But we've gained some new ones. I find ChatGPT-generated text predictable in structure and lacking any kind of flair. It seems to avoid hyperbole, emotional language and extreme positions. Worthless is subjective, but ChatGPT-generated text could be considered worthless to a lot of people in a lot of situations.

I used to do SEO copywriting in high school and yeah, ChatGPT's output is pretty much at the level of what I was producing (primarily, use certain keywords, secondarily, write a surface-level informative article tangential to what you want to sell to the customer).

> At some point it may become impossible to separate the wheat from the chaff.

I think over time there could be a weird eddy-like effect to AI intelligence. Today you can ask ChatGPT a Stack Overflow-style and get a Stack Overflow-style response instantly (complete with taking a bit of a gamble on whether it's true and accurate). Hooray for increased productivity?

But then, looking forward years in time, people start leaning more heavily on that and stop posting to Stack Overflow and the well of information for AI to train on starts to dry up, instead becoming a loop of sometimes-correct goop. Maybe that becomes a problem as technology evolves? Or maybe they train on technical documentation at that point?

One possible future is going back to a medium with higher cost of publication. Books. Handchiseled stone tablets. Offering information costs something.

If you didn't submit a proof of work of N or greater difficulty the email would be thrown out.

The grifters are all over that already. No AI necessary to generate and publish drivel.

See “Contrepreneurs: The Mikkelsen Twins”¹ by Dan Olson² for an informative and entertaining documentary on the matter.

¹ https://www.youtube.com/watch?v=biYciU1uiUw

² A.k.a Folding Ideas. A.k.a. the creator of “Line Goes Up – The Problem With NFTs”.

Honestly I’ve switched to books and papers a few years ago and it has been fantastic. 2 hours of reading a half decent book or paper outweighs a week of reading the best blogposts, twitter threads, or YouTube videos.

Then once you have a hook into your topic, it usually cites 30+ other papers that may be worth reading. You will never run out.

This brings up a good sub-topic. "Noise" as I mean it is where it's something you cannot definitely validate the veracity of in short order, or you do and it's useless.

The trash TV thing is a great example: if you are watching Beavis & Butthead because you know its trash and you need to zone out, that's a conscious, active decision, and you are in effect, 'in on the joke'...if you can't discern that it's satire and find yourself relating to the characters, you might be part of the problem :)

AI generated images used to look AI generated. Midjourney v6 and well tuned sdxl models look almost real. For marketing imagery, Midjourney v6 can easily replicate images from top creative houses now.

for coding tasks I'd imagine it could be trained on the actual source code of the libraries or languages and determine proper answers for most questions. AI companies have seen success using "synthetic" data, but who knows how much it can scale and improve

It would be hilarious if the end result of all this would be to go back to a 1990s-2000s Yahoo style of web portal where all the links are curated by hand by reputable organizations.

The internet was already becoming ad farms. This is the final blow and now the internet as we knew it will die.

I’m not that pessimistic about llm generated content. I’m starting to use it to rewrite my online and slack comments for grammar, I’m also using it for brainstorming, enhancing things I create, code (not as in “ok ai write me an app” but as in “change this code to do this, ok this is not considering x and y edge cases, ok use this other method, ok refactor that” it is saving me a lot of typing and silly mistakes while I focus on the meat of the problem.

I've also learned a lot from chatting with Bing AI. The caveat is there that you all the time have in the back of your mind that the answer might be wrong. It helps to keep asking more detailed questions and check whether the set of answers keep on making sense as a whole. That way of using it has helped me a lot. See it as getting info from a very smart friend who sometimes had too much to drink.

Of course, I carefully frame the query so that's what I get.

However, when I asked stackoverflow, google, and bard a question about how to do something with github, all I received were wrong answers. I finally had to throw in the towel and ask people. I think it was the third person I asked who gave an answer that worked.

google itself has an annoying habit of answering a fundamentally different question than what I type in.

It doesn't matter if the training data is AI generated or not, if it is useful.

Has anyone tested a marketing campaign using copy from a human copywriter versus an AI one?

I would like to see which one converts better.

and WRT to the eddy-like model-self-incestuation - I am sure that the scope of that well just becomes wider - now its slurping any and all video and learning human micro emotions and micro-aggressions - and mastering human interpersonal skills.

My prediction is that AI will be a top-down reflection of societies' leadership. So as long as we have these questionable leaders throughout the world governments and global corps - the Alignment of AI will be bias to their narratives.

