_______ __ _______
| | |.---.-..----.| |--..-----..----. | | |.-----..--.--.--..-----.
| || _ || __|| < | -__|| _| | || -__|| | | ||__ --|
|___|___||___._||____||__|__||_____||__| |__|____||_____||________||_____|
on Gopher (inofficial)
URI Visit Hacker News on the Web
COMMENT PAGE FOR:
URI Boring is good
mallowdram wrote 1 hour 13 min ago:
The computer is not boring, per se, it's neutral when unplugged. The
idea it accelerates binary primarily for prediction purposes in
extracting values from the arbitrary, this is not only exciting, it's
illusory.
The writer fails to grasp what rabbit hole we've gone down since the
70s/90s when we began applying the principles of prediction to
computation, then horizontalized them via web. This was the most
exciting time because it added a vast illusory value to the arbitrary,
it's a time of massive piracy that posed as corporatocracy.
Once this prediction became automated by AI, yes, now the piracy
becomes boring, and in turn reveals what was going on all along.
Aldipower wrote 3 hours 33 min ago:
This is a good article, but
> We keep asking them to do âintelligent thingsâ and find out a)
they really arenât that good at it, and b) replacing that human task
is far more complex than we originally thought
I never thought that. From the beginning on this were easy to uncover
marketing statements.
Hendrikto wrote 5 hours 36 min ago:
> Whenever there is hype, we shuffled into the easy path, forcing the
tech into the product without understanding its weaknesses. We are more
worried about being left behind than actually doing something of value.
otikik wrote 5 hours 38 min ago:
> Weâre here to solve problems, not look cool.
Shots fired
liampulles wrote 6 hours 1 min ago:
LLMs are useful in contexts where fuzzy and hazily accurate is
acceptable. A developer trying to hack together some solution through
trial and error for example.
They are less useful in contexts where accuracy is expected or legally
required. An audit log for example.
Many businesses have made bad judgements of where the distinction is,
some don't even recognise a distinction. This will improve over time.
DeathArrow wrote 9 hours 35 min ago:
The investment fund that acquired the company that acquired our company
requests all that all companies it owns go big on cloud and AI, no
matter what, because this raises valuation and they can sell them for
bigger profits.
I have nothing against cloud or AI per se, but I still believe in the
right tool for the right job and in not doing things just for the sake
of it. While raising valuation is a good thing, raising costs, delaying
more useful features and adding complexity should also be taken into
account.
sfn42 wrote 15 min ago:
Easy, just quit. Go work for a serious company instead of these
grifters.
mindwok wrote 10 hours 48 min ago:
I feel that with LLMs and AI, people are furiously trying to argue the
reality they desire into existence. I've never read more articles
predicting the future than on this topic (I am guilty of it, too.)
qustrolabe wrote 6 hours 9 min ago:
Predicting AGI and all those "complete replacement of jobs" are
boring example of this. What's more annoying to me are people who do
that arguing into existence about reality they want where AI is
completely useless and fake and can't do anything and make you stupid
and ten more loudest clickbait titles of recent year
fuzztester wrote 9 hours 10 min ago:
>I feel that with LLMs and AI, people are furiously trying to argue
the reality they desire into existence.
The emperor's new clothes ...
nine_k wrote 10 hours 53 min ago:
The author of "Choose boring technology" regretted the choice of the
word "boring" [1].
Anyway, boring is bad. Boring is what spends your attention on
irrelevant things. Cobol's syntax is boring in a bad way. Go's error
handling is boring in a bad way. Manually clicking through screens
again and again because you failed to write UI tests is boring in a bad
way.
What could be "boring in a good way" is something that gets things done
and gets out of your way. Things like HTTPS, or S3, or your keyboard
once you have leaned touch typing, are "boring in a good way". They
have no concealed surprises, are well-tested in practice, and do what
they say on the tin, every time.
New and shiny things can be "boring in the good way", e.g. uv [2]. Old
and established things can be full of (nasty) surprises, and, in this
regard, the opposite of boring, e.g. C++.
[1]
URI [1]: https://boringtechnology.club/#30
URI [2]: https://github.com/astral-sh/uv
tpoacher wrote 6 hours 0 min ago:
> The author of "Choose boring technology" regretted the choice of
the word "boring"
Well, yes, but only in the sense that people kept giving him beef
about how boring is a bad word in their mind, not because it was a
bad word for this context per se. Which is somewhat ironic given your
comment!
I suppose what you're getting at is the difference between boring,
and "boooooriiiiiing".
adastra22 wrote 4 hours 55 min ago:
If âboringâ coveys the wrong, and at times even opposite
meaning in listeners ears, then âboringâ was a bad word choice,
even in context.
worthless-trash wrote 3 hours 49 min ago:
There is no good choice of words, because you can not control how
people interpret language.
