Previously on this blog: my objectives / adjustment; actual OKR #1, #2, #3, #4, #5, #6, #7, #8, #9, #10, #11, #12, #13.
Another half year, another OKR. And I am absolutely certain that I failed the “Think out loud” objective, because I did not published anything on this blog during the first half of 2026.
I did write, but I could not finish any blog draft. An excuse is that 2026 Q1 and Q2 were not my best time. As written in #12, I had an accident in late 2024. And in Q1 and Q2 of 2026, I received my last leg of the dental treatment. That treatment was not cheap and not covered by the German insurance. I am thankful, that I had surplus to pay for it. If it were like 10 years ago, I probably would not be able to pay for it. My wife’s health wasn’t good as well. But there is no detail I would want to share.
Mentally, I was also not fit. But again, it is also an area I have nothing to share. But if you follow this OKR series, you probably know what really bothers me.
The world really sucks. Sometimes I also think about what is the purpose of still keeping a blog. Should I still think out loud? Who would want to read me think out loud? Or it’s more like for sending it to your friend and your friend replies you with a “two eyeballs” emoji? It appears more likely this blog is for generating human-produced training data for whatever LLMs, rather than for humans. I also think about the same thing when I do open source.
In summary, I am still fine. Nothing more to add.
Nothing
readODS 2.3.5 and rio 1.3.0.
Excluding preprints and conference submissions, I published only one paper in the last six months.
This is a critique of LLMs, which conceptualized in a conference session in 2025 that I did not attend. My role for this opinion piece was more editorial. After reading Cory Doctorow’s latest book The Reverse Centaur’s Guide to Life After AI: How to Think About Artificial Intelligence—Before It’s Too Late recently, I also self-reflected whether this is a good critique. In Doctorow’s book, he borrows the concept of Crit-Hype from Lee Vinsel. If a critique “is parasitic upon and even inflates hype,” it probably is a Crit-Hype and is not a good critique. In the domain of AI critique, if one accepts the premise that whatever AI or LLM is amazing, and then “but”s or “however”s, the piece is probably a Crit-Hype. I feel that we didn’t reproduce the BS from big tech in this piece. However, we still seem to restate the potential of LLMs for communication research, which seems to still fuel the hype. If I would want to change, we need to say it upfront that we should not replace perfectly capable humans with automated tools that cannot do humans’ jobs. I would go more deeply into the fundamentals of automation in social science research. Who would researchers want to “replace” (this is the discourse) using LLMs “at scale” (this is also the discourse), e.g., audience perspectives represented by trained coders in the case of automated content analysis, human opinions in the case of “in-silco” survey. We should also think about it in the logic of capitalistic societies. Researchers use LLMs to “replace” trained coders, usually students who are desperately in need of money to pay for the expensive tertiary education. Thee monies saved are split between the vendors of LLMs (either directly using the services from big tech companies such as OpenAI, Google, Antropic, SpaceX; or buying hardware from Nvidia and running the so-called “open-weight” models) and the researchers. Is this anti-student-workers, pro-big-tech, (and of course, pro-climate-destruction) way of using LLMs any good? ↩