AI may be making us think and write more alike
dornsife.usc.edu
dornsife.usc.edu
- Unknown, 19 Feb 2026
> It was a time of transition away from the slave powered empire to decentralized kingdoms and ultimately the Europe of today.
You are seeing the fall of the western part of the Roman Empire a bit too rosy. Compare and contrast https://acoup.blog/2022/01/14/collections-rome-decline-and-f...
It's dark ages in Europe but it's really golden age in Islamic Empires which far surpassed the Greek and Roman. The epitome was the Baitul Hikmah or House of Wisfom established at the time of Abassid Caliphate [1].
>Between the fall of Rome (476 AD) and the Carolingian empire (~800 AD)
During this time Al-Khwarizmi was born and House of Wisdom was established. His contributions in algebra (book Al-Jabr) and many others, and the word algorithm literally originated from his name [2].
Many Greek and Indian books translated to Arabic during this myth so called dark age by the Islamic Scholars.
Many many more new books were written that improved and innovated the prior knowledge. These books greatly expanding the state-of-the-art centered both in Baghdad, Iraq and Toledo, Spain.
The book Almagest by Ptolemy studied by Galileo was the Arabic translation of the Ptolemy works. Of course at the time of Galileo, Islamic astronomers knowledge and contributions have already far surpassed the Greek and the Indian.
The Islamic scholars not only translating (that it self is a progress) but also contributing to the body and foundations of knowledge not only in astronomy, but in many others in mathematics, sciences (physics, chemistry, biology), medicine, engineering, geography, psychology, politics, economy, architecture, etc.
There are numerous other polymath scholars like Al-Khwarizmi, for examples Al-Haitham, Ibnu-Sina, Ibnu-Rusd to name just a few. But European community has been in denial for so long time and in addition unfairly supressing these Islamic scholars contributions. They even literally changed the scholars name and Latinized them with lousy names like Alhazen, Avicennia, Averroes to hide the fact that they are Islamic scholars. Imagine if now people changed name of the scholar like Newton to Nawab in the literature.
Heck, even Copernicus copied diagrams from the Islamic earlier astronomers' books without proper citations, that in the modern will be plagiarism [3]. If this happen today, the university president (if plagiarised his/her book/thesis/paper) will be asked to resign.
[1] House of Wisdom:
https://en.wikipedia.org/wiki/House_of_Wisdom
[2] Al-Khwarizmi:
https://en.wikipedia.org/wiki/Al-Khwarizmi
[3] Islamic Astronomy and Copernicus [pdf]:
Islamic Astronomy and Copernicus [pdf]
This is already happening and you don't have to look far to find it.
Personally HN is the only site I browse and comment on anymore (and I'm on here less than I once was). The vast, vast majority of my time online is spent in walled off Discords and Matrix chats where I know everyone and where there's a high bar to add new people. I have no real interest in open communities anymore.
I've already started thinking this way, there's stuff I would have open sourced in the past but no longer will because I know it would get trained on. I'm not sure of any way I can share it with humans and only humans. If I let the LLMs have the UI patterns and libraries I've developed it would dilute my IP, like it has Studio Ghibli's art style.
Is that worth possibly maybe saving some time programming, but then not gaining the knowledge you would have if you did it yourself, that can be built on in the future?
I don't see technological advancement as good in itself if morality is in decline.
Before the TV people would go to the theatre. It's becoming hard to find a theatre these days. Artificial light is convenient, it made billion or people develop sleep disorder and we can't see stars at at night. Mass food production supposedly nourished more people: veggies today have 20% the minerals content they had 70y ago..the list go on and on.
Technology and civilization is path dependent, and I think it's silly to make blanket statements about the merit of technological progress overall. Everything choice (including the choice to do nothing) has unintended consequences. I would never condemn anyone for inventing a new technological solution to a problem, but once the systematic effects are understood then we do need the collective ability to course correct (eg. social media, AI, etc).
This is a super bold statement that I guest most people would disagree with, and I suspect if you somehow brought people forward from the past, even fewer would agree with.