Everything at some point will be interpreted incorrectly.
buster wrote 7 hours 56 min ago:
What you describe is the difference between tedious and simple.
Boring is good. I don't want to be excited by technology, i want to
be bored in the sense that it's simple and out of my way.
Same for KISS. I tend to tell people to not only keep things simple,
but boring even. Some new code i need to read and fix or extend? I
want to be bored. Bored means it's obvious and simple.
The difference? There are many complex libraries. By definition they
are not simple technology.
For example a crypto library. Probably one of the most complex tasks.
I would consider it a good library if it's boring to use/extend/read.
prmph wrote 8 hours 38 min ago:
Why conflate boring with old? "Boring" in this context means: proven
and stable. Yes, that would take some time to become apparent, but
the converse is not necessarily the case: a tech does not become
"boring" in a good way simply because it is old.
All this was my understanding before, so not sure why you think
"boring" was meant to be equivalent to "old"
ozim wrote 8 hours 22 min ago:
Problem is official definition is ânot interesting, tediousâ
with synonym âdullâ.
As much as I do get the idea, I can see how promoting use of
tedious to use and dull tools is something that really misses the
mark.
Well known and mature tools are still sharp and lots of them are
not tedious to use.
Ekaros wrote 7 hours 49 min ago:
What is the opposite of dull? Exiting?
I do not want my browser to be exiting. I do not want for it to
change every week. Say moving buttons to different places.
Changing how address bar operates. Maybe trying new short cut
keys...
Same goes for most useful software. I actually do want them to be
dull. And do their job and not get in between and make my day
more interesting by having to fight against it.
hbarka wrote 13 min ago:
Boring is the classic turn signal stalk on my carâs steering
column. Exciting is when Elon Muck decided to make it a push
button. He did not consider that British drivers would have to
do hand gymnastics in a roundabout. A few months later there
was a third-party Chinese made turn signal stalk add-on. And of
course the boring turn signal indicator was brought back by
Tesla. We could also talk about the wing doors on the Model X
or the exciting /s Cybertruck.
ozim wrote 6 hours 6 min ago:
Well known and mature tools are still sharp and lots of them
are not tedious to use..
I picked "sharp" not "exciting".
Dull knife doesn't do its job, you want tools that do the job
efficiently. "Boring" as it seems was interpreted as picking
tools that don't do the job efficiently. That is why original
idea creator found "Choose boring technology" most likely
misunderstood.
kmarc wrote 8 hours 46 min ago:
What you just described fits my definition of boring, which is some
function of (time passed, individual at keyboard)
Cobol was (and for some, still is) exciting at first, but _becomes_
boring once you master it, and the ecosystem evolves to fix or work
around its shortcomings. Believe it or not, even UX/UI testers can
deal with and find happiness in clicking through UIs for the tenth
thousand time (sure, last time I saw such Tester, was at around
2010).
This doesn't mean the technology itself becomes bad or stays good. It
just means the understanding (and usage patterns) solidifies, so it
becomes less exciting, hence: "boring".
But you can't sell a book with the title "Choose well-established
technology". Because people would be like, no sht, Sherlock, I don't
need a book to know that.
fnord77 wrote 12 hours 9 min ago:
> He uses the example of the dynamo, an old-fashioned term for a
powerful electric motor.
um, dynamo is a generator, it takes mechanical energy and turns into to
electricity.
adastra22 wrote 4 hours 39 min ago:
While technically correct (the best kind of correct), they are or can
be literally the same mechanism. Apply a mechanical force and you
will create DC current. Apply a voltage and you will generate motion.
sothatsit wrote 14 hours 2 min ago:
I tend to think that the reason people over-index on complex use-cases
for LLMs is actually reliability, not a lack of interest in boring
projects.
If an LLM can solve a complex problem 50% of the time, then that is
still very valuable. But if you are writing a system of small LLMs
doing small tasks, then even 1% error rates can compound into highly
unreliable systems when stacked together.
