2
points
resfirestar
Joined 1,984 karma
- resfirestar parentTo be fair big tech did do a full court press to stop site blocking when such a law (SOPA/PIPA) was proposed in the US, and they continue to oppose the MPA's attempts to get site blocking via the courts. DMCA on the other hand seems very broken, don't give the MPA the "3 strikes" regime they want and you get sued into the ground like Cox. I suspect tech CEOs don't complain about this because they don't want the same treatment.
- Yes, fetching arbitrary webpages is its own can of worms. But feels less intractable to me, it's usually easy to disable web search tools by policy without hurting the utility of the tools very much (depends on use case of course).
- If someone can write instructions to download a malicious script into an codebase, hoping an AI agent will read and follow them, they could just as easily write the same wget command directly into a build script or the source itself (probably more effective). In that way it's a very similar threat to the supply chain attacks we're hopefully already familiar with. So it is a serious issue but not necessarily one we don't know how to deal with. The solutions (auditing all third party code, isolating dev environments) just happen to be hard in practice.
- Part of the problem here is all the vendor lock in with the tools. It's a new category so it's to be expected, but currently any company that sells an enterprise cloud platform kind of needs their own AI coding tool suite to be competitive.
- I think this article misses the main reason the white wired Apple EarPods and their imitators are popular: they're great for phone calls. Wireless earbuds all have awful microphones, technical measures to mitigate this seem to be futile against the fact that they're sitting on the sides of your head while your mouth is on the front. Earpods make you sound good on calls while also being cheap and fashionable. That's the appeal for most people, I think, nothing to do with their suitability for music.
- That refers to the sandbox "escape hatch" [1], running a command without a sandbox is a separate approval so you get another prompt even if that command has been pre-approved. Their system prompt [2] is too vague about what kinds of failures the sandbox can cause, in my experience the agent always jumps straight to disabling the sandbox if a command fails. Probably best to disable the escape hatch and deal with failures manually.
[1] https://code.claude.com/docs/en/sandboxing#configure-sandbox...
[2] https://github.com/Piebald-AI/claude-code-system-prompts/blo...
- Thanks for sharing the screenshots! The fullscreen view looks a lot cleaner, will look forward to the updates. For monolingual dictionaries, I'm always using Yomitan's ability to look up words within definitions.
- Manabi Reader (OP's app) is way too "busy" for my personal preference. Opening a book and seeing it covered in highlighting and annotations by default is intimidating. To the extent that progress tracking is fun, I want it to be something that's done passively rather than covering every page of every book with paragraph splitting.
It also does not support Yomitan-style custom dictionaries, which is a shame but I understand why it would be a non-goal. Shiori (the other iOS reader app the post mentions) and Jidoujisho (the Android app winner) both have only partial support. The Yomitan+ttsu stack on desktop is unbeatable for learning by reading in my experience. I hadn't heard of Lumie but will try it out on the blog's recommendation.
(Edit: 2 pages into a book, I am not a fan of Lumi's text rendering compared to ttsu.)
- Oh true, that actually sounds useful. I often just avoid that kind of novel because I'm terrible with names.
- The addition doesn't really bother me because Calibre is already full of features that seem utterly useless, so I trust its author to add new stuff without ruining the parts that are useful to me. Still, does anyone actually use "ChatGPT bolted onto the ebook reader" type features for anything besides cheating on school assignments? Lack of web search tools makes them suboptimal for asking clarifying questions or getting recommendations. Makes some sense on a Kindle where you can't exactly alt-tab to ask ChatGPT directly, not so much in a desktop application.
Not to say that there's no use case (I'd be interested to try a LLM-aided notetaking tool), just that adding a chat box is hardly a feature.
- This is just DRAM hysteria spiraling out to other kinds of hardware, will age like fine milk just like the rest of the "gaming PC market will never be the same" stuff. Nvidia has Amazon, Google, and others trying to compete with them in the data center. No one is seriously trying to beat their gaming chips. Wouldn't make any sense to give it up.
- Gamer social movements always burn bright at first, then die when they demand too much purity to reconcile with the fundamental truths: people want to make games and, when they're good, people want to play them. Trying to stop people from using (even experimenting with!) new tools is doomed, just like the old attempts to boycott games over their business models or their creators' politics/sexuality/whatever.
- True, I'm thinking more on the indie level. And "won't notice" is my hyperbole, of course sharp eyed people who use games and software are always noticing small technical choices even when you forgot them yourself. But it shouldn't dominate the impression of the product unless it's poorly done.
- On the art part at least, there's an analogy to Unity. No one liked low effort asset-flip games, and they gave Unity a bad name because they all showed the logo on startup due to using the cheapest license. But no one noticed or complained when good games were made with Unity, or even when they selectively used off the shelf assets. Similarly, we're going to go through a period where every terrible game is going to be full of assets with the stable-diffusion slop look, but no one will notice the AI-generated textures in good games that have actual standards for art.
On the other hand, I think I'd personally vote "somewhat negative" on quests and dialogue. Many games have too much pointless filler dialogue and unmemorable sidequests. Maybe if you don't care enough to actually write it, it shouldn't be there at all.
- They will need to figure out how to give IT departments controls to limit what data can be sent to the cloud backend. Comet and Atlas have "enterprise" versions but lack anything to control this. Edge has some controls that work with Purview DLP.
