logoalt Hacker News

"I can tell you wrote this yourself"

219 points • by BrunoBernardino • today at 7:13 AM • 230 comments • view on HN

Comments

FinnLobsien • today at 10:39 AM

AI and LLMs, in my opinion, are effectively the industrialization of knowledge work.

Like it or not, industrialization brought many benefits and we couldn't imagine our modern way of living without it (just think about medicines). Industrial production made so many things so much more accessible for so many people.

But an industrialist perspective doesn't care about the process of making stuff, only about the outcome. The shoemakers deep knowledge of leathers, his intuition of all sorts of different shoes, and the pride of that competence—that doesn't matter when you can make thousands of shoes a day that cost 0.5% of something artisanal. The people working at the shoe factory only carry things to and fro and occasionally fix some machine issue.

This has tons of benefits and tons of downsides. But with physical goods, a shoe is a shoe.

With knowledge work like writing, the process is often the entire point. The point of writing a paper is what you learned in the process. The point of documented software architecture is to develop an understanding deep enough to drive coherent downstream decisions. If those things are no longer true, the asset it stands for is basically worthless.

➕ show 3 replies
borzi • today at 7:59 AM

Great post, author clearly has a bright future. What I realized over the past years is that people that want to "think for themselves" were already, and probably always, in the minority. It's much easier to delegate how to react to a certain event or piece of information to someone else (news anchors, then influencers in last couple years, now increasingly AI I guess). I've certainly been guilty of this...

➕ show 2 replies
wavemode • today at 8:13 AM

> Assignment documents that start with "you are strongly encouraged to explore the use of AI" which quickly turns into the clear understanding that you aren't just encouraged but expected to use it. That you're disadvantaged by not using it. The hardest class of my degree was spent in a brick-walled windowless computer lab prompting an AI for eight hours a day.

College education for software developers has never been particularly good, but it seems like now they've actually just given up.

Do we accept this in any other field? Does your math professor teach you to solve integrals by asking WolframAlpha?

➕ show 7 replies
MrScruff • today at 9:11 AM

AI behaves like a constraint solver. You put in the right set of constraints (on writing style, on technical details, on structure) and with a sufficiently capable model what will come out is a close approximation of something you might have produced, or at least something you might have been happy to receive from someone else. There will usually be aspects that aren’t quite right, so you add more constraints.

If you have a job/life where there are more things you want/need to achieve than you have time for, this can be a significant accelerant.

It does require the ability to judge when something is good enough of course.

This is actually no different to leading teams of humans. You guide individuals and set the standards and take responsibility for the output of the team.

➕ show 2 replies
orphereus • today at 8:18 AM

When I read other comments, I come to the conclusion that it's kinda futile to post this on HN.

➕ show 2 replies
cadamsdotcom • today at 10:38 AM

It's an oversimplification to equate the leveraging of external intelligence with not thinking for yourself.

External intelligence will get you out of a ton of intellectual grunt work so you are freed to think at a higher level. You've are freed to think about different things more core to your problem, and if you wield the tools well you can choose what you offload. But it should be viewed as a tool not a shortcut...

Consider an accountant summing numbers with a calculator vs. by hand. There's way less thinking needed to get the numbers summed right. But blindly summing numbers doesn't make the work correct even if the sums are correct...

No one should put their name on work they didn't proofread, or if they shipped whatever the model pooped out on its first go. That is not your work yet. That's what we have to instil. It is OK to draft with AI! Leverage tools, sure - but know the material, make your work your own, and don't ship your first draft.

➕ show 1 reply
asimovDev • today at 8:14 AM

Interesting article. All I want to add is that I recommend everyone to look at a Tawny Frogmouth bird that is mentioned at the bottom of the post. Beautiful creature

https://en.wikipedia.org/wiki/Tawny_frogmouth

Pavilion2095 • today at 8:44 AM

> Search for answers instead of asking AI. Ask a friend. Ask your mum about the mystery illness you're convinced you're developing and then take her advice and try drinking a glass of water.

The issue that we're outsourcing too much to AI exists, but this falls short of offering a solution.

➕ show 2 replies
jwrallie • today at 9:21 AM

I work in post-graduate education, and had the same thought multiple times. It is a compliment.

It may be that your writing style is peculiar and that triggered this thought process superficially, but most likely, the structure of your text is superb in a way current LLMs cannot handle.

Either way it is a pleasure to read non-LLM generated text (in academia or otherwise).

I recently wrote a similar length report, and the hard part was not typing the words, but realizing your thought process was superficial or incomplete and needs to be refined, something you would not realize if a computer was writing it for you.

edarchis • today at 10:40 AM

"How hard is it to right 10 pages?" Maybe he just meant that it was full of mistakes ? Maybe half mistakes and half style imperfections that prove it's written by a human. I'm sure there are plenty of tools to make AI output seems like that to avoid detection. (I'm not thinking less of OP for one mistake, I do make many mistakes myself when writing, especially with English being my 3rd language)

ForgotMyUUID • today at 11:30 AM

Take it as a compliment, of course! Reading a generated text feels like a wiping a spit on one's face. One day sending an LLM-generated text will become a mauvais ton again.

