is out. Rule of thumb, only privatise railways as much as you privatise parking. worksinprogress.co/issue/why-ja...worksinprogress.coWhy Japan has such good railways - Works in Progress MagazineJapan's railways are the finest in the world. Other countries can copy its formula. 27520
Shreeharsh Kelkar @shreeharsh.bsky.social · 20/02/2026Original link here: www.chronicle.com/article/the-... (paywalled but please use if you have a subscription).chronicle.comThe Solution to AI Cheating Is Within Our GraspWhy aren’t we talking more about watermarking? 000
Shreeharsh Kelkar @shreeharsh.bsky.social · 20/02/2026A few weeks ago, I wrote an op-ed in the @chronicle.com about how professors and universities should pursue AI watermarking if they still want to keep a powerful tool in their learning arsenal: the analytical paper. Link: open.substack.com/pub/computin... 100
Shreeharsh Kelkar @shreeharsh.bsky.social · 30/12/2025This is how factual errors should be acknowledged and corrected. 000
Reposted by Shreeharsh KelkarJeffrey Yost @justcode.bsky.social · 19/12/2025Our special issue is out! "Automation by Design: Politics, Culture & Landscape in an Age of Machines That Learn"! Eds. Colette Perold, me, Gerardo Con Diaz. Authors: David Dunning, Christos Karampatsos, Polyxeni Malisova, Eliza Pertigkiozoglou, Jason Ludwig, Megan Wiessner, & Ranjodh Singh Dhaliwal 0137
Reposted by Shreeharsh KelkarMatt Grossmann @mattgrossmann.bsky.social · 23/11/2025Google has a new tool combining Google Scholar search with LLM chatbot style questions and summaries: scholar.google.com/scholar_labs...scholar.google.com 183
Reposted by Shreeharsh KelkarWill Thomas @williamthomas.bsky.social · 30/08/2025Good to see the New York Times has a lengthy obituary for Margaret Rossiternytimes.comMargaret W. Rossiter, 81, Dies; Wrote Women Scientists Into History 06118
Shreeharsh Kelkar @shreeharsh.bsky.social · 27/08/2025This is an incredible deep dive into data centers and their water consumption. Lots of links that you can look into (though I didn't). The conclusion seems to be that data centers have to be treated like any other industrial center. andymasley.substack.com/p/i-cant-fin...andymasley.substack.comData centers don't raise household water bills at all, anywhereA lot of journalism on AI's water impacts is misleading 020
Shreeharsh Kelkar @shreeharsh.bsky.social · 22/08/2025I'm not sure that negates anything about the actual argument she makes in her piece. If anything, it shows even more why she's right about the importance of persuasion. 000
Shreeharsh Kelkar @shreeharsh.bsky.social · 22/08/2025Sentence of the day! "If I've learned one bitter lesson about this stuff over the years, it's that the best productivity hack in the world is simply liking your job." Great post by @caseynewton.bsky.social with some new tools I hadn't heard of: www.platformer.news/productivity...platformer.newsWhat I learned about productivity this yearWhat I gave up, what I kept, and what's new. PLUS: How I'm using AI 020
Shreeharsh Kelkar @shreeharsh.bsky.social · 19/08/2025Of interest to my fellow BlueSky STS people: @castac.bsky.social @asaskat.bsky.social @labortechresearchnetwork.org @cstms.bsky.social And @asociologist.bsky.social @shobitap.org @lmesseri.bsky.social 020
Shreeharsh Kelkar @shreeharsh.bsky.social · 19/08/2025My recommendation is that policy advocates should make a case for their policies by emphasizing value-tradeoffs. This is a much more honest way of making their case that also ensures that it does not get delegitimized down the line. It is more, dare I say it, … democratic. 100
Shreeharsh Kelkar @shreeharsh.bsky.social · 19/08/2025(2) plenty of policy frameworks can be expert-driven but still “democratic,” such Community Notes on Twitter designed entirely by programmers but still much beloved for its underlying values. Or the NHS's NICE committee that @ezrakleinbot.bsky.social wrote about here. www.vox.com/2020/1/28/21...vox.comIn the UK’s health system, rationing isn’t a dirty wordThe UK has one of the most equitable health care systems in the world. Here’s how. 100
