Daniel Lowd

Dear Reviewers: Don't reject because we made it look easy. Making it look easy is hard work. Don't reject because our solution is simple. Simple solutions often have the most impact. Don't reject because you can imagine additional experiments. There's a page limit.

@mmitchell_ai @yoavgo I could be wrong here, but I don’t think anyone else reads the quoted post as nuanced. They may agree or disagree with the post itself, but either way I don’t see how it’s nuanced. If it was meant that way, it is failing to communicate that to almost everyone.

@dlowd @yoavgo Yeah! I don’t think it’s good for everyone to assert LLMs are great when there are a lot of nuances and shades of grey. (And I reckon there’s a larger context.)

lol she is relentless

@ramonalvaradoq Yeah, I think many cases of both hype and hatred come down to lack of critical thinking plus lack of imagination. The interesting, nuanced takes are when people can imagine new ways of doing things, rather than arguing about if LLMs are human-equivalent and/or limitless magic.

Some of us that know the history and philosophy of computing can acknowledge how impressive LLMs are and still think their uncritical widespread adoption in many contexts, including in our work, is misguided. There can be good and bad reasons for doing or not doing things.

@mmitchell_ai @yoavgo As you know, I’m happy to discuss limitations of ML — I’ve been doing it for 20 years, to varying degrees. And I share many criticisms of big tech, which is why I’m in academia. So I’m ok critiquing tech and tech companies. It’s the particular angle that seems… unproductive.

@mmitchell_ai @yoavgo Do you genuinely think that the quoted post is a valuable addition to the discourse, one that helps people better understand LLMs? (I’m aware that there’s a lot of bad faith discourse out there, but I’m focusing on this piece because you seem to be defending it.)

@mmitchell_ai “AI writing” is such a vague concept to begin with. The goal is to express something and reach someone. Pen vs. typewriter vs. word processor vs. spoken already affect how we express ourselves in language. AI is just one more tool.

This is interesting, but many have pointed out that AI writing may seem on first skim meaningful, but on inspection is much more vacuous. This study shows serious “experimenter’s bias”: it’s (unintentionally) designed to support experimenters’ views. 1/

@mmitchell_ai Anyway, I guess this went off on a tangent mostly unrelated to your thread... But thinking about humanity and computers just sparked some thoughts in me. Be well.

@mmitchell_ai And I guess being tribal and hating other people and forming mobs is also human... but it's the part of being human that I hope we can transcend through civilization, not reinforce.

Unpopular opinion: companies continue to shove AI into everything because from their perspective, it's going better than we'd like to admit. One example is Google's AI overviews. I was one of the people loudly complaining about it in its early days when it was in the news for telling people to put glue on pizza. But the quality has improved gradually yet dramatically, and these days I find it pretty useful. I think our disdain for companies "shoving AI down our throats" is largely a selection effect — when one of these AI integrations is new and experimental, we tend to notice, but over time the kinks get worked out, it becomes a part of our workflow, and we stop noticing it. Reminds me of the classic quip that "AI is whatever doesn't work yet." To be clear, even though many of these integrations are pretty useful, the AI-in-everything trend is problematic. There are second-order effects to worry about, such as losing our skills over time when something we used to do manually gets done by AI. And in the case of AI overviews, it cannibalizes traffic to the sources that enable search engines to be profitable. So if we think these integrations harm us or harm society in the long run, it's not enough to simply yell at companies that we hate these features. They have heard us and have correctly concluded that our behavior speaks louder than our words and that we're not quitting these products over AI features. I do think there are some AI integrations we should resist, but to do so effectively we first have to get past the simplistic idea that most AI integrations are useless and companies don't know what they're doing.

I've decided to release a minimal, free online version of my upcoming "10-202 - Intro to Modern AI" course, starting January 26: modernaicourse.org. As a brief summary, this course introduces students to the elements of modern AI systems: you'll build and train a simple LLM chatbot from scratch in PyTorch, without any pre-built layers or pre-trained models at all. More information about the course, and a link to enroll, is available on the webpage link above. In the free online version, you can watch all lecture videos and submit all the assignments (all of which are autograded in the full course anyway). There won't be quizzes/exams, and there's no credit offered via CMU; this is purely for educational purposes. The online course will begin two weeks after the CMU version, and lectures and assignments will all be released with the same two-week delay after the in-person version. Note that this is a first-time offering for this course, so I certainly expect some hiccups along the way. But I hope the material will prove useful for many people.

@KeyTryer Do you follow Rodney Brooks’ predictions? They’ve been pretty accurate. Although a lot of them are just saying, “It’ll be at least 2 years before technology XYZ is anywhere, and at least 8 years before technology XYZ is everywhere.” rodneybrooks.com/blog/

A lot of kurzweilian biotech predictions have been wrong I think, but things have subtly improved like with mRNA vaccines.

I've always had a custom to read the website futuretimeline.net every new year. It's now been enough years that if you followed them for a while you can see how predictions have missed and changed over time.

@simonw The hardest task I gave Opus was to get a simple script to run every morning at 5am. It turns out that debugging Apple’s launchd rules and constraints is surprisingly tricky! But after several days of trial and error, it looks like it succeeded.

I've been throwing some absurdly difficult tasks at Opus 4.5 too, it's proving quite hard to find its limits I've had it build a from-scratch JavaScript interpreter in Python and a WebAssembly runtime too

@KeyTryer IDE? Optional, vim is usually good enough. Debugger? Optional, printf is usually good enough. LLM? Essential, unless you don't care about efficiency or correctness. [NB: I'm an academic who learned to program in the 1990s, so I don't really know what I'm talking about.]

All the rebukes I've seen online is from coding influencers and purist hipsters, but it isn't much a phenomenon I've observed irl that people don't acknowledge how much these programs are capable of doing code at this point.

idk what to tell people who don't believe this, because all the professional developers I know really don't write much code anymore, or have actively tried to resist these programs.

For a previous birthday, I bought my partner a Klein bottle (kleinbottle.com). This Christmas, I bought her a Klein bottle opener. (Yes, she still responds to my emails.)

@KeyTryer @romlib_ Carol can have a little nihilistic hedonism, as a treat.

@romlib_ You can only pull the "she's grieving and horny" card so much before her behavior becomes stupid.

Carol sucks so much

Me: …is this the part where you say you can’t advise me any more because you don’t want to be responsible for this disaster? Gemini 3: Not at all. In the spirit of culinary science (and because I am an AI and cannot taste the consequences), we are absolutely doing this.

Eventually, our house will reach an equilibrium where the dogs track as much mud back outside as they track inside.

@moyix I had to tell Gemini 3 to never say “game changer” again.

Claude has now claimed it found the "smoking gun" in these log files about half a dozen times. We need better gun control, or at least an anti-smoking campaign for the guns so they don't get lung cancer