> Assignment documents that start with "you are strongly encouraged to explore the use of AI" which quickly turns into the clear understanding that you aren't just encouraged but expected to use it. LLM is not an AI. IBM Watson is an AI, Prolog is AI, OWL ontologies, expert systems etc - technically any system where knowledge is encoded explicitly and unambiguously. Language model is an advanced vectorized search engine, a compression algorithm with a human-like dialogue system on top. LLM has no intelligence. Here is a simple gedanken proof of that: if we define an LLM as a "rational" system the latter means that it does not contradict itself, it has a complete non-contradicting picture of the world. We train an LLM on the whole literal corpus of humanity completely ignoring the fact that the progress is , in matter, a resolving contradictions. Having all books of humanity on the table is necessary but not sufficient for building a real intelligent system. Building such a system would imply identifying and resolving all of the contradictions, all the ignorance present in our knowledge. Calling an LLM an AI is the greatest misleading marketing trick in history.

I do voluntary teaching while working as an engineer at my job. I teach the interns because this is how I learned my skills. Except math and physics, 90% of skills I learned at the work place from elder colleagues. We have to surround ourselves with knowledgeable people who are able to take responsibility. Even if they will not stay in your company. I can't just sit still while this tech is destroying our and our children's future for the sake of smb's short-term profit. One might get all of the money of the world, but some things cannot be bought.

bongofizzbuzz • today at 7:56 AM

In "I,Robot" movie, it is frowned upon to drive the motorbike yourself as AI is more safer at high speeds.

I guess something similar will be mandated in our world, once AI crosses a threshold.

➕ show 6 replies
agileAlligator • today at 10:17 AM

If I am reading this right the author wrote the LaTeX diagrams by hand. I am equal parts impressed and horrified.

➕ show 1 reply
mirekrusin • today at 9:39 AM

Ask yourself if you'd judge model based on responses where it used different model subagents?

There is no single, universal yes or no answer – it depends on the context.

meekoAMH • today at 9:31 AM

I agree and disagree... i think in our zeitgeist we're catastrophising a little too..

it's just a tool, like a knife... useful to chop up your vegetables and eat your dinner, or to cause harm... it's not inherently negative... and blaming the tool rather than the actor as the dangerous one is probably the bigger issue here.

On using AI to replace thinking i agree, is a slippery slope and we will continue to see and be hurt by this going forward, and we need to try to contain the impact of this, but by the same token since i started using AI to develop, I've genuinely become a better developer. I can iterate and learn faster, I'm being forced to process analyze, understand things at a greater scale and more steps ahead, and able to better create systems that i believe will genuinely improve health outcomes for the general public.

So if it were me I'd probably change that final line in your blog to: "Please read and understand it, before you ask AI to do it for you"

Great post, do definitely keep writing!

➕ show 1 reply
dintech • today at 9:26 AM

As with many things in life, the absolutists always seem to get the airtime. You don’t need to 100% use AI or 100% condemn it. It’s possible to think for yourself on the fun or interesting things and also use AI for the boilerplate.

➕ show 2 replies
rootsudo • today at 8:44 AM

Right / write

➕ show 1 reply
p2detar • today at 8:43 AM

> The internet is riddled with AI content that is making it increasingly more difficult to tell fact from fiction.

This has been forseen and well explained at least 3 years ago, by people such as Maggie Appleton: https://www.youtube.com/watch?v=VXkDaDDJjoA

> Being able to think for ourselves has perhaps never been more important.

I wholesomely agree. This has always been important, even outside the age of AI.

> In a world where everything is so polarised and deliberately provoking, the ability to identify bias or motivation in the content we consume has perhaps never been more important.

No, that is simply futile. The best course of action is not to play the game. Given the video above, I believe we are already in the age of the Dark Forest of Internet, and you need to hide. Private communities, less public interaction, less reading of text on the Internet should be the recommended way to go. Maybe we'll see the rise of alternative human-only channels, but I think that won't be on the Internet as we know it. I'd be rather curious.

midas0m • today at 9:31 AM

Unfortunately, after the massive ingestion of text in any original format by for-profit LLMs, where creators’ originality is sold back as “thinking tokens,” I find it hard to encourage anyone to express their original thinking in writing, including in papers and blog posts.

estetlinus • today at 8:07 AM

> prompting an AI for eight hours a day

This is the life of a developer now. Don’t mistake understanding for memorizing all the nasty APIs you once had to.

