While true, it doesn’t keep you safe from sleeper agent attacks.

These can essentially allow the creator of your model to inject (seamlessly, undetectably until the desired response is triggered) behaviors into a model that will only trigger when given a specific prompt, or when a certain condition is met. (such as a date in time having passed)

https://arxiv.org/pdf/2401.05566

It’s obviously not as likely as a company simply tweaking their models when they feel like it, and it prevents them from changing anything on the fly after the training is complete and the model is distributed, (although I could see a model designed to pull from the internet being given a vulnerability where it queries a specific URL on the company’s servers that can then be updated with any given additional payload) but I personally think we’ll see vulnerabilities like this become evident over time, as I have no doubts it will become a target, especially for nation state actors, to simply slip some faulty data into training datasets or fine-tuning processes that get picked up by many models.

you could just do it yourself.

Personally, I think that wholly depends on the context.

For example, if someone’s having part of their email rewritten because they feel the tone was a bit off, they’re usually doing that because their own attempts to do so weren’t working for them, and they wanted a secondary… not exactly opinion, since it’s a machine obviously, but at least an attempt that’s outside whatever their brain might currently be locked into trying to do.

I know I’ve gotten stuck for way too long wondering why my writing felt so off, only to have someone give me a quick suggestion that cleared it all up, so I can see how this would be helpful, while also not always being something they can easily or quickly do themselves.

Also, there are legitimately just many use cases for applications using LLMs to parse small pieces of data on behalf of an application better than simple regex equations, for instance.

For example, Linkwarden, a popular open source link management software, (on an opt-in basis) uses LLMs to just automatically tag your links based on the contents of the page. When I’m importing thousands of bookmarks for the first time, even though each individual task is short to do, in terms of just looking at the link and assigning the proper tags, and is not something that takes significant mental effort on its own, I don’t want to do that thousands of times if the LLM will get it done much faster with accuracy that’s good enough for my use case.

I can definitely agree with you in a broader sense though, since at this point I’ve seen people write 2 sentence emails and short comments using AI before, using prompts even longer than the output, and that I can 100% agree is entirely pointless.

I definitely agree.

However, I think there’s certainly a point at which the usage of a given tool is too small to meaningfully impact your actual retention of a skill, and I do think that when these people are just, say, occasionally firing off an email and they feel like the tone is a bit off, having it partially rewrite it could possibly even help them then do better in the future at changing their tone on their own, so personally I think it’s a bit of a mixed bag.

But of course, when I look at all the people foregoing things like learning programming languages to ask ChatGPT to just vibe code everything for them, then talk about how they’re gonna get a job in tech… yeah, that’s 100% past the point of skills atrophying in my opinion.

See, when I have 4-6 sentences to summarize, I don’t see the value-add of a machine doing the summarizing for me.

Oh I completely understand, I don’t often see it as useful either. I’m just saying that a lot of people I see using LLMs occasionally are usually just shortening their own replies to things, converting a text based list of steps to a numbered list for readability, or just rewording a concept because the original writer didn’t word it in a way their brain could process well, etc.

Things that don’t necessarily require a huge amount of effort on their part, but still save them a little bit of time, which in my conversations with them, seems to prove valuable to them, even if it’s in a small way.

The problem is that anything even remotely related to AI is just being called “AI,” whether it’s by the average person or marketing people.

So when you go to a company’s website and you see “powered by AI,” they could be talking about LLMs, or an ML model to detect cancer, and the average person won’t know the difference between the technologies.

So if someone universally rejects anything that says it “uses AI” just because what’s usually called “AI” is just badly implemented LLMs that make the experience worse, they’re going to inevitably catch nearly every ML model in the crossfire too, since most companies are calling their ML use cases “AI powered,” and that means rejecting companies that develop models like those that detect tumors, predict protein folding patterns, identify anomalies in other health characteristics, optimize traffic routes in cities, etc, even if those use cases aren’t even related to LLMs and all the flaws they often bring.