The cost of LLMs occasionally giving you wrong answers is worth it for
answers to harder tasks, in a way that it is not worth it for smaller
tasks. For those smaller tasks, usually you can get much closer to 100%
reliability, and more importantly much greater predictability, with
hand-engineered code. This makes it much harder to find areas where
small LLMs can add value for small boring tasks. Better auto-complete
is the only real-world example I can think of.
raincole wrote 8 hours 1 min ago:
Yeah. Is it even proven that LLMs don't hallucinate for smaller
tasks? The author seems to imply that. I fail to see how it could be
true.
adastra22 wrote 4 hours 50 min ago:
No? That is trivially not the case. Ask an LLM something outside
its training data and it will hallucinate the answer. How can it do
anything else? Maybe its hallucination ends up being correct, but
not all of the time.
a_bonobo wrote 13 hours 19 min ago:
>If an LLM can solve a complex problem 50% of the time, then that is
still very valuable
I'd adjust that statement - If an LLM can solve a complex problem 50%
of the time and I can evaluate correctness of the output, then that
is still very valuable. I've seen too many people blindly pass on LLM
output - for a short while it was a trend in the scientific
literature to have LLMs evaluate output of other LLMs? Who knows how
correct that was. Luckily that has ended.
empiko wrote 3 hours 3 min ago:
> Who knows how correct that was. Luckily that has ended.
What do you mean it ended? I still see tons of NLP papers with this
methodology.
adastra22 wrote 4 hours 52 min ago:
> for a short while it was a trend in the scientific literature to
have LLMs evaluate output of other LLMs? Who knows how correct that
was.
Highly reliable. So much so that is basically how modern LLMs work
internally. Also speaking from personal experience in the projects
I work on, it is the chief way to counteract hallucination,
poisoned context windows, and scaling beyond the interaction limit.
LLMs evaluating LLM output works surprisingly well.
danpalmer wrote 10 hours 15 min ago:
> I've seen too many people blindly pass on LLM output
I misread this the first time and realised both interpretations are
happening. I've seen people copy-paste out of ChatGPT without
reading, and I've seen people "pass on" or reject content simply
because it has been AI generated.
sothatsit wrote 13 hours 11 min ago:
True! This is what has me more excited about LLMs producing Lean
proofs than written maths proofs. The Lean proofs can be proved to
be correct, whereas the maths proofs require experts to verify them
and look for mistakes.
That said, I do think there are lots of problems where verification
is easier than doing the task itself, especially in computer
science. I think it is easier to list tasks that aren't easier to
verify than to do from scratch actually. Security is one major one.
hansvm wrote 11 hours 37 min ago:
Even there it's risky. LLMs are good at subtly misstating the
problem, so it's relatively easy to make them prove things which
look like the thing you wanted but which are mostly unrelated.
sothatsit wrote 8 hours 16 min ago:
Yes, Lean only lets you be confident in the contents of the
proof, not how it was formed. But, I still think that's pretty
cool and valuable.
alberth wrote 14 hours 8 min ago:
OT: Since the author is a former Apple UX designer who worked on the
Human Interface Guidelines, I hope he shares his thoughts on the recent
macOS 26 and iOS updates - especially on Liquid Glass.
URI [1]: https://jenson.org/about-scott/
sporkxrocket wrote 14 hours 9 min ago:
I like this article, and I didn't expect to because there's been
volumes written about how you should be boring and building things in
an interesting way just for the hell of it, is bad (something I don't
agree with).
Small models doing interesting (boring to the author) use-cases is a
fine frontier!
I don't agree at all with this though:
> "LLMs are not intelligent and they never will be."
LLMs already write code better than most humans. The problem is we
expect them to one-shot things that a human may spend many
hours/days/weeks/months doing. We're lacking coordination for long-term
LLM work. The models themselves are probably even more powerful than we
realize, we just need to get them to "think" as long as a human would.
liampulles wrote 5 hours 37 min ago:
"think" is doing a lot of heavy lifting there. Let's remember that
neural networks are designed primarily not as an accurate brain
models but as matrix models that can be calculated on in a feasible
time scale on GPUs.
da_chicken wrote 13 hours 6 min ago:
The issue is one that's been stated here before: LLMs are language
models. They are not world models. They are not problem models. They
do not actually understand world or the underlying entities
represented by language, or the problems being addressed. LLMs
understand the shape of a correct answer, and how the components of
language fit together to form a correct answer. They do that because
they have seen enough language to know what correct answers look
like.
In human terms, we would call that knowing how to bullshit. But just
like a college student hitting junior year, sooner or later you'll
learn that bullshitting only gets you so far.
That's what we've really done. We've taught computers how to
bullshit. We've also managed to finally invent something that lets us
communicate relatively directly with a computer using human
languages. The language processing capabilities of an LLM are an
astonishing multi-generational leap. These types of models will
absolutely be the foundation for computing interfaces in the future.
But they're still language models.
To me it feels like we've invented a new keyboard, and people are
fascinated by the stories the thing produces.
adastra22 wrote 4 hours 41 min ago:
I could make the exact same argument about the activation loops
happening in your brain when you typed this out.
