I value the hard stance on stability and backwards compatibility over the constant churn that some JS libraries/frameworks have. I understand the need both approaches, but this is a breath of fresh air.
I also happen to think that most web apps have no business being so complex, and that too much work goes into dealing with the added complexity of running SPAs where an SPA is not needed.
Because then I can keep the bulk of the logic in a language that’s better designed than JS. Not having to write JS is a huge feature. The added complexity of HTMX is abstracted behind a single library, and the bulk of the logic stays in a better-designed language—Go, Python, Java, Kotlin, Rust, Zig, C#, anything.
Manta Cares | Sr. Front End and Back End Engineers | hybrid-SF | Full time | $145-165K + equity
At Manta Cares we're on a bold journey to demystify healthcare and transform the healthcare experience for patients and their loved ones. Our mission is much bigger than any one of us – we're here to make a real difference in people's lives.
Venture backed (pear.vc, Sozo, 1843 ventures). Come to bootstrap our SF engineering team. We are looking for experienced developers with an affinity to putting great product in the hands of users and for making people's life better when they need it the most.
We are looking for Sr. BE and FE engineers to join the Head of Engineering (me!) in San Francisco (hybrid, 2 office days a week). Because you'd be one of the first US-based engineers, you'll need strong Golang + Postgres or Typescript + React skills, or convince us that you can get there fast. It also helps if you have worked in healthcare environments and / or if you have a passion to help people of all walks of life go through their hardest times. We deal with non-ordinary problems and constraints, bringing simplicity and calmness into very complex and stressful situations.
Manta Cares | Sr. Front End and Back End Engineers | hybrid-SF | Full time | $145-165K + equity
At Manta Cares, we're on a bold journey to demystify healthcare and transform the healthcare experience for patients and their loved ones. Our mission is much bigger than any one of us – we're here to make a real difference in people's lives.
Venture backed (pear.vc, Sozo, 1843 ventures). Come to bootstrap our SF engineering team. We are looking for senior people who like go get their product in the hands of users and also want to contribute to making people's life better when they need it the most.
We are looking for Sr. BE and FE engineers to join the Head of Engineering (me!) in San Francisco (hybrid, 2 office days a week). Because you'd be one of the first US-based engineers, you'll need strong Golang + Postgres or Typescript + React skills, or convince us that you can get there fast. It also helps if you have worked in healthcare environments and / or if you have a passion to help people of all walks of life go through their hardest times. We deal with non-ordinary problems and constraints, bringing simplicity and calmness into very complex and stressful situations.
I have two data points for unschooling, my two children. After a foray in public school in SF, we decided to take one kid out at first, and the other one a year later, the older in 3rd grade, and the younger in 2nd grade. We unschooled them, which we were able to do because we had the equivalent of one full time parent available every day. It was hard work for us, mostly for my wife. It was expensive too. We hired tutors for certain things, we sent them to workshops, classes, etc… but it was always after they expressed interest (one way or another). After 3 years and one year of COVID, our kids expressed interest in going back to school to be around more kids their ages. It was weird for them (and for us!) at first, but they adapted to the schedule pretty quickly and they’re straight As kids.
It worked because in a way, they already had it in them. They both were (and still are) very curious and self-starters. They liked unschooling for a while, and to this day they still have an do-it-yourself attitude when it comes to learning things that they’re interested in —- I think most kids do, and in their case it got cemented during this period.
But they got sick of it too, they got sick of not having a clear schedule, clear goals to meet, too much uncertainty. Now they like more academic-oriented schooling, a bit of the opposite of unschooling, and they’re thriving.
We were lucky that the director of the elementary school they ended up in. She was very welcoming of our kids, and expressed that she herself had an unstructured upbringing. More interestingly, she seemed to think kids should have some unschooling period in their lives, as having free range helps them develop as people.
What does "Silicon Valley" mean here? A few people who struck gold and became billionaires and they loathe the system that got them where they are? Or the general population that work in tech companies in the Bay Area?
So google will eventually be mostly indexing the output of LLMs, and at that point they might as well skip the middleman and generate all search results by themselves, which incidentally, this is how I am using Kagi today - I basically ask questions and get the answers, and I barely click any links anymore.
But this also means that because we've exhausted the human generated content by now as means of training LLMs, new models will start getting trained with mostly the output of other LLMs, again because the web (as well as books and everything else) will be more and more LLM-generated. This will end up with very interesting results --not good, just interesting-- akin to how the message changes when kids the telephone game.
So the snapshot of the web as it was in 2023 will be the last time we had original content, as soon we will have stop producing new content and just recycling existing content.
> Google any recipe, and there are at least 5 paragraphs (usually a lot more) of copy that no one will ever read, and isn't even meant for human consumption. Google "How to learn x", and you'll usually get copy written by people who know nothing about the subject, and maybe browsed Amazon for 30 minutes as research. Real, useful results that used to be the norm for Google are becoming more and more rare as time goes by.
> We're bombarding ourselves with walls of human-unreadable English that we're supposed to ignore. It's like something from a stupid old sci-fi story.
When I read comments today I wonder if there is a human being that wrote them or an LLM.
That, to me, is the biggest difference. Previously I was mostly sure that something I read couldn’t have been generated by a computer. Now I’m fairly certain that I would be fooled quite frequently.
Mm. To me, I think ChatGPT has a certain voice, not sure about the other LLMs.
But perhaps I'm wrong. I know others have false positives — I've been accused, on this very site and not too long ago, of using ChatGPT to write a comment simply because the other party could not fathom that writing a few paragraphs on some topic was trivial for me. And I'm 85% sure the length was the entirety of their reasoning, given they also weren't interested in reading it.
>You’re definitely right about that. CharGPT is almost too accurate/structured.
I think a lot of the material was from standardized testing.
This very structured writing style. Many paragraphs, each discussing one aspect, finished by a conclusion. This is the classic style taught for (American at least) standardized testing, be it SAT, GRE, TOEFL, et al.