It can, but I don’t see that happen often in most places I see it used, at least by the average person, although I will say I’ve deliberately insulated myself a bit from the very AI bro type of people who use it regularly throughout their day, and mostly interact with people who are using it occasionally during research for an assignment, rewriting part of their email, etc, so I recognize that my opinion here might just be influenced by the type of uses I personally see it used for.

In my experience, when it’s used to summarize, say, 4-6 sentences of text, in a general-audience readable text (i.e. not a research paper in a journal) that doesn’t explicitly rely on a high level of context from the rest of the text (e.g. a news article relies on information it doesn’t currently have, so a paragraph out of context would be bad, vs instructions on how to use a tool, which are general knowledge) then it seems to do pretty well, especially within the confines of an existing conversation about the topic where the intent and context has been established already.

For example, a couple months back, I was having a hard time understanding subnetting, but I decided to give it a shot, and by giving it a bit of context on what was tripping me up, it was successfully able to reword and re-explain the topic in such a way that I was able to better understand it, and could then continue researching it.

Broad topic that’s definitely in the training data + doesn’t rely on lots of extra context for the specific example = reasonably good output.

But again, I also don’t frequently interact with the kind of people that like having AI in everything, and am mostly just around very casual users that don’t use it for anything very high stakes or complex, and I’m quite sure that anything more than extremely simple summaries of basic information or very well-known topics would probably have a lot of hallucinations.

This is precisely why I don’t think anybody should be using it for meeting summaries. I know someone who does at his job, and even he only uses it for the boring, never acted upon meetings that everyone thinks is unnecessary but the managers think should be done anyways, because it just doesn’t work well enough to justify use on anything even remotely important.

Even just from a purely technical standpoint, the context windows of LLMs are so small relative to the scale of meetings, that they will almost never be able to summarize it in its entirety without repeating points, over-explaining some topics and under-explaining others because it doesn’t have enough external context to judge importance, etc.

But if you give it a single small paragraph from an article, it will probably summarize that small piece of information relatively well, and if you give it something already formatted like bullet points, it can usually combine points without losing much context, because it’s inherently summarizing a small, contextually isolated piece of information.

that’s a different beast

I think what was being implied though was that the original poster was saying that any use or talk of AI by a company immediately invalidates it, regardless of there being any specific traits like firing workers present. (e.g. “Using AI” was the only prerequisite they mentioned)

So it seems like, based on the original wording, if they saw a hospital going “we use top of the line AI to identify tumors and respiratory diseases early” they would just disregard that hospital entirely, without actually caring how the AI works, is implemented, or affects the employment of the other people working there, even though it’s wholly beneficial.

At least, that’s just my reading of it though.

I don’t think AI is actually that good at summarizing.

It really depends on the type and size of text you want it to summarize.

For instance, it’ll only give you a very, very simplistic overview of a large research paper that uses technical terms, but if you want to to compress down a bullet point list, or take one paragraph and turn it into some bullet points, it’ll usually do that without any issues.

Edit: I truly don’t understand why I’m getting downvoted for this. LLMs are actually relatively good at summarizing small, low-context-necessary pieces of information into bullet points. They’re quite literally made as code that interprets the likelihood of text based on an input. Giving it a small amount of text to rewrite or recontextualize is one of its best strengths. That’s why it was originally mostly implemented as a tool to reword small isolated sections in articles, emails, and papers, before the technology was improved.

It’s when they get to larger pieces of information, like meetings, books, wikipedia articles, etc, that they begin to break down, due to the nature of the technology itself. (context windows, lack of external resources that humans are able to integrate into their writing, but LLMs can’t fully incorporate on the same level)

This has been my reaction for a while now. And usually, I feel like it does tend to accurately represent the thought put into a product.

When a company barely thinks about their marketing material, (the thing they often require to even make their thing seem like a purchase you “need” in the first place) and just assumes that “AI cool therefore AI good” when making their ad, then yeah, I’m going to be highly skeptical of the thought they put into their actual product.