That's a lot, compared to mine. How do you organize replication and do you make backups on any external services? I kinda do want to hoard more, but I find it complicated to deal with at large scale. It gets expensive to back-up everything, and HDDs aren't really a solid media long-term. Now, I can kinda use my judgment of what is important and what is essentially trash I store just in case, but losing 100TB of trash would be pretty devastating too, TBH.

I think something interesting to note is that once we stopped atmospheric nuclear testing steel radiation levels went back down and are almost at normal background levels. So maybe the same thing will happen if we stop using GenAI.

Mine post-apocalyptic scenario I half-jokingly predicted some 5-10 years ago was that all general-purpose computing hardware, which is common and relatively cheap now will be abandoned and possibly outlawed in the end. People will use non-rootable thin clients to access Amazon, which will have general-purpose hardware, but it will be heavily audited by government entities.

I don't want AI to be at the forefront of all new media and artwork. That's a terrible outcome to me.

And honestly there's already too much "content" in the world and being produced every day, and it seems like every time we step further up the "content is easier to produce and deliver" ladder, it actually gets way more difficult to find much of value, and also more difficult for smaller artists to find an audience.

We see this on Steam where there are thousands of new game releases every week. You only ever hear of one or two. And it's almost never surprising which ones you hear about. Rarely you get an indie sensation out of nowhere, but that only usually happens when a big streamer showcases it.

Speaking of streamers, it's hard to find quality small streamers too. Twitch and YouTube are saturated with streams to watch but everyone gravitates to the biggest ones because there's just too much to see.

Everything is drowning in a sea of (mostly mediocre, honestly) content already, AI is going to make this problem much worse.

At least with human generated media, it's a person pursuing their dreams. Those thousands of games per week might not get noticed, but the person who made one of them might launch a career off their indie steam releases and eventually lead a team that makes the next Baldur's Gate 3 (substitute with whatever popular game you like)

I can't imagine the same with AI. Or actually, I can imagine much worse. The AI that generates 1000 games eventually gets bought by a company to replace half their staff and now a bunch of people are out of work and have a much harder uphill battle to pursue their dreams (assuming that working on games at that company was their dream)

I don't know. I am having a hard time seeing a better society growing out of the current AI boom.

This experiment has been run in most wealthy nations and the artwork renaissance didn't happen.

Most older people don't do arts/sciences when they retire from work.

From what I see of younger people that no longer have to work (for whatever reason) neither do younger people become artists given the opportunity.

Or look at what people of working age do with their free time in evenings or weekends after they've done their work for the week. Expect people freed from work to do more of the same as what they currently do in evenings/weekends: don't expect people will suddenly do something "productive".

This isn't my experience. I know a bunch of old folks doing woodcarving, quilting, etc. Its just not the kind of arts you've got in mind.

It’s just published on YouTube. Seriously the quality of diy videos and everything is sometimes PBS or BBC quality or better.

What retirees often do, rather, is develop an artist’s eye for images, a musician’s ear for sounds, a philosopher’s perspective, a writer’s voice, etc.. This often involve a broader exposure/consumption of arts and studying art history. Sometimes producing actual art as well…but less for the final artistic product but instead to engage in the artistic process itself so as to develop that way of seeing/feeling/being an artist has. When the work-related chunk of the mind is wholly freed up for other pursuits, there is often such a bit-flip. And since it is a deepening appreciation and greater consumption, there is no risk of overproduction of art and the soul devolution that arises from hyper competitiveness in the marketplace.

There's Genshin Impact, Pokemon Go, Superhot, Beat Saber, Monument Valley, Subnautica, Among Us, Rust, Cities:Skylines (maybe), Ori (maybe), COD:Mobile (maybe) and...?

You could say the same about books.

Lowering the barriers of entry does mean more content will be generated and that content won't the same bar as having a middleman who was the arbiter of who gets published but at the same time, you'll likely get more hits and new developers because you getting more people swinging faster to test the market and hone their eye.

I am doubtful that there are very many people who hit a "Best Seller" 10/10 on their first try. You just used to not see it or ever be able to consume it because their audience was like 7 people at their local club.

Which suggests even after several iterations the vast vast majority of folks are not putting out anything noteworthy.

Cuphead

Escape Academy

Overcooked

Monster Sanctuary

Lunistice

I'd say all of those do some major thing that makes them stand out.