Was going to post something similar. There may be a need for a way to confirm ( not detect, which is its own field ) organic content. I hate the thought, because I assume I know where that goes privacy-wise.
Mm. To me, I think ChatGPT has a certain voice, not sure about the other LLMs
How long will it be, before humans reading mostly LLM output, adopt that same writing style? Certainly, for people growing up today, they will be affected.
I remember an HN comment six months or so ago by someone who said they were intentionally modeling their writing on ChatGPT's style. The person said that they were not confident about writing and that they were trying to get better by imitating AI.
One of the many surprising things to me about ChatGPT when it was first released was how well, in its default style, it imitated the bland but well-organized writing style of high school composition textbooks: a clearly stated thesis at the beginning, a topic sentence for each paragraph, a concluding paragraph that often begins "In conclusion."
I mentioned that last point—the concluding "In conclusion"—as an indicator of AI writing to a university class I taught last semester, and a student from Sweden said that he had been taught in school to use that phrase when writing in English.
If I see HN comments that have final paragraphs beginning with "In conclusion" I will still suspect that an LLM has been used. Occasionally I might be wrong, though.
I was taught in high school that using "In conclusion" to open your conclusion was cliche and really almost like an unnecessary slap in the face to the reader. Your composition should end with a conclusion, yes. There was a standard formula for that, yes. But it's not necessary to literally label it as such.
Many of the disliked essay writing cliches are good speech tropes. The difference between reading and listening is that in reading you can skim and skip and rewind, so you don't need structured signposts to guide you through the content. In listening you do. You can't see the last paragraph coming when listening to a speech.
An entertaining informative style of speech can detract from clearly communicating substance. (Of course, the audience rarely wants substance.)
I've intentionally changed some parts of I've comments I've written just because upon reading them back, it felt very close to ChatGPT's style at certain sentences.
I understand. A few months ago, I posted a comment here that attracted several down votes. The content, I thought, was completely innocuous, and I couldn’t figure out at first why some people didn’t like it. Only later did I realize that I might have polished it a little too much and it came out reading like ChatGPT.
How long will it be, before humans reading mostly LLM output, adopt that same writing style?
From what I’ve seen (tutoring high school kids), the picture is much bleaker. They use ChatGPT to write for them but they have no writing style of their own. They can barely put a sentence together just to write the prompt!
As humans, we like to copy things we consider good. I know I tried to use writing styles I thought were funny with varying results. I think I will say never ( or as close to never as possible ).
Given how inhumanly capable it is of seeing both sides of a topic, I guess we'll find out when we see "sublety" and "nuance" go the way way of "literally" and radically change meanings to their near opposite.
> How long will it be, before humans reading mostly LLM output, adopt that same writing style? Certainly, for people growing up today, they will be affected.
Just imagine 180M users of chatGPT having an estimated 1B sessions per month. The model is putting 1-2Trillion tokens into people's brains. People don't assimilate just the writing style and ideas, but also take actions into the real world influenced by the model. Sometimes they create useful discoveries or inventions that end up on the internet and in the next scrape. Full cycle.
That’s the first output from ChatGPT. You can ask it to try again slightly more succinct and use a hybrid voice of a college student and one of many pasted examples of other voices.
When I fed that prompt to my Claude 3.5 based CarlinBot, it refused to write a poem and instead wrote a rant about why it hates tangerines (warning, profanity):
Fuck tangerines, man. Those little orange bastards are a pain in the ass to peel. You spend 10 minutes trying to get that leathery skin off, your fingernails getting all sticky with that goddamn citrus juice. And then when you finally wrestle one of those fuckers open, you got all those little white strings hanging off everywhere. It's like dental floss from hell.
And don't even get me started on how those tangerine slices always shoot juice in your eye when you try to eat them. It's like getting maced by a tiny orange asshole. You ever get tangerine juice in your eye? Shit burns like the fires of hell itself. Makes you want to claw your own goddamn eyeballs out.
Nah, fuck tangerines and their whole stupid family tree. Oranges, clementines, satsumas - they can all go straight to fruit hell as far as I'm concerned. Give me a nice safe banana or an apple any day over those sadistic little citrus grenades. Tangerines are the work of the devil, plain and simple. Fuck writing poems about them little torture devices.
On what does a tangerine wait
Each morning below the winters sun as it awakes?
Do twisted dreams linger, of what it might mean to be a taste on the memory of a forgotten alien tongue?
Is its sacred role seen -- illuminated amongst the greens and unique chaotic chrominance bouncing ancient wisdom between the neighboring leaves?
The tangerine -- victim, pawn, and, ultimately, master ; its search for self in an infinitely growing pile of mixed up words truly complete. There is much to learn.
I was listening to a podcast/article being read in the authors' voice and it took me an embarrassingly long time to realize it was being read by an AI. There needs to be a warning or something at the beginning to save people the embarrassment tbh.
I think it will eventually be good public policy to make it illegal to post massive amounts of texts produced by AI without disclosing it.
As with all illegal things on the internet, it's difficult to enforce, but at least it will make it more difficult/less likely
How about articles written by human charlatans? Claiming they are 'doctors' or 'scientists'. Or posters claiming something that didn't happen? Like a... pro bullshtter claiming he was denied apartment renting because of his skin color. He could make a lot of money if that was true. But poster is still taking ads place, payed by poor 'suffering' minority. Another example 'influencers' who pretending, or really being, experts advise you on forums about products. The tell mostly the truth, but avoid some negative details and competing products and solutions. Without disclosing their connections to businesses.
Shorter version: intentional bullshtting never ends, it's in human, and AI, nature. Like it or not. Having several sources used to help, but now with flood of generated content it may be not the case anymore. If used right this has real affect on business. That's how small sellers live and die on Amazon.
Sure, but for me there isn't anything fundamentally different between a LLM reply and a spammers reply / SEO-vomit. Both are low quality useless junk that gives the masquerade of resembling something worth engaging with.