The only time it wouldn’t raise red flags for me is when it’s used in more of a, I guess you could call it a transitional manner. Like in Coca Cola’s “Masterpiece” ad where they mostly just used it to make the transitions between relatively different scenes look a little more natural, but it was only used for a few frames each time, rather than comprising the vast majority of the promotional material itself.

That ad required many actual talented human artists, and would not have been even physically possible with AI alone, so it evokes a different reaction in my opinion.

Of course, then Coca Cola marketing execs released their complete stock footage-looking AI slop ad a bit later, so it doesn’t seem like that’s a trend that’ll hold up.

And the worst part is, I’m not even sure if they believe it, or if they’re just lying to try and pump the value of the coins they’re investing in that claim to be capable of doing that in the future.

And honestly, I don’t know which I dislike more. Deliberate ignorance, or actual stupidity.

Not that long ago. Many still do, although you’ll primarily find them in more niche spaces within the overarching crypto community.

In fact, just a few years back, I used to be one of them. Of course, later on I became disillusioned with the promises of crypto after learning more about socialism, thinking more closely about how the system fundamentally worked, and realizing that it was effectively just a slightly more distributed variant of capitalism that would inevitably fall to the same structural failings, that being capital accumulation.

To clarify the reasoning that was often used, including by myself, the reason people specifically thought blockchains would make microtransactions better is because they thought that it would lead to more user freedom, and open markets. If you can buy a skin now, then sell it later when you’re done with it, then the effective cost of the skin is lower than in a game where you are unable to sell, for instance.

Obviously the concept of selling in-game items isn’t novel in any way, but the main selling point was that it could be tradeable on any marketplace (or peer-to-peer with no marketplace at all), meaning low to no fees, and they items could be given native revenue-share splits, where the publisher of a game would get a set % of every sale, leading to a way for them to generate revenue that didn’t have to be releasing new but low quality things at a quick pace, and could then allow them to focus on making higher quality items with a slower release schedule.

Of course, looking back retrospectively:

1. Financializing games more just means people play them more for money than for enjoyment
2. This increases the incentives for hacking accounts to steal their items/skins
3. Game publishers would then lose profits from old accounts being able to empty their skins onto the market when they quit the game instead of those skins being permanently tied to that account

There are a small subset of people who legitimately just don’t understand game development fundamentals though, and they actually believe that things would just be fully interchangeable. As in, you buy a skin in Fortnite, and you can then open up Roblox and set it as your player model.

Those ones are especially not the brightest.

To be fair, I do believe their research was based on how convincing it was compared to other Reddit commenters, rather than say, an actual person you’d normally see doing the work for a government propaganda arm, with the training and skillset to effectively distribute propaganda.

Their assessment of how “convincing” it was seems to also have been based on upvotes, which if I know anything about how people use social media, and especially Reddit, are often given when a comment is only slightly read through, and people are often scrolling past without having read the whole thing. The bots may not have necessarily optimized for convincing people, but rather, just making the first part of the comment feel upvote-able over others, while the latter part of the comment was mostly ignored. I’d want to see more research on this, of course, since this seems like a major flaw in how they assessed outcomes.

This, of course, doesn’t discount the fact that AI models are often much cheaper to run than the salaries of human beings.

Yes, unless that service is the kind of thing you think you might pick up later.

For instance, you might use LinkedIn to find a job, but that can still be something you might need in the future, because it’s unlikely you’ll hold that one job forever, and intermittently posting during your existing job could actually help your future prospects.

By contrast, if you used a random site to create a fancier resume, yeah, that account can go straight in the digital wastebasket when you’re done with it. You can always make a new account if you need to make a new resume, and it probably won’t rely on your old account’s data to get that job done.

Ah yes, the “mistake” of deliberately stealing an entire family’s belongings. Easy mistake to make.