Transformer architectures are not replicas of human brain
architrcture, but they are not categorically different either.
rbranson wrote 8 hours 55 min ago:
Is it bullshitting to perform nearly perfect language to language
translation or to generate photorealistic depictions from text
quite reliably? or to reliably perform named entity extraction or
any of the other millions of real-world tasks LLMs already perform
quite well?
da_chicken wrote 4 hours 27 min ago:
Picking another task like translation which doesn't really
require any knowledge outside of language processing is not a
particularly good way to convince me that LLMs are doing anything
other than language processing. Additionally, "near perfect" is a
bit overselling it, IMX, given that they still struggle with
idioms and cultural expressions.
Image generation is a bit better, except it's still not really
aware of what the picture is, either. It's aware of what images
are described as by others, let alone the truth of the generated
image. It makes pictures of dragons quite well, but if you ask it
for a contour map of a region, is it going to represent it
accurately? It's not concerned about truth, it's concerned about
truthiness or the appearance of truth. We know when that
distinction is important. It doesn't.
zoom6628 wrote 9 hours 33 min ago:
THIS !
bigstrat2003 wrote 13 hours 29 min ago:
> LLMs already write code better than most humans.
If you mean better than most humans considering the set of all
humans, sure. But they write code worse than most humans who have
learned how to write code. That's not very promising for them
developing intelligence.
adastra22 wrote 4 hours 43 min ago:
I have been writing code for nearly 30 years. I have my name on a
couple of ingenious patents.
I have fully integrated LLMs into May daily workflow, and am
routinely surprised at the innovative solutions that Claude
sometimes comes up with. There is a lot of generated code that is
far beyond my own capabilities, even as an expert in the field in
which Iâm working.
tibbar wrote 14 hours 21 min ago:
I think this is, essentially, a wishful take. The biggest barrier to
models being able to do more advanced knowledge work is creating
appropriately annotated training data, followed by a few specific
technical improvements the labs are working on. Models have already
nearly maxed out "work on a well-defined puzzle that can be feasibly
solved in a few hours" -- stunning! -- and now labs will turn to
expanding other dimensions.
adastra22 wrote 4 hours 46 min ago:
There are plenty of ways of writing more capable software stacks
using LLMs which donât rely on reinforcement learning. If anything,
the AI labs have too much of a focus on building larger models with
bigger or better sets of labeled data, where algorithmic changes will
let you do more with the same tools.
mikepalmer wrote 14 hours 41 min ago:
"LLMs are not intelligent and they never will be."
If he means they will never outperform humans at cognitive or robotics
tasks, that's a strong claim!
If he just means they aren't conscious... then let's don't debate it
any more here. :-)
I agree that we could be in a bubble at the moment though.
vjvjvjvjghv wrote 2 hours 22 min ago:
The word âneverâ is a dangerous one. I remember that computers
would âneverâ beat humans at chess. And people would never use
the internet for banking.
fakwandi_priv wrote 8 hours 59 min ago:
I have heard people on both ends of the spectrum:
- LLM's are too limited in capabilities and make too many mistakes
- We're still in the DOS era of LLM's
I'm leaning more towards the the 2nd, but in either case pandora's
box has been opened and you can already see the effects of the
direction our civilization is moving towards with this technology.
Aldipower wrote 3 hours 32 min ago:
DOS actually worked pretty well, at least it worked honest.
pjmlp wrote 3 hours 17 min ago:
I had lots of fun with DOS, even if I was the only PC guy in a
circle of friends that were Amiga owners.
stephenlf wrote 15 hours 17 min ago:
Great take. I personally find the thought of spec-driven development
tedious and boring. But maybe thatâs a good thing.
akagusu wrote 2 days ago:
I also agree that boring is good, but in our current society you won't
get a job for being boring, and when you get a job, it's is guaranteed
you are not being paid to solve problems.
kmarc wrote 8 hours 44 min ago:
Some early retirees who started learning Cobol just 8 years ago would
very much disagree with you :-)
com2kid wrote 14 hours 8 min ago:
> but in our current society you won't get a job for being boring,
One can argue that every other field of engineering outside of
Software Engineering, specializes in making complex things into
boring things.
We are the unique snowflakes that take business use cases and build
castle in the clouds that may or may not actually solve the business
problem at hand.
voxelghost wrote 12 hours 39 min ago:
We're everything from the Architect to the concrete guy,to the
framer, carpenter, sparky, and plumber.
... and if it all falls down, don't blame us - you clicked the EULA
/s
Tagbert wrote 14 hours 43 min ago:
One of my main job functions is to watch out for and solve problems.
keyle wrote 15 hours 16 min ago:
> and when you get a job, it's is guaranteed you are not being paid
to solve problems
That's just your experience, based on your geolocation and chain of
events.
DIR <- back to front page