In fact the really bad spammers were already re-using prompts/templates, think of how many of those recipe novellas shared the same beats. "It was my favorite childhood comfort food", "Cooked with my grandma", blah blah blah
Really? People want to have discussions with other people. I don’t want the output of aggregate data that some tech company worth billions (or the wannabes) might offer. It is truly weird that this needs to be said.
I don’t want this to come across as too negative of a sentiment, but (…) a lot of online discussions are just people repeating opinions they heard elsewhere they agree with. AI is, in this regard, not that different. And marketing is a big part of it, so there are already companies with lots of weight behind making sure that people talk about only certain topics with certain viewpoints (i.e. the Overton window).
Actually original commentary in a discussion is bloody hard to come by.
Sure but the output of an LLM is _never_ original.
Human output signal might be wildly different from person to person if judged on originality. But LLM output is then pure noise. The internet wad already a noisy place but humans are “rate limited” to a degree an LLM is not.
OP is pretty on point. While internet is full of SEO junk, it was far more prevalent back in 2010-2014-5, where the main SEO strategy was to dump 500 words articles in web directories.
The difference is that back then there was an effort from companies like Google to fight the spam and low quality content. Everyone was waiting Matt Cutts( back then head of web spam and search quality at Google) to drop a new update so they can figure out how to step up their game. So at one point you could't afford to just spam your domain with low quality content because you would be penalised, and dropped from the search engines.
There is nothing like that today everybody is on the bandwagon of AI, somehow chatting with pdf documents is now considered by the tech bro hype circle as a sign of enlightenment a beginning of a spark of intelligence...
this is mainly to prolong time on site / impressions that can be served. of course 98% of the banners on those pages are served by doubleclick (google) and thus google makes more money, the crappier the page.
> A recipe is a statement of the ingredients and procedure required for making a dish of food. A mere
listing of ingredients or contents, or a simple set of directions, is uncopyrightable. As a result, the Office cannot register recipes consisting of a set of ingredients and a process for preparing a dish. In contrast, a recipe that creatively explains or depicts how or why to perform a particular activity may be copyrightable. A registration for a recipe may cover the written description or explanation of a process that appears in the work, as well as any photographs or illustrations that are owned by the applicant. However, the registration will not cover the list of ingredients that appear in each recipe, the underlying process for making the dish, or the resulting dish itself. The registration will also not cover the activities described in the work that are procedures, processes, or methods of operation, which are not subject to copyright protection.
Recipes were an easy way to avoid some copyright claims. Copy the list of ingredients, and write a paragraph about how your grandmother made it from a secret recipe that turned out to be on the back of the box.
----
I can still think of content farms and the 2010s and the sheer bulk of junk they produced.
> The former “content creator” — that’s what Demand CEO Richard Rosenblatt calls his freelance contributors — asked to be identified only as a working journalist for fear of “embarrassing” her current employer with her content farm-hand past. She began working for Demand in 2008, a year after graduating with honors from a prestigious journalism program. It was simply a way for her to make some easy money. In addition to working as a barista and freelance journalist, she wrote two or three posts a week for Demand on “anything that I could remotely punch out quickly.”
> The articles she wrote — all of which were selected from an algorithmically generated list — included How to Wear a Sweater Vest” and How to Massage a Dog That Is Emotionally Stressed,” even though she would never willingly don a sweater vest and has never owned a dog.
> “Never trust anything you read on eHow.com,” she said, referring to one of Demand Media’s high-traffic websites, on which most of her clips appeared.
Be VERY careful using Kagi this way -- I ended up turning off Kagi's AI features after it gave me some comically false information based on it misunderstanding the search results it based its answer on. It was almost funny -- I looked at its citations, and the citations said the opposite of what Kagi said, when the citations were even at all relevant.
May I ask how you know those 999 answers were correct, and how would you have been sure to catch a mistake, misinterpretation or hallucination in any of those?
It's not only Kagi AI but Kagi Search itself has been failing me a lot lately. I don't know what they are trying to do but the amount of queries that find zero results is impressive. I've submitted many search improvement reports in their feedback website.
Usually doing `g $query` right after gives me at least some useful results (even when using double quotes, which aren't guaranteed to work always).
Happens about 200 times a day (0.04% of queries), very painful for the user we know, still trying to find root cause (we have limited debugging capabilities as not storing much information). it is on top of our minds.
Yeah, that's totally fair. I just think about all the people to whom I've had to explain LLM hallucinations, and the surprise in their faces, and this feature gives me some heebie-jeebies
Eventually the only purpose of AI as is the only purpose of computers is to enhance human creativity and productivity.
Isn't an LLM just a form of compressing and retrieving vast amounts of information? Is there anything more to it than that?
Don't think LLM itself will ever be able to out compete competent human + LLM. What you will see is that most humans are bad at writing books so they will use LLM and you will get mediocre books. Then there will expert humans that use LLM and are experts to create really good books. Pretty much what we see now. Difference is future you will a lot more mediocre everything. Even worse than it is now. I.e, if you look at Netflix there movies all mediocre. Good movies are the 1% that get released. With AI we'll just have 10 Netflix.
> Don't think LLM itself will ever be able to out compete competent human + LLM
Perhaps, perhaps not. The best performing chess AI, are not improved by having a human team up with them. The best performing Go AI, not yet.
LLMs are the new hotness in a fast-moving field, and LLMs may well get replaced next year by something that can't reasonably be described with those initials. But if they don't, then how far can the current Transformer style stuff go? They're already on-par with university students in many subjects just by themselves, which is something I have to keep repeating because I've still not properly internalised it. I don't know their upper limits, and I don't think anyone really does.
Oh man. Want to know an LLM's limits? Try discussing a new language feature you want to build for an established language. Even more fun is trying to discuss a language feature that doesn't exist yet, even after you provide relevant documentation and examples. It cannot do it. It gets stuck in a rut because the "right" answer is no longer statistically significant. It will get stuck in a local min/max that it cannot easily escape from.
This is a limit of an LLM's architecture. It is based on statistics and can only answer statistical questions. If you want it to provide non-probable answers, an LLM won't work.