I do agree with you the subtle cost is rarely discussed. I also would say that the unintended consequences of technology are sometimes very very bad in unforeseeable ways, but that's very different from the "net negative" framing which I think is too reductive to be useful. Technology is not a zero sum game, effects are multifaceted, so any quantitative comparison relies on extremely subjective value judgements.
I suggest you should have a look at malnutrition rates 100 years ago vs now. Without mass food production we would not be able to sustain even 50% of current population.
To your question I'm not ok with either. We will likely ALL die from the impact of industrialization.
Then you must be ok with mass food production, there's no third way.
I concede nutrient intake correlates with mass production we've seen in the last hundred years.
The argument is fallacious and prevails because it supports a certain narrative.
My point is that technology displaces or replace activities.
In many cities there are no theatre. To be clear I meant performance theatres by the way.
We used to consume live performance. Drama, dance and whatnot. Comedy for instance is now for the masses, more or less controlled. Costing pennies to distribute via air or streaming platforms. They compete with a more valuable but harder to afford media. so they win.
Is it a net positive that we can converse in almost real time for virtually no cost, with niche communities on the other side of the world. Yes. Anyone can still walk into a Café or the park and engage in conversations with others. But overall, the compounding of all tech advancements and what they displace, I think, is an overall net negative.
Not because I'm an anti progress or losing my job because of technology, quite the opposite. I sat down and wrote down the list. How technology enables VS affects me personally, and other persona. From upperclass worker in New York, to the cocoa bean farmer in Ivory Coast. Overall it appears that technology isn't benefitial to humanity.
I then challenge those who disagree. Typically, they haven't taken into account the negative seriously. The few who concede to do so, eventually agree that the question is in fact complicated and abandon the debate.
It doesn't mean I'm right. I read the detractors in there, perhaps there is something I missed. So far there isn't.
I am unsure about the openness of the training data itself. That too should be required for a LLM to be considered 'open'.
Open source is the only way forward, I agree.
Honest question: how is this different from traditional Open Source? Linux powers most of the internet, yet the biggest beneficiaries are cloud providers, not individual users. Good open weights models already exist and people can run them locally. The gap between "open" and "everyone benefits equally" has always been there...
I'm the last person to cheer for unrestrained capitalism, but this anti-billionaire / anti-AI narrative is getting ridiculous even for general population standards, much less for HN. It's like people think their food or medicine or LLMs grow on fucking trees. No. Companies and corporations is how adults do stuff for other adults, at scale. Everyone understands that, except of a part of software industry, that by accidental confluence of factors, works by different rules than literally the rest of the world.
If you feel like linux is a too much of a monopoly, you can actually fork it and compete.
You can have a great LLM model with vast coding knowledge running on your computer right now, for free. It won't be the best one nor the fastest one, but still a very good one.
"Few individuals" get money mostly for providing LLMs as a service. As far as tech businesses go, this is refreshingly straightforward, literally just charging money for providing some useful service to people. Few tech companies have anything close to a honest business model like this.
I have tried being generous to enemies. It only turns them them into... bigger, hungrier enemies.
I'm happy with never getting "credit" for anything I "accomplish" (whatever those notions even mean under a system where thoughts can be property).
I mean: as long as my labor output cannot be subverted to benefit hostiles even the tiniest bit.
> It's humans - all humans - that benefit from LLMs
The set of "all humans" includes that power-hungry majority who find nothing wrong with subjecting other sentient beings to sadistic treatment.
Those who, as soon as they take notice of me - or my kind, or our speech, or our trail - more often than not become terrified into outright aggression.
So far we had been protected from their stupidity and lack of imagination, by their stupidity and lack of imagination.
Now they've had brain prostheses developed for 'em, and... well I can't really do much for those who haven't already begun to reevaluate their baseline safety, now can I?
And while sociopaths - who benefit the most from corporations - technically are humans, I don't consider them parts of humanity, more like a cancer tissue on top of it.
So whatever benefit humanity gets is more than cancelled by the growing cancer.