Shreeharsh Kelkar @shreeharsh.bsky.social · 19/08/2025In this new post, I argue that (1) democracy is not about the public versus the experts; rather, it’s a shared culture of remaining open to different voices and causes (but without any guarantees that any cause will win out) 100
Shreeharsh Kelkar @shreeharsh.bsky.social · 19/08/2025Rather, we have experts and regular people on ALL sides of an issue. 100
Shreeharsh Kelkar @shreeharsh.bsky.social · 19/08/2025Advocates for policies and policy frameworks will often argue that their proposal is more “democratic” while the one they oppose is “technocratic.” But conflicts in many industrial societies today are NOT between publics on one side and experts on the other. 100
Shreeharsh Kelkar @shreeharsh.bsky.social · 19/08/2025New post on my Substack about why the dichotomy between "democracy" and "technocracy" is a false one that is premised on a definition of "democracy" as public input as opposed to expert input. computingandsociety.substack.com/p/democracy-.... Thread below.computingandsociety.substack.com"Democracy" is not the opposite of "technocracy"Consulting regular people as opposed to experts does not make something more "democratic." Democracy is an ethos. Advocates should justify policy preferences by describing the value tradeoffs. 163
Shreeharsh Kelkar @shreeharsh.bsky.social · 12/08/2025I forgot to thank you and Con for being such good stewards and editors! So, thank you to you both! 010
Shreeharsh Kelkar @shreeharsh.bsky.social · 12/08/2025Very happy that "Just Code" is now published. It was a great journey from conference to book. I'm very happy to be among the company of all the writers in the anthology. 120
Reposted by Shreeharsh KelkarLuísa Reis Castro @anthrobite.bsky.social · 06/08/2025🦟 Can the Mosquito Bite? 🦟 The Multispecies Transmutation of Wolbachia Mosquitoes as Biotechnologies of Epidemic Control in Rio de Janeiro How does the use of a bacterium in vector control reconfigure biopolitical relations? New article at @estsjournal.bsky.social! #STSsky #AnthroSky 1165
Shreeharsh Kelkar @shreeharsh.bsky.social · 04/08/2025Would be curious to see other people's thoughts! 000
Shreeharsh Kelkar @shreeharsh.bsky.social · 04/08/2025But calling everything "power" just confuses all the distinctions we should be making between institutions and different organizing machineries. 110
Shreeharsh Kelkar @shreeharsh.bsky.social · 04/08/2025What is his contribution? It's that the essence of modernity is that science and scientific experts create what one might call the “organizing machinery” or dispositifs of daily life. 100
Shreeharsh Kelkar @shreeharsh.bsky.social · 04/08/2025The problem with Michel Foucault's work: I wrote a post on my favorite philosopher explaining why his habit of calling his contribution as "power" makes the term "power" essentially meaningless. computingandsociety.substack.com/p/what-was-m...computingandsociety.substack.comWhat was Michel Foucault guilty of? He neutered the word "power" of any analytical heft whatsoever.How to do a Foucaldian analysis: replace "power" with a technical term like "dispositif." Also, you must actually suggest an alternative organizing machinery to replace the dispositif you find unjust. 220
Shreeharsh Kelkar @shreeharsh.bsky.social · 31/07/2025E.g. I wrote a whole piece on *why* we have post-truth and most of its citations just say "we live in a post-truth age (five articles with the word post-truth in title including mine)." It may be the pressure of just having to cite something; it could also be that we produce too many papers. 020
Shreeharsh Kelkar @shreeharsh.bsky.social · 31/07/2025I feel this is true of most citations; they may not be this egregious but most tend to be broad references to a topic rather than actually engaging with specific arguments of the papers. 110
Shreeharsh Kelkar @shreeharsh.bsky.social · 31/07/2025I tried it once. I think it worked overall but I didn't do it again even though I love having everyone share one physical text. One thing I would change if I did it again would be to keep the number of pages on the lower side so that it's not too thick (and this also reduces its cost for students). 010
