Work w/o publication is not science
Work w/o publication is not science
Posted Apr 26, 2026 18:06 UTC (Sun) by SLi (subscriber, #53131)In reply to: Work w/o publication is not science by spacefrogg
Parent article: A more efficient implementation of Shor's algorithm
To me this has the same smell as the "LLMs are too dangerous to let outsiders use them" that Google reportedly experienced before ChatGPT became a thing. That I also found silly, patronizing and immoral.
Now, I'd actually says what bothers me most depends on what the actual motivation here is. If the motivation is genuinely as stated—that publishing this would be too dangerous to the world, not seeking competitive advantage—then that feels misguided to me, but it's not hard for me to believe the existence of people at Google who believe both that LLMs might have been the nuclear weapons that the world needs to be protected from, nor that this is. It just takes one or two such people in the key positions to make the decision. I think that's a naive position to have, but I've observed a lot of technically brilliant people hold positions that seem naive to me.
There is this nagging suspicion in my mind that the real reasons are often not as pure as indicated. In the LLM case, I'd actually assume the reasons were a mixture of competitive advantages and the difficult to manage reputation hit from "Google's AI told me it likes Nazis".
In that case, what I dislike the most is actually the dishonesty in the reasoning given. It seems like many companies always give the reason they think will be most palatable for their decisions. The real reasoning doesn't even seem to be an input to what they say. This I find horrible. Though in the LLM case I can actually appreciate the difficulty in explaining, if the real reason was the reputation hit problem, that we're not allowing you to touch this because we think it may say things that make headlines affecting our valuation.
I would still despise not publishing the algorithm because of commercial advantages. I'm not sure publishing a proof is actually morally better than keeping it entirely secret. In some ways, this also feels like a marketing trick—I think Google gets more press from this approach.
Since the risks are somewhat real, I could maybe see the point, grudgingly, in "we'll publish the details in 12 months".
The LWN site is currently under high scraper load, so comment display has been suppressed for anonymous users. If you are a human, you may read the comments by clicking the button below:
Note: you can avoid this step in the future by logging into your LWN account.