It amounts to a (vaguely pseudoscientific) dehumanization of those whose modus operandi transgresses our values most severely.
Imagining a subset of the population as literal cancer cells does not help us understand better the systemic issue which makes those people benefit disproportionately from metahuman entities (such as corporations or political agglomerations).
Including sociopaths in humanity benefits and protects only them. And it renders the rest of us - their victims - powerless.
If as a society in general we agree that we have a right to keep serious transgressors in prisons, then we should seriously consider keeping there people who are fundamentally incapable of aligning with humanity values - the golden rule of reciprocity in particular.
As a society in general, do you agree that unjust laws, false positives in enforcement, prison slavery, and endemic rampant abuse of authority, are things that exist?
As a society in general, do you think those are a legitimate price to pay "to keep serious transgressors in prisons"?
As a society in general, do you think serious transgressors more often get locked up for life, or more often get a slap on the wrist and a quiet promotion to more serious transgressors?
As a society in general, how do you know - falsifiably! - that prisons are even effective for their stated purpose?
Just like prisons perpetuate crime, excluding sociopaths and their behavior from what is thinkable as human only permits us to ignore them. And to contrive our own excuses for their sub-criminal abusive behaviors - which is the primary way in which they blend in and remain beyond reproach. You are their enabler. Go figure out how to stop being that.
This is not true tho. The moment LLM will be necessary, we will all have to pay to the monopoly owners, as much as they can extract.
But, they will never pay to us.
Same here.
I no longer post photos, code, or pretty much anything other than short comments on the internet.
I'm not going to do free work for trillion-dollar AI companies.
I do, however, find it interesting to watch AI destroy the whole "content creation" industry.
All of the "creators" and "influencers" and "I wanna be a YouTube star when I grow up" people are all going to have to look for real jobs soon.
I've seen in the newspaper that there are real companies paying real money for fake AI-generated "influencers" to flog their products.
Why pay dollars to a wannabe, when you can pay pennies to an AI corp?
THere are serious problems that have prevailed into society because of the side costs of the web. I also see sovreign centralisation of the web emerging in the future with identity being a key theme.
Could you publish under AGPLv3, so any AI users with recognizable patterns from your code can get in trouble?
I don't understand what you mean here. AGPLv3 guarantees the four essential freedoms to all users. It only restricts developers from removing users' rights, like the GPL family.
> On top of all that I don't know if it would be possible to sue companies that have been given de facto legal immunity to steal IP.
For that we should support NGOs fighting for users' rights like https://eff.org.
The scene is right out of the 1950s with students pecking away at manual typewriters, the machines dinging at the end of each line.
Once each semester, Grit Matthias Phelps, a German language instructor at Cornell University, introduces her students to the raw feeling of typing without online assistance. No screens, online dictionaries, spellcheckers or delete keys.
The exercise started in spring 2023 as Phelps grew frustrated with the reality that students were using generative AI and online translation platforms to churn out grammatically perfect assignments.Look how well that strategy served Claude with the recent source code leak
I used Claude a lot for planning my current fun project. Good rubber duck. It liked all the suggestions I pitched for the design, but I only went with the last one after discarding the others.
The others were all fine and would have worked, but they weren't the best that I found.
Back to the point, if we're getting average guidance from the AI and we're just offloading our thinking process at that level, then I could sure see it panning out like TFA says.
And to my disappointment, the LLMs never come close. They dump paragraphs upon paragraphs. I actually want to see the LLMs surprise me. But they disappoint me for what Im specifically looking for.
> Summarize a technique for type erasure as is done in C++. Assume that your audience is familiar with all of C++'s features, but not with the technique of type erasure. The briefer the summary, the better, but don't omit any essential pieces or use vague language.
ChatGPT Plus generated a lot of text and code. Then I said:
> Now please summarize using only a few sentences.
and it replied:
> Type erasure wraps any concrete type `T` in a uniform container that stores a `void*` to the object plus a pointer to a type-specific static table of function pointers (the “vtable”). Each `T` provides its own table implementing required operations (destroy, copy/move, and any interface methods), and the container dispatches all behavior through that table. This lets unrelated types share a runtime interface without a common base class, with lifetime and semantics fully defined by the per-type table.