➕ show 1 reply
maxglitchers • today at 10:05 AM

The big realisation of "Why?" rings so true here.

Cognitive offloading will be the topic of the next 100 years. It shocks me how readily people around me are willing to hand off any and all thinking to something else.

We need to be able to clearly articulate the value in friction - the "not knowing" and recontextualise llms as scaffolding, not 'the answer'.

The problem is a combination of a willingness to offload and the llms desire to output.

I've noticed my brain rewiring to disregard AI generated anything: It's a slop font, a softened sentance, a needless extra paragraph, somebodies vibed idea that they've clearly not used outside a golden path.

Intelligence is under threat and the alarm bells are not sounding loud enough.

You article is a diamond in the rough of the internet. Thank you.

atoav • today at 8:19 AM

I also write documents like that myself and have been often praised for the quality of how I structure my thinking and writing. Having studied philosophy, worked as a programmer and freelanced as a graphic designer, I have an intersection of experiences that results in clearly structured documents.

Now since AI has reached a certain quality level my worry is that what I would have written normally before comes across as AI written to the uninitiated. It started with the em-dash — a piece of typography I often used, until it came to infamy as a tell on AI-written texts.

I still think AI often makes bad taste decisions, regardless of whether it is around text or graphic design. However I have to conceded that there are not many out there whose own decisions were better. But at least they were their own. I think as a consequence people will become even more lonely and isolated, since now you can't even be sure that the words a person chose to write to you, were their own.

komali2 • today at 9:31 AM

> It's soul crushing, and I don't want to outsource so much of my thinking to it.

Been thinking about this a lot lately. For the last maybe 200,000 years, humans, as social and curious creatures, have seemed to develop our satisfaction cycles around exploring, creating, and doing useful labor that is helpful to other people, or at least recognizable to other people as interesting and hard.

This is basically how most humans have spent most of their time, throughout every economic system experimented with in the last 200,000 years. Whether dictated by the State, the Corporate Machine, or just the raw physical requirements to get calories and maintain shelter, we've been filling our time with labor, and attaching meaning to it (artificial or otherwise).

I think about "Man's Search for Meaning," and how Frankl rationalized his way to find meaning in being subjected to the worse things humans have ever done (the book was partially inspired by his surviving a nazi concentration camp). If the meaning isn't inherent to something - such as our labor - we'll create it, and rationalize it. Thus we've been able to derive at least some level of the same satisfaction from carrying a deer we killed into a village as we have when pressing the widget making button in the widget making factory. The alienation has been increasing, of course, but we work around it.

I think AI era presents us with maximum alienation. There really is no point to human participation in this labor (or soon there won't be). What's the point of doing a thing when a machine can do it better than you? Our last sliver of rationalization has been taken away from us, but our economic system hasn't caught up with that fact, so not only do we need to deal with the maximum alienation, we also need to deal with the existential dread of not being able to justify our labor in the "your labor has to have commercial value in order to eat" economic system.

I understand that optimists are saying that AI is just the new automated loom, that there'll be some job losses at first but there'll be new jobs created by the new technology and the opportunities it opens, but there's also people saying that the AI are only going to get smarter, and as they get smarter, I can't reconcile this idea that they'll somehow create new jobs for us, rather than eliminate more and more. Ironically enough, the unsolved frontier is moving into blue collar jobs, because we can't make plumbing robots yet - but there certainly are not enough plumbing jobs to go around once every white collar job is automated.

So, what do we do with ourselves then?

➕ show 2 replies
throwfaraway135 • today at 9:05 AM

I totally agree with the sentiment, but is it just me? technical 10 pages are A LOT of work, even when it's your primary focus. Writing always reminds me of the quote by Blaise Pascal.

> If i had more time i would have written a shorter letter.

mtrx • today at 9:53 AM

[dead]

knotruby671 • today at 8:33 AM

[flagged]

sambejk • today at 7:22 AM

I like the blog and fully agree that you should be able to make a report without AI usage. On the other hand, it is probably not bad to let AI check for grammar and spelling mistakes. In this blog, I can read at least one: "right" instead of "write". It could be an amazing reviewer to bounce ideas off and catch any glaring errors

➕ show 12 replies
selcuka • today at 11:53 AM

> how hard is it to _right_ 10 pages yourself

I can tell you wrote this yourself.

selcuka • today at 11:52 AM

> how hard is it to _right_ 10 pages yourself?

I can tell you wrote this yourself.

philipallstar • today at 11:39 AM

Why know things when you can use the machine? Why feel things when heroin is so much easier?

pluc • today at 11:19 AM

> At first this shocked me. Our reports are limited to 10 pages - so the appendices are getting a big workout around here - how hard is it to right 10 pages yourself?

That was unintentionally funny.

➕ show 1 reply
nashashmi • today at 9:41 AM

> Apparently the future of engineering - at least in my field - is to become a glorified babysitter.