Your brain is also based on statistics. We also get stuck in a rut because the "right" answer is no longer statistically significant.
And yet this is not what limits our cognition.
Current LLMs are slow to update with new info, which is why they have cut-off dates so far in the past. Can that be improved to learn as fast (from as little data) as we do? Where's the optimal point on inferring from decreasing data before they show the same cognitive biases we do?
(Should they be improved, or would doing that simply bring in the same race dynamics as SEO?)
Even humans are not good at this. The US military has a test (DLAB) to figure out how good you are at taking in new information in regards to language -- to determine if it is worth teaching you new languages. Some humans are pretty good at this type of thing, but not all. Some humans can't even wrap their heads around algebra but will sell you a vacuum cleaner before you even realize you bought it.
The problem with LLMs is that there is one and it is always the same. Sure, you can get different ones and train your own, to a degree.
> They're already on-par with university students in many subjects just by themselves, which is something I have to keep repeating because I've still not properly internalised it.
That’s because it’s not really true. There are glimpses of this but it trips up too often.
This is a weird take. The paren comment said that, the Internet will not be the same with LLM generated slop.
You're differentiating between LLM generated content and LLM + human combination.
Both will happen, with dire effects to the internet as a whole.
Yeah, but the layout of singular value decomposition and similar algorithms and how pages rank among it is changing all the time. So, par for course. If aspect become less useful people move on. Things evolve, this is a good thing
My experience is that AI tends to surface original content on the web that, in search engines, remains hidden and inaccessible behind a wall of SEOd, monetized, low-value middlemen. The AI I've been using (Perplexity) thumbnails the content and provides a link if I want the source.
The web will be different, and I don't count SEO out yet, but... maybe we'll like AI as a middleman better than what's on the web now.
> So the snapshot of the web as it was in 2023 will be the last time we had original content, as soon we will have stop producing new content and just recycling existing content.
I’ve seen this take before and I genuinely don’t understand it. Plenty of people create content online for the simple reason they enjoy doing it.
They don’t do it for the traffic. They don’t do it for the money. Why should they stop now? Is not like AI is taking away anything from them.
You ask an LLM to do it. Not sarcasm, they’re quite good at ranking the quality of content already and you could certainly fine tune one to be very good at it. You also don’t need to filter out all of the machine written content, only the low quality and redundant samples. You have to do this anyways with human generated writing.
I just tried asking ChatGPT to rate various BBC and NYT articles out of 10, and it consistently gave all of them a 7 or 8. Then I tried today's featured Wikipedia article, which got a 7, which it revised to an 8 after regenerating the respose. Then I tried the same but with BuzzFeeds hilariously shallow AI-generated travel articles[1] and it also gave those 7 or 8 every time. Then I asked ChatGPT to write a review of the iPhone 20, fed it back, and it gave itself a 7.5 out of 10.
I personally give this experiment a 7, maybe 8 out of 10.
ChatGPT has a giant system prompt that you have no control over. Try using Llama and create a system prompt with clear instructions and examples. If you were going to use a model in a production system you would also want to either fine tune it or train a BERT-like model as a classifier that just outputs a score. Maybe even more than one for ranking along different dimensions.
Except AI in search is taking away significant traffic from everywhere, and it hits small blogs as well as nonprofits like encyclopaedias the hardest, while misrepresenting and “remixing” the actual content.
I’ve given up on the internet as a place to share my passions and hobbies for the most part, and while LLM’s weren’t the only reason, this current trend is a significant factor. I focus most of my attention on talking directly with people. And yes that does mean the information I share is guaranteed to be lost to time, but I’d rather it be shared in a meaningful manner in the moment than live on in an interpreted zombie form in perpetuity.
I have a blog. Been writing on that for 7 years. Should I care if AI in search is taking away traffic? If yes, why? I’m not writing for traffic. I write because I enjoy doing it. People find my way mostly thanks to other people linking to my site. And a solid % of traffic comes from RSS anyway.
I think giving up on the web because of AI is the wrong move. You should still create and focus more on connecting with others directly, when online. Get in touch, write emails, sign guestbooks.
I’m personally having great exchanges daily with people from all over via email and that won’t stop because of stupid ChatGPT or whatever.
And don’t get me wrong, it’s awesome to spend more time offline so if you want to do down that path it’s great.
The only reason to put things you write online is to make it available to others. If writing simply for my own enjoyment or reference I write in my notebooks, as I do all the time. I never stopped doing that.
A lot of people who create content don't want their content to feed AI. They love what they do and they don't want their work to support a system whose purpose is to debase and commoditize that work. The only way to avoid that is to never publish to the web, everything published to the web feeds AI. That is the web's purpose now.
Also there are plenty of people who create content because they love it, and also need to be able to make a living at it, because doing so at the level of quality they want is time consuming and expensive.
But mostly because even people who produce content because they love it want to share that content with the world and that will be nigh impossible when the only content anyone sees, and that any platform or algorithm surfaces, is AI generated. Why put in the effort and heart and work to create something only for an AI to immediately clone it for ad revenue? Why even bother?
> The only way to avoid that is to never publish to the web, everything published to the web feeds AI. That is the web's purpose now.
And in doing that you also prevent real humans from accessing that same content. Look, I have no simpathy for AI companies. I wrote about it before on my site, will probably write again. The current situation sucks. But giving up is not the right answer imo.
> Also there are plenty of people who create content because they love it, and also need to be able to make a living at it, because doing so at the level of quality they want is time consuming and expensive.
Fair but those are the minority. I'd argue the vast majority of people create content because they enjoy the process and earn a living in other ways. I run a newsletter where I interview people with blogs and so far, after a year running it, not a single person has told me they blog for a living. Every single one is doing it for passion. And I suspect that's true for the vast majority of people out there. The bulk of internet content (when it comes to creative content that is) is created by people who do it as a hobby.
> But mostly because even people who produce content because they love it want to share that content with the world and that will be nigh impossible when the only content anyone sees, and that any platform or algorithm surfaces, is AI generated. Why put in the effort and heart and work to create something only for an AI to immediately clone it for ad revenue? Why even bother?