Sorry, I wasn’t referencing textbooks specifically. I was moreso referencing the reading materials a lot of kids would want for things like ELA classes in middle/high school, many of which are often lent by larger libraries, since many schools can’t afford to maintain 30+ copies of individual books for each class, especially if that class is reading multiple books per semester, and changing books entirely every year.

Most schools now rely on digital interfaces for their local library like Libby, but of course, when physical branches are shutting down, you end up shifting all physical demand to digital demand as well, which exceeded most libraries’ capacities, since they could only afford to buy (on a subscription basis only) some of the ebook licenses that publishers sell in the quantities required.

I believe textbooks may have been implicated, but I don’t believe it was the bulk of the books that the Archive made available.

You mean as in everyone who owns a book could digitize it and contribute it to the library to be lent out one at a time?

Technically that’s possible, but the real argument being made by rightsholders (such as the publishers suing the Internet Archive) is that they don’t have the right to digitize it and lend it out, because that would be them replicating the work, and thus not just lending out the same copy, even if it’s identical in practice in terms of how many people can access it, and what its content is.

Under current copyright law, you’re going to be sued into oblivion if you try that.

Though to be fair, the main case being made in court that really holds water is that the Internet Archive lent out unlimited copies of digitized copyrighted works during the pandemic when many libraries where physically shut down and unable to offer books. Practically speaking, they did the morally correct thing by providing access to materials that would otherwise have been available, barring the extreme circumstances of the pandemic, but since the publishers thought they deserved to profit from that by selling every student who needed reading material in closed libraries a fresh copy of the book for $20, the Archive is now facing legal consequences, because that’s technically still illegal.

However, if you want a communal library, you kind of get that with things like Little Free Libraries, where you can contribute any book, and books regularly cycle through the neighborhood over time, groups like BuyNothing, where you can very easily have people request and hand off things they no longer want themselves, including books, and you can always technically just start a local group that gets books and lends them like a traditional library would, although some libraries just accept donations of your used books and can lend them out without any additional administrative effort or separate entity set up in your community. That depends on your local library though, if you have one at all.

Chrome is relatively limited in scope compared to, say, a user on an instance of degoogled chromium just using the same Google services along with all the other browsing they do. The extra data that’s gathered is generally going to be things like a little more DNS query information, (assuming your device isn’t already set to default to Google’s DNS server) links you visit that don’t already have Google’s trackers on them (very few) and some general information like when you’re turning on your computer and Chrome is opening up.

The real difference is in how Chrome doesn’t protect you like other browsers do, and it thus makes more of the collection that Google’s services do indirectly, possible.

Perplexity is still being pretty vague here, but if I had to guess, it would essentially just be taking all the stuff that Google would usually get from tracking pixels and ad cookies, and baking that directly in to the browser instead of it relying on individual sites using it.

Is this phone also more secure?

Probably not.

Apple & Google have spent considerable amounts of time building out hardware security infrastructure for their products that I find it extremely unlikely Fairphone would have been able to match.

For example, the popular alternative Android OS GrapheneOS only supports Google Pixels, because: (Emphasis added by me)

“There are currently no other devices meeting even the most basic security requirements while running an alternate OS. GrapheneOS is very interested in supporting a non-Pixel brand, but the vast majority of Android OEMs do not take security seriously. Samsung takes security almost as seriously as Google, but they deliberately cripple their devices when unlock them to install another OS and don’t allow an alternate OS to use important security features. If Samsung permitted GrapheneOS to support their devices properly, many of their phones would be the closest to meeting our requirements. They’re currently missing the very important hardware memory tagging feature, but only because it’s such a new feature”

If even Samsung, the only other phone brand on the market they consider close to meeting their standards, doesn’t support every modern hardware security feature, and deliberately cripples their security for alternate OS’s, as a multi billion dollar company, I doubt Fairphone has custom-built hardware security mechanisms for their phones to the degree that Google has.

Yet another case of “AI” becoming a meaningless word to describe generalized computing techniques that have been in place for years