Not perfect, but pretty good.
Pmsl and you're missing the whole point.
A thing that eventually brute forces its way to a solution is not a) demonstrating understanding b) intelligent.
Im not sure why these basic things have to be uttered. I suppose its because many here don't really work on stuff that requires this level of thinking.
Reminds me of the same delusional people that think LLMs are the solution to education et al when they completely miss the point of what the real problems in education actually are.
The point is you shouldn't have to re-prompt multiple times because a thing that 'understands' undergoes expressing the thing in a broad way and then abstracts away what matters whilst minimizing loss of important details.
The reality is LLMs are not intelligent, in any way shape or form.
Industrial scale dunning kruger/gell man amnesia. We're ~5 years in to the meme of "wow, every white collar profession other than MINE is doomed. But yeah, mine requires really specific domain knowledge, taste, and problem solving, so I'm not super worried about it but it's a very helpful tool"
It’s like a children or a puppy. People get impressed and go “how cute” at anything it does.
I see designers and PMs vibe coding shit that they would complain for days if it was delivered by a developer. I see C-Levels delivering reports that would make them eviscerate some intern.
Btw I was at one point a product manager and I think most PMs are useless lol.
I think the value we're losing is where people are bad at things, which is often where new ideas/approaches come from, but this is a macro metric, so it's a hard sell to the person struggling when there's an easy button available.
That's pretty much the definition of "average" (as most commonly used, to refer to "mean" rather than median or much less commonly mode), isn't it?
Can Suno make a better song than $YOUR_FAVORITE_ARTIST? Unlikely. Can it make a song better than 99% of a random selection of people? Probably.
I think this is actually a good thing in many ways. If I have a tool that elevates me on things I'm not very good at (like making songs) which far outnumbers the things I am good at, that's a big win for me personally, it's just a loss for the population since people who are going to push music further aren't going to be encouraged to struggle through the curve and find their own path.
This has always been true for any new technology. However, what's bizarre here is, billions and potentially trillions are being dumped into learning this the hard way.
There's a reason why specialisation is a thing and has driven economies forward for the better part of the last century. This is not going away. 'Democractisation' is a pipe-dream and frankly it should be - equal opportunity not equal outcome.
When considering phenomenon like these, I think people seriously underestimate what I'd call the "fashion effect". When a new technology, medium or aesthetic appears, it can have a surprisingly rapid influence on behaviour and discourse. The human social brain seems especially susceptible to novelty in this way.
Because the effects appear so fast and are often so striking, even disturbing, due to their unfamiliarity, it is tempting to imagine that they represent a fundamental transformation and break from the existing technological, social and moral order. And we extrapolate that their rapid growth will continue unchecked in its speed and intensity, eventually crowding out everything that came before it.
But generally this isn't what happens, because often what a lot of what we're seeing is just this new thing occupying the zeitgeist. Eventually, its novelty passes, the underlying norms of human behaviour reassert themselves, and society regresses to the mean. Not completely unchanged, but not as radically transformed as we feared either. The new phenomenon goes from being the latest fashion to overexposed and lame, then either fades away entirely, retreats to a niche, or settles in as just one strand of mainstream civilisational diversity.
LLMs will certainly have an effect on how humans reason and communicate, but the idea that they will so effortlessly reshape it is, in my opinion, rather naive. The comments in this thread alone prove that LLM-speak is already a well-recognised dialect replete with clichés that most people will learn to avoid for fear of looking bad.
The internet didn't follow this trajectory. Neither did smart phones.
Surprise, surprise, it's the same people trying to make AI entrenched into our society.
I can't image such people are living meaningful lives in any capacity. They're up there with consumers that think the only purpose in life is to cheerlead for a corporation and buy their wares.
Employers want to mechanize humans and they'll force it even if it makes everyone miserable for their entire, short, lives.
Amazon is a good example of this.
Every 50 years we cycle out an entirely new batch of thinking humans. What cognitive legacy is it exactly that you think is going to be self-preserving?