Why even bother? Because there are people out there who care. And the assumption that "the only content anyone sees, and that any platform or algorithm surfaces, is AI generated" is a wrong one imo. I can assure you that there are PLENTY of people out there who still value original content, still value connecting with real human beings doing things because they love the craft. Assuming everything is doomed is not helpful.
Is it going to be harder? Yes. Are there solution? Yes.
In an infinitely large world with an infinitely large number of monkeys typing an infinite number of words on an infinite number of keyboards, "just index everything and threat it as fact" isn't a viable strategy any more.
We are now much closer to that world than we ever were before.
> new models will start getting trained with mostly the output of other LLMs
That is a naive, flawed way to do it. You need to filter and verify synthetic examples. How? First you empower the LLM, then you judge it. Human in the loop (LLM chat rooms), more tokens (CoT), tool usage (code, search, RAG), other models acting as judges and filters.
This problem is similar to scientific publication. Many papers get published, but they need to pass peer review, and lots of them get rejected. Just because someone wrote it into a paper doesn't automatically make it right. Sometimes we have to wait a year to see if adoption supports the initial claims. For medical applications testing is even harder. For startups it's a blood bath in the first few years.
There are many ways to select the good from the bad. In the case of AI text, validation can be done against the real world, but it's a slow process. It's so much easier to scrape decades worth of already written content than to iterate slowly to validate everything. AlphaZero played millions of self games to find a strategy better than human.
In the end the whole ideation-validation process is a search for trustworthy ideas. In search you interact with the search space and make your way towards the goal. Search validates ideas eventually. AI can search too, as evidenced by many Alpha model (AlphaTensor, AlphaFold, AlphaGeometry...). There was a recent paper about prover-verifier systems trained adversarially like GANs, that might be one possible approach. https://arxiv.org/abs/2407.13692v1
Communal spaces are fine, communal spaces will continue to be fine. Forums are fine. IRC is fine. The only thing that's dying is Google. Google is not the Internet.
> this also means that because we've exhausted the human generated content by now as means of training LLMs, new models will start getting trained with mostly the output of other LLMs
There is also a rapidly growing industry of people whose job it is to write content to train LMs against. I totally expect this to be a growing source of training data at the frontier instead of more generic crap from the internet.
Smaller models will probably stay trained on bigger models, however.
> growing industry of people whose job it is to write content to train LMs against
Do you have an example of this?
How do they differentiate content written by a person v/s written by LLM, I'd expect there is going to be people trying to "cheat" by using LLMs to generate content.
> How do they differentiate content written by a person v/s written by LLM
Honestly, not sure how to test it, but this is B2B contracts, so hopefully there's some quality control. It's part of the broad "training data labeling" business, so presumably the industry has some terms in contracts.
ScaleAI, Appen are big providers that have worked with OpenAI, Google, etc.
They just give you such an insight into another human being in this raw fashion you don’t get through a persona built website.
My own blog is very similar. Haphazard and unprofessional and perhaps one day slurped into an LLM or successor (I have no problem with this).
Perhaps one day some other guy will read my blog like I read Makoto Matsumoto’s. If they feel that connection across time then that will suffice! And if they don’t, then the pleasure of writing will do.
And if that works for me, it’ll work for other people too. Previously finding them was hard because there was no one on the Internet. Now it’s hard because everyone’s on it. But it’s still a search problem.
Print-on-demand means that paper books will be just as flooded with LLM sludge as eBook stores. I think we are at risk of regressing back to huge publishers being de-facto gatekeepers, because every easily accessible avenue to getting published is going to get crushed under this race to the bottom.
Likewise with record labels if platforms like Spotify which allow self-publishing get overwhelmed with Suno slop, which is already on the rise (there's some conspiracy theories that Spotify themselves are making it, but there's more than enough opportunistic grifters in the world who could be trying to get rich quick by spamming it).
> The Fifty Shades trilogy was developed from a Twilight fan fiction series originally titled Master of the Universe and published by James episodically on fan fiction websites under the pen name "Snowqueen Icedragon". Source : https://en.wikipedia.org/wiki/Fifty_Shades_of_Grey
The AI is already tainted with human output.... If you think its spitting out garbage it's because that's what we fed it.
There is the old Carlin bit about "for there to be an average intelligence, half of the people need to be below it".
Maybe we should not call it AI rather AM, Artificial Mediocrity, it would be reflection of its source material.
The issue is that the AI shit is flooding out anything good. Nearly any metric you can think of to measure "good" by is being gamed ATM which makes it really hard to actually find something good. Impossible to discover new/smaller authors.
Scale matters. The ability to churn out bad writing is increasing by orders of magnitude and could drown out the already small amount of high quality works.
While it's true that the volume of bad writing is increasing, our ability to analyze and refine this sludge is also improving. Just as spell check and grammar check give instant feedback why not AI instant feedback about writing quality / originality / suitability / correctness / … ? If instant feedback can improve spelling and grammar why not these other things?
> But this also means that because we've exhausted the human generated content
Putting aside the question of whether dragnet web scraping for human generated content is necessary to train next gen models, OpenAI has a massive source of human writing through their ChatGPT apps.
I use LLM output from kagi too. But given the rate of straight-up factually incorrect stuff that comes out of it, I need it to come with a credible source that I can verify. If not, I'm not taking any of it seriously.
AlphaGo learned to play Go by playing with itself. Why couldn't LLM do the same? They got plenty of information to be used as a starting point, so surely they can figure out some novel information eventually.
LLMs aren't logically reasoning through an axiomatic system. Any patterns of logic they demonatrate are just recreated from patterns in input data. Effectively, they can't think new thoughts.
> Effectively, they (LLMs) can't think new thoughts.
This is true only if you assume that combining existing thought patterns is not new thinking. If they can't learn a certain pattern from training data, indeed they would be stuck. However, their training data keeps growing and updating, allowing each updated version to learn more patterns.