Yeah, the more tools we get into our hands the more we incline towards glorified baby sitters. This includes calculators as well. It takes a little bit of talent to wrestle back control with new visions and challenges. And then we become micromanagers. We think the other person should follow my vision rather than I follow my people’s vision.

dingdongditchme • today at 9:19 AM

> how hard is it to right [sic] ten pages yourself?

I have read this sentiment here before: sometimes it's the mistakes make you pass as human. So the titletext could also be a sick burn.

jesse_dot_id • today at 10:57 AM

> At first this shocked me. Our reports are limited to 10 pages - so the appendices are getting a big workout around here - how hard is it to right 10 pages yourself?

Unfortunate typo

darksim905 • today at 9:51 AM

I want to like this post, but finding typos in the first few paragraphs after reading comments here where others said it was a great post makes me sad.

difosfor • today at 9:12 AM

> how hard is it to right 10 pages yourself?

Still difficult to choose the right words apparently ;)

Sorry, this made me chuckle. I sympathize with their thinking and it was well written besides this one mistake.

khalic • today at 10:51 AM

> ... how hard is it to right 10 pages...

apparently, not that easy to write

cnewey • today at 10:53 AM

From the article;

> How hard is it to right ten pages yourself?

Quite hard, apparently.

carlosjobim • today at 11:36 AM

First day of driving school: "Let's call a taxi!".

If you are a student/studying like the author of the article, there are only two goals you should have. Either to learn things, and then you have to quit the class or school if your professor tells you to use AI and go somewhere where they can teach you.

Or you're there to get a juicy diploma so you can get a juicy job, in which case by all means use AI or whatever means to get it.

I guess the third case is also if you're young and want to have sex with other young people, you should go to school.

bluerooibos • today at 9:42 AM

> how hard is it to right 10 pages yourself?

Good idea to sneak a few spelling mistakes into your writing heh.

nickm12 • today at 8:35 AM

So was it a complement or an insult? I feel like I've been left hanging.

blueblazin • today at 9:41 AM

This reads like a LinkedIn post :\

echelon_musk • today at 10:24 AM

> how hard is it to right 10 pages

I can tell you wrote this yourself.

athrowaway3z • today at 8:33 AM

I see its another entry in the long list of "use the tool for just the right amount at the right time" plea.

Its not that i disagree with the author about his worries; its just that i've read the same personal opinion & anecdote more than a dozen times, and they all seem to disagree where the line is.

'Our blessed productivity multiplier, their barbarous lack of thinking'

cisc • today at 9:17 AM

> how hard is it to right 10 pages yourself?

Seems like it's pretty hard.

TheLudd • today at 7:58 AM

> how hard is it to right 10 pages yourself?

Is this a joke?

Rodmine • today at 8:45 AM

[dead]

benzguo • today at 10:07 AM

[dead]

alex43578 • today at 8:06 AM

I don’t get this knee-jerk association of AI = lies or wrong. AI-written content can certainly make mistakes, but so can human content.

The author says “That you're disadvantaged by not using it… Apparently actually knowing things doesn't matter nearly so much as knowing how to prompt a machine to do the thing.”, which is hilarious coming from someone taking an engineering degree.

Just like how CAD and FEA have supplanted hand-drafting and manual calculations, so too will LLMs supplant hand-writing a report.

Newsflash: you will absolutely be disadvantaged by not using tools to advance and accelerate your work. For FEA, you need to understand and validate the results - you don’t get extra credit for doing the math (or in the case of writing, doing the writing) by hand. Use an LLM, understand how to prompt it, validate/read/refine the results.

➕ show 2 replies
Xinti • today at 8:31 AM

And this nonsense discredited idea:

"Think about what the creators of the content you consume want you to think2. Treat the things you consume like the Year 7 English class where you were taught how to identify persuasive techniques. Think about things as much as possible."

You assume that there is some Machiavellian scheme that is deliberately orchestrated like the Illuminati to do...stuff...and things...

Businesses want to make money. Anything you see around that would be network effects of doing so. It's not a fucking scheme that someone planned out like some nonsensical conspiracy theory: companies are people too (more specifically composed of people).

Okay, I'll bite: you presume, that you go from "read my artisanal science report" to "yeah man... like ideas come from people...from THE MAN...man...". Cool I can tell your nonsense just ran off the track of your own terrible thinking and lack of focus: nice, now I can tell how they could tell you wrote things.

You're a scientist right? Cool beans. Prove to me. absolutely PROVE TO ME, that your approach is somehow better, and by better I mean statistically better, you know high p values, peer reviewed, most citations, cohort studies, etc.

I'm betting that you could try to prove this (literally) personal opinion of yours and wouldn't get anywhere. I've heard the argument and no one has any level of proof approaching this.

➕ show 1 reply

🔗 View 4 more comments