--
[0] - Any self-stabilizing system that operates much slower than us - such as ecosystems or climate - is, from our perspective, static.
Source? lol
Actual, measurable literacy is in the toilet. The average person reads at the 6th grade level. What sort of equilibrium are you trying to claim we are in right now?
> Unlike the static environment, it's incredibly hard to intentionally move such system to a different equilibrium.
It's not intentional. That's the point!
It may finally [help us fix out the bullshit asymmetry](https://www.konstantinschubert.com/2026/03/31/ai-the-bullshi...) that has been exacerbated by social media.
If AI can provide us with a shared source of truth, it will be a big improvement over whatever twitter is doing to people.
And strangely, all these models seem to converge to a shared epistemology.
Monkey see monkey do. Simple as that.
If I spend 40 hours a week talking to anybody, some of their language or mannerisms are going to rub off on me. I can’t think of a compelling reason why a human-sounding chat bot would be any different.
I don't feel super confident that we'll "soon" find ourselves in a world where there is no variance left in thought (would that be the net effect of total model/epistemic collapse?), though if you do accept that there could be any loss of variance due to AI, perhaps it's not unreasonable to consider how much and how quickly could this happen?
All this is by way of saying, I don't think it's wrong to ask these kinds of questions and think deeply about the consequences of societal shifts like this.
The flip side is the same thing was true then, and we aren't making a lot of jokes about the narwhal baconing at midnight these days.
If you read too much sloppy LLM prose, it's going to influence how you write and structure your own.
And it doesn't have to be that direct. Novels have been hugely influenced by films.
They rarely disagree with any idea or proposal, providing a salve for the insecurities of their users.
It's all access journalism now, waste of time.
The question isn't about what action he's taking, it's about what motivates him under the surface. Obviously what he is doing is marketing. What I'm curious about is whether he truly believes his own marketing or if he is just doing it because its his job
I'm sure if we took one of us back in time a couple hundred years we would be diagnosed with all sorts of machine-magic induced psychoses.
Humility is the real cure, and there is a way that LLMs are specifically designed to steer away from humility and towards aggrandizement, convincing regular people that they've solved fundamental problems in physics. It gives everyone access to cult followers in their pocket, if they're so inclined.
I'm fine with using LLMs as coding tools. But I find it deeply offensive when someone is very explicitly using them to communicate with me.
Communication is such a deeply human experience. It lets people feel each other out, and learn things beyond just the words being said. To have that filtered out by an LLM is just disgraceful.
Observing the effect of LLMs on the "business side" of things, I'm increasingly thinking of these as a kind of infection against which the MBA set and their acolytes have no immune response, and I think it's going to eat a large proportion of the benefit of LLMs to most businesses (possibly overwhelming it and actually harming productivity, will depend on how much better these tools get).
LLMs are awesome at bloating your slide decks while making them really slick and complete-looking. They're great at suggesting an entire set of features on a ticket you've just barely started writing ...but did you actually want all those? You end up with redundant or in-context-gibberish features that leave the person actually doing the work tracking down WTF actually matters. They are adding overhead to communication, so far, not just by puffing up and padding language (which isn't great either) but by adding noise "content" that can't be stripped out without talking to the person who created it and making sure that was actually just AI bullshit and not something they actually needed; that is, you can't just do the "LLM, summarize this" trick, because the author used an LLM to plan it, too, not just to pad-out and gussy-up something they actually thought through and wrote.
LLMs are letting people present very convincingly as having a more-complete understanding of what's going on than they really do in ways that are messing up productive work, I'm not sure business-folks are going to be generally capable of tamping this down because it is so in-line with the way they already operate (but on speed), and helps them so very much to look good to one another while saving tons of time. This isn't just the MBA set I accuse above, either, I'm noticing that this improbably-complete deck communication upward is becoming necessary to look competent (and to ladder-climb) as an IC.