Google really missed the opportunity of becoming ChatGPT. LLMs are the best interface for search but not yet the best interface for ads so it makes sense for them to not make the jump. ChatGPT and Claude are today what Google was in 2000 and should have evolved to.
Mind you they will be trained on what humans have filtered as being acceptable content. Most of the trash produced by ML that hits the web is quickly buried and never referenced.
The incentives will be largely gone when SEO-savvy AI bots will produce 10K articles in the time it takes you to write one, so your article will be mostly unfindable in search engines.
Human generated content will be outpaced by AI generated content by a large margin, so even though there'll still be human content, it'll be meaningless on aggregate.
We can adapt. There's already invite-only and semi-closed online communities. If the "mainstream" web becomes AI-flooded, where you'd you like to hang out / get information: the mainstream AI sludge, or the curated human communities?
The LLM is trained by measuring its error compared to the training data. It is literally optimizing to not be recognizable. Any improvement you can make to detect LLM output can immediately be used to train them better.
GANs do that, I don't think LLMs do. I think LLMs are mostly trained on "how do I recon a human would rate this answer?", or at least the default ChatGPT models are and that's the topic at the root of this thread. That's allowed to be a different distribution to the source material.
Observable: ChatGPT quite often used to just outright says "As a large language model trained by OpenAI…", which is a dead giveaway.
This is the result of RLHF (which is fine-tuning to make the output more palatable), but this is not what training is about.
The actual training process makes the model output be the likeliest output, and the introduction phrase you quoted would not come out of this process if there was no RLHF. See GPT3 (text-davinci-003 via API) which didn't have RLHF and would not say this, vs. ChatGPT which is fine-tuned for human preferences and thus will output such giveaways.
This seems like it would only work if you deliberately rank AI-generated text above human generations.
If the AI generations are correct, is it really that bad? If they're bad, I feel like they're destined to fall to the bottom like the accidental Facebook uploads and misinformed "experts" of yesteryear.
Where would the AI get the data necessary to generate correct answers for novel problems or current events? It's largely predictive based on what's in the training set.
> Where would the AI get the data necessary to generate correct answers for novel problems or current events?
In a certain sense, it doesn't really need it. I like to think of the Library of Babel as a grounding thought experiment; technically, every truth and lie could have already been written. Auguring the truth from randomness is possible, even if only briefly and randomly. The existence of LLMs and tokenized text do a really good job of turning statistics-soup into readable text.
That's not to say AI will always be correct, or even that it's capable of consistent performance. But if an AI-generated explanation of a particular topic is exemplary beyond all human attempts, I don't think it's fair to down-rank as long as the text is correct.
Are you suggesting that llms can predict the future in order to address the lack of current event data in their training set? Or is it just implicit in your answer that only the past matters?
We would lose the long tail, but if I were a search engine, I would have a mode that only returned results on a whitelist of domains that I would have a human eyeball every few months.
If somebody had a site that we were not indexing and wanted to be, they could pay a human to review it every few months.
You can make as many sites as you like, but I would still ask a human to review them and make a judgment call on whether other humans might be interested in the content before indexing them.
You can record as many albums as you like as well, but the DJ needs to like your music before they play it on the radio.
I guess what I’m saying is I don’t want the Internet to become a Top 25 radio station cranking out scam entertainment for the masses. I want “small pirate and indie radio” to be the norm. If you want top 25’s, go back to centralized, curated media.
The thing with the AI content boom is that if there’s 1000x more of it than there is genuine indie stations, it gets harder to find the real content. Piping things through a top25 filter doesn’t fix that, or actively makes it worse due to the incentives to monopolize / plan the system.
For example, i've been appending "site:reddit.com" to some of my Google queries for a while now —especially when searching for things like reviews— because, otherwise, Google search results are unusable: ads disguised as fake "reviews" rank higher than actual reviews made by people, which is what i'm interested in.
I wouldn't be surprised if we evolve some similar adaptations to deal the flood of AI-generated shit. Like favoring closer-knit communities of people we trust, and penalizing AI sludge when it sips in.
It's still sad though. In the meantime, we might lose a lot of minds to this. Entire generations perhaps. Watching older people fall for AI-generated trash on Facebook is painful. I hope we acted sooner.
Ok, i'll concede that you have a very good point there. Trust can be (and is being) exploited.
I guess for me, so far at least, some sites feel much more legit and human that the obviously bot-ridden mess that are the likes of Twitter/Instagram/FB. Like for example here or on Lobsters (and more on the latter) i have the feeling that it's mostly people talking with people. On the couple of relatively-small subreddits i visit, i feel the same too.
But i could be wrong of course. Maybe the tone of an HN poster is super easy for an LLM to copy; there's a reason why "shit HN says" exists after all. The only reason i have to believe otherwise is that, in comparison, Instagram or Twitter bots are so obvious and bland, and those companies have way more resources to throw at AI than HN or reddit :P
This event is predicted in Sydney Dekker’s book “Drift into Failure”, which basically postulates that in order to prevent local failure we setup failure prevention systems that increase the complexity beyond our ability to handle, and introduce systemic failures that are global. It’s a sobering book to read if you ever thought we could make systems fault tolerant.
We need more local expertise is really the only answer. Any organization that just outsources everything is prone to this. Not that organizations that don't outsource aren't prone to other things, but at least their failures will be asynchronous.
Funny thing is that for decades there were predictions about how there was a need for millions of more IT workers. It was assumed one needed local knowledge in companies. Instead what we got was more and more outsourced systems and centralized services. This today is one of the many downsides.
The problem here would be that there's not enough people who can provide the level of protection a third-party vendor claims to provide, and a person (or persons) with comparable level of expertise would be much more expensive likely. So companies who do their own IT would be routinely outcompeted by ones that outsource, only for the latter to get into trouble when the black swan swoops in. The problem is all other kinds of companies are mostly extinct by then unless their investors had some super-human foresight and discipline to invest for years into something that year after year looks like losing money.
