Posts by 22@octodon.social
 (DIR) Post #ATiAm8c9lmDJz9MerA by 22@octodon.social
       2023-03-17T15:59:51Z
       
       0 likes, 0 repeats
       
       @simon I do like @deafferret’s point because it seems like it’s probably only a matter of time till a profit-focused Apple product manager angling for promotion pitches, “sure reselling customer data is much less profitable than selling phones, but let’s do it” and the bean counters ok it. I agree not an issue right now but it’s one of those things that in the fullness of time will happen suddenly overnight (a gray swan)?@monkeyninja
       
 (DIR) Post #ATiBAr95r4qd2wYatE by 22@octodon.social
       2023-03-17T16:02:18Z
       
       0 likes, 0 repeats
       
       @simon racking my brain to remember—did I have the phone in airplane mode to test out the filters and did I delete the app immediately afterwards? Or gulp did I forget and just gave the world’s biggest data vacuum cleaner high-res scans of my face?? 🤦 I couldn’t have been that silly could I 😝 @monkeyninja
       
 (DIR) Post #ATiCy1FRBR1oWQWycy by 22@octodon.social
       2023-03-17T16:23:52Z
       
       0 likes, 0 repeats
       
       @simon dang I love this, I was just thinking idly of how one could possibly connect foreign language learning dictionaries and grammar textbooks into LLMs and this is a promising mechanism. @wim_v12e may be interested too (if not apologies!).(I hesitate to boost since your humorous “nightmare scenario” will be taken literally by folks shook by all this…)
       
 (DIR) Post #ATiFpyKuQQ1L2MwdQ8 by 22@octodon.social
       2023-03-17T16:56:31Z
       
       0 likes, 0 repeats
       
       @simon @monkeyninja could someone comment on the failure mode where, even if the majority of data brokers that Uber and Google and Grindr contract with respected US/EU privacy laws, that it just takes one data broker to sell all our info to [insert government/criminal agency] breaking the law?I mean, it seems painfully obvious that state security agencies currently enjoy total access to all the data YouTube app for example collects about us, and so TikTok is just more of the same?
       
 (DIR) Post #ATlGPInMvx1i6Dz0UK by 22@octodon.social
       2023-03-19T03:47:06Z
       
       0 likes, 0 repeats
       
       @simon 9. No, this American does not. Isn’t a squib something that stunt actors use to simulate blood? 😅
       
 (DIR) Post #ATvTcGJoh8rvASGm6C by 22@octodon.social
       2023-03-24T00:34:34Z
       
       0 likes, 0 repeats
       
       Yes, I’ve missed Yegge’s full-on rant mode!“One of the craziest damned things I hear devs say about LLM-based coding help is that they can’t “trust” the code that it writes, because it “might have bugs in it”.Ah me, these crazy crazy devs.Can you trust code you yeeted over from Stack Overflow? NO!Can you trust code you copied from somewhere else in your code base? NO!Can you trust code you just now wrote carefully by hand, yourself? NOOOO!All you crazy MFs are completely overlooking the fact that software engineering exists as a discipline because you cannot EVER under any circumstances TRUST CODE. That’s why we have reviewers. And linters. And debuggers. And unit tests. And integration tests. And staging environments. And runbooks. And all of goddamned Operational Excellence. And security checkers, and compliance scanners, and on, and on and on!So the next one of you to complain that “you can’t trust LLM code” gets a little badge that says “Welcome to engineering motherfucker”. You’ve finally learned the secret of the trade: Don’t. Trust. Anything!“ https://about.sourcegraph.com/blog/cheating-is-all-you-need
       
 (DIR) Post #ATvTcH7RicZFeNUPSK by 22@octodon.social
       2023-03-24T00:38:48Z
       
       0 likes, 1 repeats
       
       “A raw LLM is like a Harvard CS grad who knows a lot about coding and took a magic mushroom about 4 hours ago, so it’s mostly worn off, but not totally.”Ladies and gentlemen. The inimitable Steve Yegge 😂 https://about.sourcegraph.com/blog/cheating-is-all-you-need
       