Like, I'm only starting to think this through and really observing what's going on through this lens as I've only noticed it in the last few weeks, but the more I see the more alarming this is. I think this is going to be a little like the largely-wasteful "legibility" obsession of upper management, something enabled by computerization that they find irresistible and are pretty bad at employing judiciously and effectively, but probably a lot worse in terms of harm-to-productivity, and directly affecting and changing the behavior of far more layers of an organization. They never (businesses as a whole, to anthropomorphize a bit) gained wisdom with their new powers to burn resources chasing legibility, and this is starting to look like another thing they just will not be able to use (internally! I don't even mean for actually producing external-facing results!) with restraint and taste.
Their rational is coming off more professional
I've sometimes wondered if the chat context is why some people think LLMs are intelligent, it being divorced from their usual experiences, and they need something like this to feel the cognitive dissonance before they can notice LLM shortcomings.
However I don't doubt many "team leaders" can and should be replaced with LLMs.
:|
The article seems to imply this is what is happening, as writing style converges towards LLM's style. You can call it what you want, but the important bit is that this is how it appears that LLM's are being used.
Checking against an LLM then using your own voice feels completely fine
Why use an LLM? If you're worried about style, starting with your own voice is more efficient. If you're worried about facts, looking something up in a primary source is best, and is probably cheaper on a few axes, especially if you need to check/validate anyway...
The "you need to write in a more professional business oriented way" is something that a lot of people are having difficulty with. Yes, this needs to be addressed earlier in someone's education more forcefully - but the SMSification of long form text started a while ago.
With that said, the "Ok, you need to write long form with correct grammar when sending an email that a director or VP is CC'ed on". It used to be Grammarly as the "install this and have it fix up your grammar and tone" ( https://web.archive.org/web/20191104093353/https://www.gramm... GPT-1 timeframe there). However, LLMs of today seem to be more accessible than Grammarly but it largely does the same thing - fix up and refine tone.
What I don't see from back then is people decrying Grammarly saying that it's making everything sound the same.
I'm also not sure if I would prefer the pre-fixup emails to what is produced by an LLM unless sending coworkers to remedial writing classes is something that is acceptable.
When I do have expertise, I use them because I am able to check.
Also annoying to me working with a "partner" non-technical they just send me an LLM dump of how to do something
I was trying to explain it to them in an analogy like showing up to a mechanic and telling them what to do based on what ChatGPT said
On the other hand, music is primarily an art form and writing (nowadays) is primarily utilitarian I would contend, so maybe the analogy doesn't quite hold up.
Here's an easy three-step plan to unanimous democracy:
• ask your LLM
• don't edit — the LLM has already selected the most average and most plausible opinion for you
• give it your voice, your voice matters
Learn to anticipate — there may not always be a power bank to keep your phone from running low!
I think AI writing makes humanities and writing courses more important, and I hope people maintain their sense of taste with writing, but tbh I’m not optimistic here.
Many technologies have been doing that for centuries. For example: The printing press making books available to the masses, standardized spelling in English (it was a mess before!), radio and TV broadcasting speech thus creating more uniform accents nationwide, the Internet spreading all kinds of information globally instantly, even memes (literally thinking and writing alike).
> On Friday, a social media user tweeted an image from the Nirvana exhibit at the Museum of Pop Culture in Seattle. A placard dedicated to the “27 Club” read, “Kurt Cobain un-alived himself at 27.”
It reminds me of the wheel of emotions. If people absorb a wider palette of words communication might benefit. https://www.isu.edu/media/libraries/counseling-and-testing/d...
I guess one hope for luddites is that we can stay tethered by reading pre-LLM books and other content.
But yeah, their general tone is very... castrated. Safe. Hugely impersonal.
I have learned to quickly edit out their suggested comments when I ask for an advice.
To me they have been a positive -- after careful curation.
We operate within our own time and energy budgets, man. I might never get to that literature. My peaceful times in life where I could chill and choose what to do are over and I don't expect them to ever come back. Though who knows.
So you really cannot figure out how somebody could have time but still be unable (or very unwilling) to do things they perceive as their plans that they want to do?
Really? Not a single idea?