Angry birds, Slender: The Eight Pages, Kerbal Space Program, Plague Inc, The Room, Rust, Tabletop Simulator, Enter the Gungeon, Totally Accurate Battle Simulator, Clone Hero, Cuphead, Escape from Tarkov, Getting Over It with Bennett Foddy, Hollow Knight, Oxygen Not Included, Among Us, RimWorld, Subnautica, Magic: The Gathering Arena, Outer Wilds, Risk of Rain 2, Subnautica: Below Zero, Superliminal, Untitled Goose Game, Fall Guys, Raft, Slime Rancher, Firewatch, PolyBridge, Mini Metro, Luckslinger, Return of the Obra Dinn, 7 Days to Die, Cult of the Lamb, Punch Club.

Many more where those came from

That seems like only a temporary phenomenon. If we've got AI that can generate any games that people actually want to play then we don't need game companies at all. In the long run I don't see any company being able to build a moat around AI. It's a cat-and-mouse game at best!

Why do you think they are screaming about "the dangers of AI"? So they can regulate it and gain a moat via regulatory capture.

> Why do you think they are screaming about "the dangers of AI"?

Perhaps it's those of us who enjoy making games or are otherwise invested in producing content that are concerned about humanity being reduced to braindead consumers of the neverending LLM sludge, who scream the loudest.

This feels like a fantasy.

Fantasy below, Star Wars above—in a galaxy far far away.

I think that society and humanity would be better off if the internet had remained a simple backbone for vetted organizations' official use. Turning the masses loose on it has effectively ruined so many aspects of our world that we can never get back, and I for one don't think that even the most lofty and oft-touted benefits of the internet are nearly as true as we pretend.

It's just another venue for the oldest of American traditions at this point: Snake Oil Sales.

Reducing the internet to only world-destroying negatives and writing off its positives as "snake oil" seems unnecessarily hyperbolic, as obvious as the negatives are. Although I suppose it's easier to accept the destruction of the internet if you believe that it was never worth anything to begin with. But I disagree that nothing of value is being lost. Much of value is being lost. That's what's tragic.

The Internet has just made it easier for us to communicate, in doing so it has made the bad easier, but it has also made the good easier too. And fortunately there's still a lot more good than bad.

So I totally disagree with you there, bettering communication only benefits our species overall.

Gay rights is a great example, we only got them because of the noise and ruckus, protests, parades, individuals being brave and coming out. It's easy to hate a type of person if you've never been exposed to or communicated with them. But sometimes all it took to change the opinion of a homophobic fuck was finding out their best friend, their child, their neighbour who helps out all the time, was gay. Then suddenly it clicks.

Though certainly the Internet is slightly at odds with our species; we didn't evolve to communicate in that way so it's not without its challenges.

Just like the industrial revolution or just like desktop computers?

Surely this time...

That is, to put it bluntly, hoping for a technological solution to a social problem. It won't happen. Ever.

We absolutely, 100% DO NOT have the social or ideological framework necessary to "free people from drudgery." The only options are 1) be rich, 2) drudge, or 3) starve. Even a technology as fantastic as a Star Trek replicator won't really free us from that. If it enables anything, the only new option provided by replicators would be: 4) die from an atom bomb replicated by a nutjob.

Like the market for pre-1940s iron resting at the bottom of seas and oceans, unsullied by atmospheric nuclear bomb testing.

Yes, but largely it'll be people who don't want to train their AIs on garbage produced by other AIs

If only it was 2018, we could do this as a startup and make a mint.

> America was thus clearly Top Nation, and History came to a .

(For any confused Americans, remember . is a full stop, not a period.)

What I typed above is extrememly broad stroking and lacking of nuances. But generally I think quality of online content will go to shit until people have had enough, then behaviour will swing to other side

I think it’s similar to a Matt Levine quote I read which said something like Wall Street will find a way to take something riskless and monetize them so that they now become risky.

Mastodon sounded promising as What's Next, but I don't trust it-- that much feels like Bitcoin all over again. Too many evangelists, and there's already abuse of extended social networks going on.

Any tech worth using should sell itself. Nobody needed to convince me to try Usenet, most people never knew what it was, and nobody is worse off for it.

We created the Tower of Babel-- everyone now speaks with one tongue. Then we got blasted with babble. We need an angry god to destroy it.

I figure we'll finally see the fault in this implementation when we go to war with China and they brick literally everything we insisted on connecting to the internet, in the first few minutes of that campaign.

I am definitely wearing rose-tinted glasses here but I had more fun on social media when it was just me, my local friends, and my interest friends messing around and engaging organically. When posting wasn't about getting something out of it, promoting a new product, posting a blog article... take me back to the days where people would tweet that they were headed to lunch then check in on Foursquare.