 (DIR) Post #AU3Ah6QvOdQ7CqJy4m by 22@octodon.social
       2023-03-27T19:07:57Z
       
       0 likes, 0 repeats
       
       @simon reminds me of something Moxie Marlinspike (the Signal guy) tooted a few years back and that’s stuck with me:Many trends in modern programming language design seem to focus on developers pressing fewer keys on the keyboard. To me, that's a strange priority.For large systems where the industry spends most of its time, I think "readability" is much more important than "writability."For example, even simple features like "type inference" feel like misplaced priorities to me.People say "it's annoying I have to write String foo = new String()," but realistically, you're more often writing "String foo = bar.getBaz()"If that becomes "val foo = bar.getBaz()"...what is "foo?""The compiler can figure it out!" they say. But what I care about is whether someone looking at the code can figure it out.We're writing 3 fewer characters one time, at the cost of less information for the ~years people will have to read and understand it.Languages designed for "writability" have function signatures that require you to press very few keys on the keyboard:"def foo(bar):”It's nice to press so few keys to write it, but error prone for the thousands of times people have to read/modify it.Duck typing, no checked exceptions, type inference, last-statement-returns, etc etc require us to press fewer keys on the keyboard, but that is not the problem I see most organizations dealing with. Or if it is, I think giving everyone a copy of Mavis Beacon would do less damage.https://nitter.net/moxie/status/1256640033383051266
       
 (DIR) Post #AU7bj41MoJmvhNMkRU by 22@octodon.social
       2023-03-29T22:29:43Z
       
       0 likes, 0 repeats
       
       @simon what do we think of this notion that, reading the manual (even long, annoying manuals like Lodash or SQLite or MDN or Python standard library) has value in helping hone your mental model on the range of problems a library can solve and its border (tasks it can’t solve)?I know I will absolutely never memorize jq syntax or ImageMagick flags, so I don’t at all mind asking ChatGPT for “the answer”, but I do think if I didn’t know OAuth say, or regexp, there’d be value in investing sufficient time to build a mental model of how exactly that works.
       
 (DIR) Post #AUASOk00m1uGvRIQIy by 22@octodon.social
       2023-03-31T07:11:42Z
       
       0 likes, 1 repeats
       
       Unexpected bonding over #LaTeX typesetting: with the bond trader sitting next to me. You never know when someone’s a closet LaTeX writer.
       
 (DIR) Post #AUNuMYCuRzSd7pRcoK by 22@octodon.social
       2023-04-04T21:40:31Z
       
       0 likes, 0 repeats
       
       Thinking about stable diffusion’s “latent image space” https://arstechnica.com/tech-policy/2023/04/stable-diffusion-copyright-lawsuits-could-be-a-legal-earthquake-for-ai/ and about @simon’s notes about how to get an LLM to avoid hallucinations:“Want them to work with specific facts? Paste those into the language model as part of your original prompt!” — https://simonwillison.net/2023/Apr/2/calculator-for-words/and reflecting how, all the times LLM gives you “the right answer”, it’s because your input prompt metaphorically put it in the latent word space which mapped to output that wound up being correct.That is, every fruitful interaction you’ve had with LLM was maybe a few input words away from a hallucination.I wonder if it’d be at all helpful for an LLM system to show you how slightly different input prompts (leading to slightly different coordinates in the word-embedded space) lead to different outputs, as a way to build your mental model of how to use the machine? If you can see some right and some wrong answers in the set of generated output, maybe you can more quickly learn what makes a prompt good?
       
 (DIR) Post #AUNuMZ8gz9gK12TlsO by 22@octodon.social
       2023-04-06T19:05:53Z
       
       0 likes, 0 repeats
       
       I also think it might have been @simon who was asked in an interview, “are humans large language models?” Can’t find the link now but I thought that was maybe a jokey question till I noticed how often I hallucinate wrong answers.I was reviewing my coworker’s code and was looking at a line that was supposed to remove duplicates (array.filter((x, i, a) => a.indexOf(x) === i) yes it made such an impression I can recreate it from scratch) and had an entire comment written up with thought experiments showing that it may be wrong, till I remembered I could hit a single button and start typing and test things myself—and of course the code worked and I had hallucinated an objection.There’s also the examples of me “checking” kids’ math homework and convincing myself that a sum of two large numbers was something it totally wasn’t.Remembering these foibles makes me take more seriously the question “am I an LLM?” because, reader, I know I bullshit and hallucinate a lot, but just, I often have mechanical aids (like Dev Consoles and calculators) that quickly debunk my mental illusions, so I forget the mountains of bullshit I dream up because they get instantly vaporized by easy reality checks. This probably isn’t a reason to be more forgiving of LLMs because even if bullshit generation is a big part of what’s going on between my ears, computers can do mechanical verification a lot better. I think there’s room for significant improvement here as LLMs combine text generation with “reality checks”.
       
 (DIR) Post #AUTIFsMrqDL7pJH35s by 22@octodon.social
       2023-04-09T05:08:13Z
       
       0 likes, 0 repeats
       
       @mekkaokereke Is the specific feature/bug we're talking about here that a smaller group like Black Twitter can focus broader cultural attention on a specific phenomenon (Pete Davidson in this case)?
       