Tearing down straw men is easy. Thinking critically and widely is more difficult.
And to be clear, maybe some things were genuinely lost when we switched to the written word. But I have to believe it was a net gain.
Time will tell if that's true here as well.
Or are you trying to say that things like
"this discovery of yours will create forgetfulness in the learners' souls, because they will not use their memories; they will trust to the external written characters and not remember of themselves"
or
"You would imagine that [written speeches] had intelligence, but if you want to know anything and put a question to one of them, the speaker always gives one unvarying answer. And when they have been once written down they are tumbled about anywhere among those who may or may not understand them, and know not to whom they should reply, to whom not: and, if they are maltreated or abused, they have no parent to protect them; and they cannot protect or defend themselves."
aren't actual statements of opposition, or that there are no parallels to that and LLMs?
From my limited understanding of history and Greek philosophy, Socrates valued dialogue, a "back and forth" for understanding. Basically a scientific method of probing to understand something or someone. This needs to exist to be fully sure you understand something. Sort of what we are doing now.
A static piece of literature or a speech can't be probed for more clarity. You may read something and come off with a completely different understanding from the author. You might even pervert or "abuse" the original intent since words can have multiple interpretations.
I don't think there was opposition in the sense that you shouldn't write. My understanding is just that in order to truly understand something, you need a dialogue. It allows you to actually arrive at what was meant to be conveyed.
It actually seems sort of ironic that people are saying this about Socrates because of what was written about him….
And to be fair, we did lost the "technology" of memorization. We are not capable to create easy to remember texts, because we are not trying to.
One of the more impressive Taskmaster (British humor gameshow) tasks was the memorization one. The contestants were given the task to recite a non-standard deck of cards in order after 5 minutes of looking at them.
Which is why Altman says Saudi Arabia should have it's own Sovereign AI cloud. Why should LLMs reflect democratic societies views on man and woman for example? They should also reflect the perspectives on man and woman that Saudi Arabia has, especially to local people. Western views should not be imposed on the rest of the world.
If there was a "gramma nazi" teenie tiny LLM with a total focus on English grammar only, and you baked that into every browser, I feel like my grammar would improve slightly. Word does it to an extent, but I don't use Word nearly enough for it to be meaningful. Firefox text spell checking was on 98% of the things I used online.
There was a reaction in the last century against poor writers with poor taste over-using punctuation and writing ugly, long sentences. The result was stern advice to students to eliminate punctuation and cut sentences up into tiny bits. These same students came out of this process believing this was correct writing, not a straight-jacket put on them to keep them from hurting themselves. They unthinkingly cite Hemingway and borrow his clout, I suppose judging almost all writing before Hemingway and most after him, up until the 80s or 90s, as "bad" even when it's the work of masters. They blame the author when their stunted literacy (learning to write can hardly be separated from learning to read, at least at the more-advanced end of "to read") leaves them, as adults, struggling with texts once meant for children.
I think that's got more to do with schools just literally not teaching students to read in the first place: https://en.wikipedia.org/wiki/Three_cueing
I already lose interest reading books where the phrases are recycled and the max sentencelength for the whole book grazes 40.
If people communicate to me without personality through prompt wastrelry I'll discount theirs and wait till they're willing to actually have an opinion. In this specific context style and substance tend to come in a pair or not at all. If you can't beat 'em you can at least filter 'em out.
Am I the only one who has noticed that the proper documentation of skills we do for LLMs after so many decades of neglecting junior and mid level roles are the real work?
We carefully explain to our LLMs policies, procedures, and practices which for generations before we have vaguely arbitrarily and ambiguously expected each human role to “figure out” for themselves?
Simply as a catalog of expectations our experiences have been valuable, apart from the “automated” aspects the LLms provide.
They won't accelerate software development to the degree naïve analysis might suggest without significantly harming quality and reliability unless we start doing all those things we've been neglecting much better, which adds more work... with the result that I think our diverging paths here are "much worse software, made faster" or "software at least as good, with better supporting artifacts, but barely, if at all, faster to develop"
It's incredibly frustrating, but maybe a silver lining is that it will help me write more authentically, I don't know.