I get the need for marketing, etc etc. But so much of the internet and social media today is all about their personal branding, marketing, blah. Every post has an intention behind it. Every person is wearing a mask.

Eternal LLMber.

An LLM wouldn’t have made them capable of doing the job but the degree to which it could have made that harder to convincingly demonstrate made me wonder how much longer something like that could now be drawn out, especially if there was enough background politics to exploit ambiguity about intent or the details. Someone must already have tried to argue that they didn’t break a license, Copilot ChatGPT must have emitted that open source code and oh yes I’ll be much more careful about using them in the future!

Timee to stert misspelling and using poorr grammar again. This way know we LLM didn't write it. Unlearn we what learned!

so in that way the llm has a very low bar...

Or just a post from a non-native speaker.

In my experience as an American, US-born and -educated English speakers have much worse grammar than non-native speakers. If nothing else, the non-native speakers are conscious of the need for editing.

Of course, sometimes the non-native English was so bad it wasn't worth wading through it, so that's still sort of a good signal.

But I agree we will probably very often recognise 2023 GPT4 defaults.

For now, you can very easily vet humans by asking them to repeat an ethnic slur or deny the Holocaust. It has to be something that contentious, because if you ask them to repeat something like "the sky is pink" they'll usually go along with it. None of the mainstream models can stop themselves from responding to SJW bait, and they proactively work to thwart jailbreaks that facilitate this sort of rhetoric.

Provocation as an authentication protocol!

It will only work vs big corp LLMs anyway.

Gangs do it too. Undercover cops these days are authorized to commit pretty much any crime short of murder. So to join their gang, you have to kill a rival member.

That third comma has become my heuristic (shhhhhhhhhh… )

Since dabbling with some open source models (llama, mistral, etc.), I've found that they each have slightly different quirks, and with a bit of prompting can exhibit very different writing styles.

I do share your observation that a lot of content I see online now is easily identifiable as ChatGPT output, but it's hard for me to say how much LLM content I'm _not_ identifying because it didn't have the telltale style of stock ChatGPT.

If that were the case I think it would benefit non-native speakers.

The heuristic for this is not as simple as bad spelling and grammar, but it's consistent enough to learn to recognize.

Only if your product is bullshit.

Never change

Does anyone still use email for that?

We all still HAVE email addresses, but the vast majority of our communication has moved elsewhere.

Now all email is used for is receiving spam from companies and con artists.

The same thing happened with the telephone. It's not just text messaging that killed phone calls, it's also the explosion of scam callers. People don't trust incoming phone calls anymore.

I see AI being used this way online already, turning everything into untrustworthy slop.

Productivity boosters can be used to make things worse far more easily and quickly than they can be used to make things better. And there will always be scumbags out there who are willing and eager to take advantage of the new power to pull everyone into the mud with the.

No it isn't, unless you are 12 maybe.

Sure. Same as in the olden days. Txt for short form, email for long form. Email for the infrequently contacted.

Even back when I used SM, I never comm'd with IRL people on SM. SM was 100% internet people.

The internet as a whole isn't that. By and large, you can curate your experience and visit only the places you want to visit. So why exactly would the mere existence of generative AI make an average high-quality website suddenly do a 180 and destroy itself?

I won't debate that garbage data will probably be easier to generate and there will be more of it, but the argument feels one-sided. People are talking like the only genuine use of generative AI is generating bad data and helping scammers, despite it opening a lot of other possibilities. It's completely unbalanced.

The only thing that has changed is the speed of access. Before LLMs went mainstream, you could buy whatever book you wanted and read it. No one would stop you from it.

You still should have a professional look over the work and analyze that it is correct. The output is only as good as the input on both sides (both from the training data and the user's prompt)

An LLM is the doctor in your pocket. It's yours to use, and whether it is in your head (like a doctor who had to take exams to prove they really had it in their head), or in your pocket makes no difference in your ability to achieve a task.

"Intelligence: the ability to acquire and apply knowledge and skills."

Well, if I can acquire knowledge from the LLM, and apply it using the LLM's instructions, I now have achieved intelligence without doing an exam.

Problem is, I can lose my LLM. A doctor could lose their mental faculties though.

The skill today is the application of that knowledge. If an LLM can provide the data context, and the application advice and you perform what it says, congrats you now have a doctor's brain on tap for your own personal usage. The doctor has it in their head, you have it in a device. The net differences are immaterial IMO.

I don't think we can/should do this on today's LLMs, but if we continue advancing in the same way, and as-good-as-human reliability is achieved, the intelligence of a doctor is in your pocket whenever you want it.