Shreeharsh Kelkar @shreeharsh.bsky.social · 25/06/2025Thanks to @markfabian.bsky.social for interviewing me. It was really fun to talk about my research. In the interview, I expand on what happened to MOOCs, how they became domesticated in that the wild hopes associated with them (access, big data!) abated but some concrete gains remained. 032
Reposted by Shreeharsh KelkarMark Fabian @markfabian.bsky.social · 23/06/2025New episode of ePODstemology! @shreeharsh.bsky.social from Berkeley is on to discuss EdTech like MOOCs (remember those?), how it has evolved, and what effect it might have on the university of the future! www.buzzsprout.com/1763534/epis...buzzsprout.comWill EdTech change the university? - ePODstemologyDr Shreeharsh Kelkar from UC Berkeley on to discuss massive online open courses or “MOOCs” and other varieties of education technology. Are they destined to displace the traditional university, or are... 101
Shreeharsh Kelkar @shreeharsh.bsky.social · 10/06/2025But hopefully, this also means in the long run the importance of courts will decline and policy negotiations will be resolved through political channels rather than by suing each other. 100
Shreeharsh Kelkar @shreeharsh.bsky.social · 10/06/2025That's a great point! I think the US is unusual among developed countries in the deference given to the courts and how much of a role courts play in adversarial policy negotiations so of course, strangely enough, here they are now as well. 100
Shreeharsh Kelkar @shreeharsh.bsky.social · 01/05/2025Of interest to: @kevincarey1.bsky.social @mmasnick.bsky.social @castac.bsky.social @geomblog.bsky.social @caseynewton.bsky.social @asaskat.bsky.social @lmesseri.bsky.social 010
Shreeharsh Kelkar @shreeharsh.bsky.social · 01/05/2025I would be curious about comments from people who have explored this technology in depth or have thoughts on the logic of collective action and standardization here. 100
Shreeharsh Kelkar @shreeharsh.bsky.social · 01/05/2025I suspect also that watermarking doesn't make the cut because of progressives’ fundamental aversion to a solution that involves policing and enforcement. But policing and enforcement can actually lead to better assignments and more learning! 100
Shreeharsh Kelkar @shreeharsh.bsky.social · 01/05/2025I wonder why there isn’t more discussion of this among instructors within universities where the conversation around ChatGPT often drifts to questions around the evilness of AI companies or about the different schools of pedagogy. 100
Shreeharsh Kelkar @shreeharsh.bsky.social · 01/05/2025This is a pedagogy-agnostic solution that would be laser-focused on the university’s main mission, teaching. It's a win for instructors who want to preserve more traditional assignments, for those who wish to incorporate LLMs into assignments, and also for AI companies. 100
Shreeharsh Kelkar @shreeharsh.bsky.social · 01/05/2025They must organize together and push AI companies to implement watermarking, a technology that makes LLM-produced text have a “fingerprint,” which can then be detected by other software. This is is difficult because ALL LLM companies have to buy in but universities have leverage. 100
Shreeharsh Kelkar @shreeharsh.bsky.social · 01/05/2025I wrote about what universities could do about the problem of students rampantly using ChatGPT (or LLMs or generative AI or whatever we call them now)to do college writing assignments. computingandsociety.substack.com/p/how-do-you.... Thread on the main argument follows.computingandsociety.substack.comHow do you solve a problem like ChatGPT?LLMs have disrupted college teaching and not in a good way. But there is a solution, it just involves technology, enforcement, and sanctions--three things that make progressives deeply uncomfortable. 122
Shreeharsh Kelkar @shreeharsh.bsky.social · 14/03/2025I will miss not hearing his voice in my RSS feeds anymore. RIP, Kevin. 000
Shreeharsh Kelkar @shreeharsh.bsky.social · 14/03/2025Have many links in the piece to some of his pieces that I had bookmarked as useful, especially from a teaching perspective. 100