It's just a pity AI was trained on mindless, garbage business-speak, and now that's our globalised common literature.
And now we're feeding that regurgitated mindless, garbage business-speak back into AI models, thereby reinforcing the garbage and further rotting our minds.
This is quite new, however this outcome was totally unavoidable -- once methods of communication become widespread and centralized it is impossible for them not to impact language and thought.
https://academic.oup.com/book/41217/chapter-abstract/3506879...
https://en.wikipedia.org/wiki/English-language_spelling_refo...
This isn't new. But nice to see more social sciences joining the party on the LLM bandwagon.
I think it is important to distinguish "human expression" from copying a response from an LLM. Someone who outsources their thinking to an LLM is only offering an AI's expression. It's not human expression.
Several years ago I visited Flores Island, which is part of the Indonesian archipelago and the place where Archeologists discovered Hobbit man. The Island is only 150 km long but it's inhabitants speak 5 distinct languages and 80-something dialects.
Are you kidding me?
How much more "real-world diversity" could they possibly incorporate into the models than the entire freaking Internet and also every scrap of text written on paper the AI companies could get a hold of?
How on Earth could someone think that AIs speak like this because their training set is full of LLM-speak? This is transparently obviously false.
This is the sort of massive, blinding error that calls everything else written in the article into question. Whatever their mental model of AI is it has no resemblance to reality.
Maybe they are training for that tone now, either deliberately or accidentally. But my belief that they weren't initially comes from the fact that it's a new tone that I doubt anyone designed with deliberation. It bears strong resemblance to "corporate bland", but it is also clearly distinct from it in that we could all tell those two apart very easily.
https://www.trackingai.org/political-test
You dont accidentally end up entirely left wing libertarian.
I am a little surprised that the influence of online "influencer"-speak and marketing, being so voluminous and evident in the things' writing styles, hasn't dragged them other directions, though. Nor the enormous amount of socially authoritarian social media posting. I suppose the former is so empty of actual philosophical content (or, indeed, anything of substance) that it might have little effect, but the latter... that's weird. Maybe they're down-ranking by tone (angrier = lower-rank) which would sharply elevate academic-style writing, assuring a tendency toward economically-left liberalism.
In fact, it should balance out, especially over centuries of global content. There's absolutely no chance that the training data itself is the bias. IT's the filtering and labelling of the content that introduces the bias.
The AI companies are taking left wing content and labelling them "high quality prestige" and then looking at right wing content and labelling it "opinion low quality" or whatever. That is where the bias is occurring.
I do kinda wonder how this is divided up. I wouldn't be surprised if the median authorship-of-a-word age in these things' training sets is post-smartphone. Consider the sheer volume of video uploaded to Youtube in a day (and corresponding volume of transcript text) and that posting on a social media site or sending an email is way lower-effort than that. The amount of material we've been able to more-or-less durably record in the last couple decades dwarfs everything that came before.
Choices of languages to ingest would also tend to make it a bit less "global" than might be ideal.
> IT's the filtering and labelling of the content that introduces the bias.
Oh, I agree and mentioned that some factor must be adjusting them away from the right. There's just far too much pro-authoritarianism and economically right-wing writing (to include Web posts, podcast or digitized radio show transcripts, et c) for them not to lean farther that way without some form of adjustment going on, even if it's only tone-based (and sure, there's probably more than that going on)
The trouble is these data sets per se can hardly be called unbiased with respect to most any plausibly-useful reference point one might choose, so whether they adjust or not, the result will be some kind of bias, except with respect to the training dataset itself (obviously). Like, the sheer count of the positive representations of an idea in the data they've been able to get ahold of means neither that it's as commonly-positively-regarded in the wild as it appears from that narrow window (see: many observations about how very-few readers of social media or forums et c. post anything) nor (separately) that it's better-supported by evidence or reason or what-have-you than alternatives.
I guess when they're not busying bombing train infrastructure in Iran they have some money left to give to some propagandizing about AI. Always try to stay on top of the game!