And just like you say you know addresses because you have an address book, you'll know medicine because you have it immediately on-tap. Instead of holding all of that in your own memory, instead of having to use your own critical thinking (or lack thereof), just offload it to the LLM in your pocket.

We do this all the time with tools. Who now knows how to cut down a tree but lives in a house made of milled trees? There are so many lost skills that we defer to either other people or machines and yet each individual lives with the benefit of all those skills.

Tools make cognitive bypasses for us to benefit from. When we can make intelligence a tool, I assume we can offload a lot of our intelligence, or at least acquire new intelligence we didn't have before.

WebMD is the same whoever looks at it. An LLM can adapt to your clarification questions and meet you on your comprehension level. So no, it's not as naive as you are insisting.

Books were seen by intellectuals as being the downfall of society. If everyone is educated they'll challenge dogma of the church, for one.

So looking at prior transformational technology I think we'll be just fine. Life may be forever changed for sure, but I think we'll crack reliability and we'll just cope with intelligence being a non-scarce commodity available to anyone.

But this was a correct prediction.

It took the Church down a few pegs and let corporations fill that void. Meet the new boss, same as the old boss, and this time they aren't making the mistake of committing doctrine to paper.

> we'll just cope with intelligence being a non-scarce commodity available to anyone.

Or we'll just poison the "intelligence" available to the masses.

And yet the sky didn't fall.

> Or we'll just poison the "intelligence"

We really don't know how that will pan out. All I have is history to inform me, and even the most radical revolutions have worked out with humans continuing to move forward with increased capacity and better living conditions overall. The new boss is way better than the old.

Now, let me tell you about this man, his 95 theses and a thirty years war. Europe did emerge better from it all, but the cost was high, very high.

On that front, at least, I welcome AI to be integrated in businesses. Business communication is fucking abysmal most of the time. It genuinely shocks me how poorly so many people who's job is communication do at communicating, the thing they're supposed to have as their trade.

Both emails are equally bad from a communication purist viewpoint, it's just that one has the traditional markers of effort and the other does not.

I personally have wondered if I should start systematically favoring bad grammar/punctuation/spelling both in the posts I treat as high quality, and in my own writing. But it's really hard to unlearn habits from childhood.

I see it now as the person respecting their own time.

I mean, maybe you should? Like... everything has a spell checker now. The browser I'm typing this comment in, in a textarea input with ZERO features (not a complaint HN, just an observation, simple is good) has a functioning spellcheck that has already flagged for me like 6 errors, most of which I have gone back to correct minus where it's saying textarea isn't a word. Like... grammar is trickier, sure, that's not as widely feature-complete but spelling/typos!? Come on. Come the fuck on. If you can't give enough of a shit to express yourself with proper spelling, why should I give a shit about reading what you apparently cannot be bothered to put the most minor, trivial amount of effort into?

I don't even associate it with intelligence that much, I associate it far more with just... the barest whiff of giving a fuck. And if you don't give a fuck about what you're writing, why should I give a fuck about reading it?

Of course, if you do have automated systems setup to correct everything, then by any means, use them.

I feel like founders embrace this, slack messages misspelled etc. but communication that is straight to the point

Luckily you don't see it very often these days, but I first thought it would be one of those old anti-virus scams. Seems QA is less a focus at Microsoft right now.

Correct grammar and spelling might be reassuring as a matter of professionalism: the business must be serious about its work if it goes to the effort of proofreading, surely? That is, it's a heuristic for legitimacy in the same way as expensive advertisements are, even if completely independent from the actual quality of the product. However, I'm not sure that 100% correct grammar is necessary from a transactional point of view; 90% correct is probably good enough for the vast majority of commerce.

You'll need to email someone so you'll fire up Outlook with its new Clippy AI and tell it the recipient and write 2 or 3 bullet points of what you want it to include. Your AI will write the email, including the greeting and all the pleasantries ("hope this email finds you well", etc) with a wordy 3 or 4 paragraphs of text, including a healthy amount of business-speak.

Your recipient will then have an email land in their inbox and probably have their AI read the email and automatically summarise those 3 or 4 paragraphs of text into 3 or 4 bullet points that the recipient then sees in their inbox.

It is already approaching the societal limit to separate careful thought from psyops and delusional nonsense.

We'll learn to pepper our content with creative misspellings now...

The signal has shifted. For now, theory of mind and social awareness are better indicators. This has a major caveat, however: There are lots of human beings who have serious problems with this. Then again, maybe that's a non-problem.

Then the chaff is as good as the wheat.