 (DIR) Post #AUguJsg8QCoUAprt1U by 22@octodon.social
       2023-04-15T23:04:36Z
       
       0 likes, 0 repeats
       
       @firstadopter it might be helpful to read https://simonwillison.net/2023/Apr/12/code-interpreter/  via @simon who said of it:“If you're a programmer and you're still thinking that all of this ChatGPT stuff is a waste of your time, I strongly suggest reviewing this example. It's over-hyped, sure - but it's not something anyone in our profession should continue to ignore” https://nitter.net/simonw/status/1645981621244534784Everyone bullish about LLMs all have their scars from how it completely failed them (fond memories of spending 15 minutes looking for nonexistent books on Qing dynasty highway systems, or the nonexistent SQLite just-in-time query compiler 😂) but have seen plenty of genuinely useful synthesis, hence the hype.
       
 (DIR) Post #AUpjowR4oSKvlinF0i by 22@octodon.social
       2023-04-20T05:26:08Z
       
       0 likes, 0 repeats
       
       @simon is this approach of making a custom backup table for each table better in general (or maybe specifically for app like Datasette) than JSON-based changes tracking per https://www.budgetwithbuckets.com/es/blog/2018/08/27/sqlite-changelog.html/ ? Or are there cases where Matt’s approach is better?I ask because I’m definitely in the category of people who use backup strategies and who don’t roll their own 😅 so having a Python library that’ll automatically add the relevant tables and let me navigate the history sounds like one of those incredible “FINALLY” moments, and want to know a bit more about the space of trade offs here and where sqlite-history sits in that space.
       
 (DIR) Post #AVBjOVzAnzwNgSG6wS by 22@octodon.social
       2023-04-27T03:09:35Z
       
       1 likes, 0 repeats
       
       ‘A century ago, we forced teenagers into compulsory high school to prevent them from being able to fraternize with older adults because we were afraid that 16yos would compete with adults for jobs as the Great Depression was unfolding. Fifty years ago, moral panics around comic books normalized a world in which we restricted children’s access to content. Can we admit that much of this content was political in nature and those who restricted it opposed those politics? Now we’re back to book banning and “don’t say gay” frames. This is not about children’s mental health. This is about preventing children from being active members of our contemporary political polis. This is about using rhetoric around children’s “innocence” to ensure that they don’t encounter views that politicians don’t want them to have. This isn’t new. This is as old of a strategy as it gets.I care deeply about children’s mental health. And there’s a lot that can and should be done. Let’s start with giving every child access to mental healthcare. Let’s make talking to a counselor free. Let’s ensure that children can talk to a trained therapist without being surveilled by their parents (or even needing parental permission).I am deeeeeeply worried about social and structural conditions that increase mental health crises. Let’s eradicate food scarcity. Let’s make it possible for parents to stay home with newborns and sick children without being docked pay or losing their jobs. Let’s build a social safety net.I also fully know how frustrating it is to see your own child struggling and escaping into a zombie state in front of a screen. But parents, please take a deep breath and look at the situation more holistically. Why is this giving them pleasure? What are they escaping? What social itch are they scratching? And are you able to create other paths for pleasure, escape, and socialization?’ —@zephoria in https://zephoria.medium.com/protect-elders-ban-television-2b18ab49988b
       
 (DIR) Post #AVDPYNseBaGOfsYSYa by 22@octodon.social
       2023-05-01T15:33:35Z
       
       0 likes, 0 repeats
       
       @simon @jeffjarvis this is the message of the classic xkcd 1425: “In CS, it can be hard to explain the difference between the easy and the virtually impossible”As a developer who works daily with very highly skilled and technical people (financial asset traders), the opacity for them of what I do continually surprises me. The gentle “I’m not sure how long this will take but…” is just as often followed by a request that’s a one-liner change or a request that needs three teams to coordinate work on.
       
 (DIR) Post #AW5CWAPopeFOHRFlNQ by 22@octodon.social
       2023-05-27T14:22:01Z
       
       0 likes, 0 repeats
       
       @simon so lucky! We’ll be heading out soon to watch for whales! My favorite fellow earth-beings, other than elephants 😁 thanks for posting this!
       
 (DIR) Post #AWiWjjm8wVK1iJnGca by 22@octodon.social
       2023-06-15T12:38:09Z
       
       0 likes, 1 repeats
       
       @wingo can I ask, I think I understand that this is about the garbage collector proposal for WebAssembly https://github.com/WebAssembly/gc/blob/main/proposals/gc/Overview.md but I’m surprised to hear that JVM folks are stakeholders! Is their hope that with wasm/gc JVM (or at least Java and Kotlin) code can run in browsers/Deno/etc.?
       
 (DIR) Post #AWjLSQwmmvBw3fHITA by 22@octodon.social
       2023-06-15T23:01:01Z
       
       0 likes, 0 repeats
       
       @simon I use Cloudflare and I am vaguely confident it matches most of these desiderata?