Simon @spoltier@qoto.org

Over on Whatever today, I talk about how the translation of fiction from one language to another isn't just a matter of exact words, and why I now have it as a contract point that the translations of my novels have to be done by humans, not by machine or "AI":

https://whatever.scalzi.com/2025/02/03/how-translation-works-book-title-edition/

Fascinating Hacker News comment from Tom Gally, a professional translator (Japanese to English) who uses LLMs as part of his workflow, which he describes in detail: https://news.ycombinator.com/item?id=42897856

Wrote a bit more about this on my blog here: https://simonwillison.net/2025/Feb/2/workflow-for-translation/

The release of DeepSeek R1 really shook things up last week. Today, @ahl and I will be joined by Andy Hock and James Wang from wafer-scale innovator Cerebras to talk about their experiences with DeepSeek -- and the future of AI disruption. Join us, 5p Pacific!

https://discord.gg/QrcKGTTPrF?event=1335772904205455391

I'm back in the Bay Area for this great workshop at UC Berkeley – if you're in the area and interested in LLMs & Cog Sci, come along!

https://simons.berkeley.edu/workshops/llms-cognitive-science-linguistics-neuroscience

cony hawk never misses...

#skateboard #skatepark #skating #trafficcone #animatedgif

Published some notes on Dario Amodei's new essay on DeepSeek, mainly to highlight some new-to-me details he included about Claude 3.5 Sonnet

https://simonwillison.net/2025/Jan/29/on-deepseek-and-export-controls/

High touch, high dose, small group programs consistently show the absolute biggest learning gains in efficacy research and this is the model of so many programs in science that are now being cut for "diversity"

After years of the crypto industry claiming to fight for liberty, privacy, and free and open information and tech, I'm very excited for our new "pro-crypto Congress" to get to work protecting access to medical care (including reproductive and gender-affirming care), data privacy, press freedom, freedom of expression, net neutrality, and the right to private and secure communications.

#USpolitics #USpol #crypto

My New Year's resolution was to read the local news (the Solothurner Zeitung). While I am *technically* moderately fluent in German, this is making me even more impressed by everyone who regularly reads complex documents in a language that isn't their first.

Like, you people can just... do this? Without being constantly tempted to reach for the little Translate button?

Have we forgotten already that Panama was *actually* invaded just a couple decades ago by a perfectly respectable Republican president? The US has never given a fuck about international law.

I tried the same medium-complexity Python rewriting prompt against both R1 and o1 and I think R1 did a better job of it - added some notes about that to my post https://simonwillison.net/2025/Jan/27/llamacpp-pr/

Pebble Watches are coming back! https://ericmigi.com/blog/why-were-bringing-pebble-back

The ongoing process of "enshittification" in corporations and other institutions can also be viewed through this lens. Companies (and other complex organizations) have gained access to increasingly sophisticated metrics to measure their user activity, and now have the technology to optimize those metrics for their own benefit (in the near-term, at least). Initially, this sort of optimization can add value for both the provider and consumer of the service; but the majority of their customer base are not playing the optimization "game", and many would have initially signed up for an experience that was more qualitative and less optimized, and containing more of the intangible benefits that even the most advanced metrics fail to capture.

So, are metrics good or bad for individuals and communities? It is a very complex question. In some areas, we are under-using metrics; in others, we are over-using them; and in yet other areas, they are being deployed in too imbalanced a fashion to be beneficial in the long term. Perhaps what we need is more of a discussion of meta-goals and meta-metrics: not just how to pursue a goal or to optimize a metric, but to try to analyze (both qualitatively and quantitatively) how similar or distinct these two objectives actually are, and to locate a stable balance between them that can be broadly accepted within the community in question. (5/5)

In our latest episode, Ashley shares the story of the program she's been running for four years -- a program made possible by federal "diversity in science" funding. I'm going to be promoting this one more than usual because it's so personal and important to us, and I hope you'll give it a share and a listen.

https://www.changetechnically.fyi/2396236/episodes/16501906-the-nih-pays-off-beyond-our-dreams

do you work at a company you like that is actively hiring Software Engineers? do you know someone who does?

consider: referring me to them / them to me! my website homepage now has a "Hire Me!!" section with info about me, what I'm looking for, and how to contact me!

https://esoterra.dev/