> The problem here would be that there's not enough people who can provide the level of protection a third-party vendor claims to provide, and a person (or persons) with comparable level of expertise would be much more expensive likely.
Is that because of economies of scale or because the vendor is just cutting costs while hiding their negligence?
I don't understand how a single vendor was able to deploy an update to all of these systems virtually simultaneously, and _that_ wasn't identified as a risk. This smells of mindless box checking rather than sincere risk assessment and security auditing.
Kinda both I think, with an addition of principal agent problem. If you found a formula that provides the client with an acceptable CYA picture it is very scalable. And the model of "IT person knowledgeable in both security, modern threats and company's business" is not very scalable. The former, as we now know, is prone to catastrophic failures, but those are rare enough for a particular decision-maker to not be bothered by it.
Depressing thought that this phenomena is some kind of Nash equilibrium. That in the space of competition between firms, the equilibrium is for companies to outsource IT labor, saving on IT costs and passing that cost savings onto whatever service they are providing. -> Firms that outsource, out-compete their competition + expose their services to black swan catastrophic risk.
Is regulation that only way out of this, from a game theory perspective?
The whole market in which crowdstrike can exist is a result of regulation, albeit bad regulation.
And since the returns of selling endpoint protection are increasing with volume, the market can, over time, only be an oligopoly or monopoly.
It is a screwed market with artificially increased demand.
Also the outsourcing is not only about cost and compliance. There is at least a third force. In a situation like this, no CTO who bought crowdstrike products will be blamed. He did what was considered best industry practice (box ticking approach to security). From their perspective it is risk mitigation.
In theory, since most of the security incidents (not this one) involve the loss of personal customer data, if end customers would be willing to a pay a premium for proper handling of their data, AND if firms that don’t outsource and instead pay for competent administrators within their hierarchy had a means of signaling that, the equilibrium could be pushed to where you would like it to be.
Those are two very questionable ifs.
Also how do you recognise a competent administrator (even IT companies have problems with that), and how many are available in your area (you want them to live in the vicinity) even if you are willing to pay them like the most senior devs?
If you want to regulate the problem away, a lot of influencing factors have to be considered.
Also a major point in the Black Swan. In the Black Swan, Taleb describes that it is better for banks to fail more often than for them to be protected from any adversity. Eventually they will become "too big to fail". If something is too big to fail, you are fragile to a catastrophic failure.
I was wondering when someone would bring up Taleb RE: this incident.
I know you aren't saying it is, but I think Taleb would argue that this incident, as he did with the coronavirus pandemic for example, isn't even a Black Swan event. It was extremely easy to predict, and you had a large number of experts warning people about it for years but being ignored. A Black Swan is unpredictable and unexpected, not something totally predictable that you decided not to prepare for anyways.
That is interesting, where does he talk about this? I'm curious to hear his reasoning. What I remember from the Black Swan is that Black Swan events are (1) rare, (2) have a non-linear/massive impact, (3) and easy to predict retrospectively. That is, a lot of people will say "of course that happened" after the fact but were never too concerned about it beforehand.
Apart from a few doomsdayers I am not aware of anybody was warning us about a crowd strike type of event. I do not know much about public health but it was my understanding that there were playbooks for an epidemic.
Even if we had a proper playbook (and we likely do), the failure is so distributed that one would need a lot of books and a lot of incident commanders to fix the problem. We are dead in the water.
I think it was "predicted" by Sunburst, the Solarwinds hack.
I don't think centrally distributed anti-virus software is the only way to maintain reliability. Instead, I'd say companies to centralize anything like administration since it's cost effective and because they actually aren't concerned about global outage like this.
JM Keynes said "A ‘sound’ banker, alas! is not one who foresees danger and avoids it, but one who, when he is ruined, is ruined in a conventional and orthodox way along with his fellows, so that no one can really blame him." and the same goes for corporate IT.
Many systems are fault tolerant, and many systems can be made fault tolerant. But once you drift into a level of complexity spawned by many levels of dependencies, it definitely becomes more difficult for system A to understand the threats from system B and so on.
Do you know of any fault tolerant system? Asking because in all the cases I know, when we make a system "fault tolerant" we increase the complexity and we introduce new systemic failure modes related to our fault-tolerant-making-system, making them effectively non fault tolerant.
In all the cases I know, we traded frequent and localized failure for infrequent but globalized catastrophic failures. Like in this case.
You can make a system tolerant to certain faults. Other faults are left "untolerated".
A system that can tolerate anything, so have perfect availability, seems clearly impossible. So yeah, totally right, it's always a tradeoff. That's reasonable, as long as you trade smart.
I wonder if the people deciding to install Crowdstrike are aware of this. If they traded intentionally, and this is something they accepted, I guess it's fine. If not... I further wonder if they will change anything in the aftermath.
There will be lawsuits, there will be negotiations for better contracts, and likely there will be processes put in place to make it look like something was done at a deeper level. And yet this will happen again next year or the year after, at another company. I would be surprised if there was a risk assessment for the software that is supposed to be the answer to the risk assessment in the first place. Will be interesting to see what happens once the dust settles.
- This is system has a single point of failure, it is not fault tolerant. Lets introduce these three things to make it fault-tolerant
- Now you have three single points of failure...
It really depends on the size of the system and the definition of fault tolerance. If I have a website calling out to 10 APIs and one API failure takes down the site, that is not fault tolerance. If that 1 API failure gets caught and the rest operate as normal, that is fault tolerance, but 10% of the system is down. If you go to almost any site and open the dev console, you'll see errors coming from parts of the system, that is fault tolerance. Any twin engine airplane is fault tolerant...until both engines fail. I would say the solar system is fault tolerant, the universe even moreso if you consider it a system.
tldr there are levels to fault tolerance and I understand what you are saying. I am not sure if you are advocating for getting rid of fault handling, but generally you can mitigate the big scary monsters and what is left is the really edge case issues, and there really is no stopping one of those from time to time given we live in a world where anything can happen at anytime.