- >Some were the sorts of teething issues that one might expect to get better over time, such as trouble integrating AI with existing IT infrastructure. Others were more fundamental. AI tools create new tasks and responsibilities, such as “post-deployment monitoring”, which involves “auditing to make sure [the tool] is still performing at the level of accuracy [that was] on the tin,” as she put it.
This is the kind of process that happens with any new technology. Hinton probably just didn't know because he's never worked outside of academia. A common problem with people commenting about "the future of work", AI-related or otherwise.
- >In the case of LLMs, we have service that are aiming to replace both the browser and the search engine
Most people already experience the internet as an integrated browser+search engine (and often, OS) experience from a single advertising company, Google, and it has been this way for over a decade.
>And the result is: ad-infused, tweaked to align with investor priorities, censored by the current politics of wherever the company is based service machinery that's constantly extracting personal information so it can learn better ways to refocus its priorities.
Exactly.
This is not to say I like this outcome, but how is it not massive hyperbole to invoke apocalyptic sci-fi? I expect we'll plod along much as before: some people fiercely guarding their personal info, some people taking a "privacy is dead anyway" approach, most people seeing personal computers as a means to some particular ends (scrolling social feeds and watching Netflix) that are incompatible with thinking too hard about the privacy and information environment implications.
- I'd push back on drawing a sharp line between "institutional abuse" and "genuine erotic connection among equals". As the essay points out, the MeToo campaign did use call-outs against individuals in service of its goal. Some of those callouts were alleging criminal conduct, but on the other end of the spectrum you had much more dubious stuff, or completely unsubstantiated rumors that some person was "bad". I agree that stopping institutional abuse is a noble goal, but the MeToo practice of naming and shaming personal friends in anonymous spreadsheets is the type of thing that builds the internal panopticon: what if our personal circumstances changed so that there's a power imbalance, or someone misinterpreted them? If you accept that practice on political grounds because it's a useful weapon against the "enemies of liberation" (as the author put it), can you really claim to want people to change their attitudes about sex? It doesn't work nearly as well if we stop seeing sexual behavior as inherently scandalous.
- The main part I object to in this essay is the ideological carveout. The author is seemingly willing to defend the #MeToo movement because it was in the service of a mission "to end a long-standing and long-permitted norm of sexual abuse within institutions", and "cancel culture" (I'm also putting it in quotes as I agree it's a very loaded term) because the backlash to it was helpful to the right and detrimental to the left. If you agree with the reasoning, then, all of the behavior being criticized is okay? In that case I don't see how or why anyone would ever change their behavior. The author's friend who wanted her to apologize to the hairdressers probably has a strong belief that being sexualized at work is a serious problem faced by women. From the right, many Christians strongly believe that criticizing behaviors like premarital sex is part of the social immune system that keeps family and community bonds strong.
I think there's a meaningful difference between being a genuine liberal who wants to change how American society thinks about sex, and being a partisan who wants to use puritan callouts as a cudgel on your enemies while ensuring that your own behavior is never subject to criticism. The essay displays an awareness of the tension, but decisively chooses the partisan path.
- Interesting. I've found Claude Code very useful for answering questions about codebases. I think that kind of functionality is on the whole more useful than static AI-generated documentation, but maybe there's a place for an always-ready and google-able starting point.
It badly needs to split up the pages for different parts of large codebases. The golang/go page is way too long and the table of contents sidebar makes you watch a scrolling animation to expand subsections.
- >What strikes me as odd is the decision to position itself as just another AI‑enabled web browser, picking a fight with big techs and better‑funded startups whose users are less hostile (and sometimes enthusiastic) about adding AI to web browsing.
By this logic why have a web browser at all if it means competing with better-funded rivals? Firefox got started "picking a fight with" Microsoft at the height of its power, the asymmetry didn't stop them then. But Firefox users at the time were a group that was excited for new ideas, not hostile. Now the project spends years blocking useful stuff like installable web apps while the vocal part of the userbase treats every new feature or API as proof that Mozilla is a mere puppet of Google.
- Maybe they're trying it with all sorts of models and we're just hearing about the part that used the Anthropic API.
- It's like writing an essay for a standardized test, as opposed to one for a college course or for a general audience. When taking a test, you only care about the evaluation of a single grader hurrying to get through a pile of essays, so you should usually attempt to structure your essay to match the format of the scoring rubric. Doing this on an essay for a general audience would make it boring, and doing it in your college course might annoy your professor. Hopefully instruction-following evaluations don't look too much like test grading, but this kind of behavior would make some sense if they do.
- The way columns like this work is that the author pretends to have a rational basis for their claims that some feature of the masses' private behavior they don't like is going to destroy society, the trick being that they set the starting point of human civilization as portrayed in their narratives and on their charts at around 1950 (or even later in this example, besides the throwaway Max Weber quote from which the title is derived). This ensures that the audience can comfortably ignore the existence of previous degenerate sorts of trends, which have been sanitized by living generations' nostalgia for those cultural moments. Then, after establishing the existence of a serious threat, they try to sell the audience on their preferred solution. Back when we said "Christian" instead of "pro-social", the demand was often temperance, or sometimes a religious reform that that either seems obvious or blasphemous to modern readers. Today the sorts of people who read columnists have reverence for a political ideology instead of a god, so aside from the largely unchanged admonitions against vice, you get arguments that too much construction regulation hollows out our moral virtues. I don't really mind the admonitions against vice (though I find them uninteresting), but the government passing laws in an attempt to fix our morals has a terrible track record. We ought to stick to cold rationality whenever possible.