This instance really seems like a human related error around deployment standards...and humans will always make mistakes.
well, you usually put a load balancer and multiple instances of your service to handle individual server failures. In a basic no-lb case, your single server fails, you restart it and move on (local failure). In a load balancer case, your lb introduces its own global risks e.g. the load balancer can itself fail, which you can restart, but the load balancer can have a bug and stop handling sticky sessions when your servers are relying on it, and now you have a much harder to track brown-out event that is affecting every one of your users for a longer time, it's hard to diagnose, might end up with hard to fix data issues and transactions, and restarting the whole might not be enough.
So yeah, there is no fault tolerance if the timeframe is large enough, there are just less events, with much higher costs. It's a tradeoff.
The cynical in me thinks that the one advantage of these complex CYA systems is that when systems fail catastrophically like CrowdStrike did, we can all "outsource" the blame to them.
It's also in line with arguments made by Ted Kaczynski (the Unabomber)
> Why must everything collapse? Because, [Kaczynski] says, natural-selection-like competition only works when competing entities have scales of transport and talk that are much less than the scale of the entire system within which they compete. That is, things can work fine when bacteria who each move and talk across only meters compete across an entire planet. The failure of one bacteria doesn’t then threaten the planet. But when competing systems become complex and coupled on global scales, then there are always only a few such systems that matter, and breakdowns often have global scopes.
crazy how much he was right. if he hadn't gone down the path of violence out of self-loathing and anger he might have lived to see a huge audience and following.
I suppose we wouldn't know whether an audience for those ideas exists today because they would be blacklisted, deplatformed, or deamplified by consolidated authorities.
There was a quote last year during the "Twitter files" hearing, something like, "it is axiomatic that the government cannot do indirectly what it is prohibited from doing directly".
Perhaps ironically, I had a difficult time using Google to find the exact wording of the quote or its source. The only verbatim result was from a NYPost article about the hearing.
>I suppose we wouldn't know whether an audience for those ideas exists today because they would be blacklisted, deplatformed, or deamplified by consolidated authorities.
Be realistic, none his ideas would be blacklisted. They sound good on paper, but the instant it's time for everyone to return to mudhuts and farming, 99% of people will return to Playstations and ACs.
He wasn't "silenced" because the government was out to get him, no one talks about his ideas because they are just bad. Most people will give up on ecofascism once you tell them that you won't be able to eat strawberries out of season.
"would be blacklisted, deplatformed, or deamplified by consolidated authorities"
Sorry. Not true. You have Black Swan (Taleb) and Drift into Failure (Dekker) among many other books. These ideas are very well known to anyone who makes the effort.
The only thing that got Unabomber blacklisted is that he started to send bombs to people. His manifesto was dime a dozen, half the time you can expect politician boosting such stuff for temporary polling wins.
Hell, if we take his alleged (don't have vetted the genealogy tree) cousins, his body count isn't even that impressive.
I think a surprising amount of people already share this view, even if they don't go into extensive treatment with references like Dekker presumably does (I haven't read it).
I suspect most people in power just don't subscribe to that. which is precisely why it's systemic to see the engineer shouting "no!" when John CEO says "we're doing it anyway." I'm not sure this is something you can just teach, because the audience definitely has reservations about adopting it.
You can't prevent failure. You can only mitigate the impact. Biology has pretty good answers as to how to achieve this without having to increase complexity as a result, in fact, it often shows that simpler systems increase resilliency.
Something we used to understand until OS vendors became publicly traded companies and "important to national security" somehow.
> if you ever thought we could make systems fault tolerant
The only possible way to fault tolerancy is simplicity and then more simplicity.
Things like crowsdtrike have the opposite approach. Add a lot of fragile complexity attempting to catch problems, but introducing more attack surfaces than they can remove. This will never succeed.
As an architect of secure, real-time systems, the hardest lesson I had to learn is there's no such thing as a secure, real-time system in the absolute sense. Don't tell my boss.
I haven't read it, but I'd take a leap to presume it's somewhere between the people that say "C is unsafe" and "some other language takes care of all of things".
I have one of these for earthquake preparedness, along with large containers with water. I think i got it 10 years ago from Costco, and I sure hope not having to try the food out.
Since Covid times we’ve been keeping a healthy stock of shelf-stable food that we’re continuously restocking. Our food supply chain seems too concentrated on fewer and fewer suppliers that in turn rely on long haul transportation, which makes it somewhat easy to disrupt and end up with stuff missing from stores for long periods. It feels safe to have a buffer of decent food. The end-of-times-bucket from Costco is therefore only for Zombie apocalypse or major earthquake :D
And others might be able to buy a house for the first time. The issue with houses being the main form of families to acquire wealth is that newcomers to the market have it much much harder to enter participate in it as time goes on and house values increase.
So people losing some "paper" savings is a price to pay so other people can also own.
But in reality, when homeowners struggle and sell their homes is mostly large real-estate funds that buy them. So it's possible that homes losing their value means owners lose savings but newcomers still can't enter the home ownership virtuous cycle.
The principle paid down on those repayment mortgage came from homeowner income and was diverted away from other savings and investments.
Punishing millions of people who just sought housing security because, for reasons beyond their control, the Fed decided to cut rates after the GFC and keep them low for 15 years is just not tenable.
Fortunately, in the UK at least, it looks like (nominal) house prices will remain stagnant until real earnings catch up with rates. A better all round solution.
> So it's possible that homes losing their value means owners lose savings but newcomers still can't enter the home ownership virtuous cycle.
if a house is cheap, but is still valuable (as new owners _want_ to buy because it's worth it while it's cheap), then this means it's a mis-pricing. And mis-pricing means that any investing entity will want to buy as well. In aggregate, this _should_ push the price back to the correct equilibrium.
Therefore, housing being expensive today is a reflection of how many people value it high. It might also indicate that the reason it was cheap before was a mispricing, and those who got in early was merely lucky.
I also happen to think that most web apps have no business being so complex, and that too much work goes into dealing with the added complexity of running SPAs where an SPA is not needed.