LLM fatigue is real
My day job basically requires me to use LLMs (or, it is strongly encouraged) on the software that I develop. Now I understand why LLM fatigue and LLM burn-out happens. It is because you constantly have to babysit the LLM to keep it from screwing things up.
Babysitting is exhausting work. At one time in my life, as a 20-something-year-old, I was responsible for a group of 30 kids age 5 to 12. Every day I went home so drained that I could barely stay awake long enough to have dinner. I know from experience that school teachers perform some of the most labor-intensive work in the modern world.
It is what they call “emotional labor.” When other people are unable (as with kids) or unwilling (as with a lazy roommate) to live or play or work with you as a cooperative peer or team member, you essentially have to spend mental energy keeping an eye on them and making sure they aren’t doing something that is disruptive. That mental energy is a real energy cost, it really does consume calories and make you feel tired.
The worst part about it is that the LLM never really “learns.” Kids can learn to work and play with you. Lazy roommates can learn to live with you. But an LLM never really learns, and it keeps making the same mistakes over and over again. People who work with LLMs often end up keeping a directory of scripts they can deploy with a quick copy-paste so they can quickly re-explain something to an LLM. So the fatigue of watching over it and making sure the LLM doesn’t do something wrong is a constant emotional pressure, and it never gets better with age.
#tech #AI #LLMs #LLM #VibeCoding #AIFatigue #BurnOut
Remote
Ramin Honary
@ramin_hal9001@fe.disroot.org
Born in the United States, I am a professional software engineer, and have been since 2008. I do full-stack app development, mostly in Python and JavaScript, and also work on embedded computers, mostly programmed in C/C++. I am passionate about functional programming languages, especially Haskell, Scheme, and Emacs, so this is mostly what my posts are about.
I care deeply about human rights and justice for the poor, persecuted, and underprivileged people.
**Warning:** I will occasionally re-share posts containing images of the victims of the Gaza holocaust **which do not have content warnings.** I do this because I feel we should not look away from the horrors of genocide, we have collectively forfeit our right to not see the shock and horror of these images.
I am strongly opposed to war, fascism, and any ideology driven by hatred. I reject all forms of violence except self defense (war and terrorism are never self defense). Climate change is an issue of human rights because it will cause the most harm to the poor and underprivileged. I am especially concerned about large corporations and nation-states using AI as a means of controlling people through targeted advertising, mass surveillance, and "pre-crime" policing. Most of the political posts that I boost are about climate change, and the abuse of AI technology.
Since late 2023, I have been especially horrified by the genocide being committed by the United States and Israel against the people of Palestine, although I now understand that these crimes against humanity have been ongoing almost continuously since 1948. **A large number** of the posts I boost are about the ongoing genocide against the people of Palestine, as well as all the other innumerable crimes against humanity committed by the states of the US and Israel throughout the Middle East and elsewhere. Usually I am careful to only boost posts that respect content warning rules.
I also love retro-computing, especially computers from the late 70s to early 90s, in particular old Apple computers, but I love all old computers from that era. I will often boost post on these topics.
I am also passionate about free/libre software, especially Linux. Most of my posts are about functional programming languages, retro-computing, and Linux. I will occasionally author and boost posts about this.
I also enjoy procedurally generated art (which do not use LLMs), what you might call "algorithmic art," and photography, and will probably boost posts sharing artwork like this fairly often as well.
#tech #Software #Computers #Linux #Emacs #Lisp #Scheme #ProgrammingLanguage #ProceduralArt #GenerativeArt #AlgorithmicArt #Gaza #Genocide #MiddleEast #War #ClimateChange #GlobalWarming
I care deeply about human rights and justice for the poor, persecuted, and underprivileged people.
**Warning:** I will occasionally re-share posts containing images of the victims of the Gaza holocaust **which do not have content warnings.** I do this because I feel we should not look away from the horrors of genocide, we have collectively forfeit our right to not see the shock and horror of these images.
I am strongly opposed to war, fascism, and any ideology driven by hatred. I reject all forms of violence except self defense (war and terrorism are never self defense). Climate change is an issue of human rights because it will cause the most harm to the poor and underprivileged. I am especially concerned about large corporations and nation-states using AI as a means of controlling people through targeted advertising, mass surveillance, and "pre-crime" policing. Most of the political posts that I boost are about climate change, and the abuse of AI technology.
Since late 2023, I have been especially horrified by the genocide being committed by the United States and Israel against the people of Palestine, although I now understand that these crimes against humanity have been ongoing almost continuously since 1948. **A large number** of the posts I boost are about the ongoing genocide against the people of Palestine, as well as all the other innumerable crimes against humanity committed by the states of the US and Israel throughout the Middle East and elsewhere. Usually I am careful to only boost posts that respect content warning rules.
I also love retro-computing, especially computers from the late 70s to early 90s, in particular old Apple computers, but I love all old computers from that era. I will often boost post on these topics.
I am also passionate about free/libre software, especially Linux. Most of my posts are about functional programming languages, retro-computing, and Linux. I will occasionally author and boost posts about this.
I also enjoy procedurally generated art (which do not use LLMs), what you might call "algorithmic art," and photography, and will probably boost posts sharing artwork like this fairly often as well.
#tech #Software #Computers #Linux #Emacs #Lisp #Scheme #ProgrammingLanguage #ProceduralArt #GenerativeArt #AlgorithmicArt #Gaza #Genocide #MiddleEast #War #ClimateChange #GlobalWarming
1236 Followers
851 Following
50 Posts
CodeBerg:
GitHub:
NeoCities:
Open post
Apropos of nothing: Big thanks to Shiro Kawai for creating Gauche Scheme, which so far has been one of the best Scheme implementations I have used. It has worked well on every computer I have tried it on so far ({Linux, FreeBSD, MacOS} x {x86_64, i386, aarch64}). It is stable, very fast, and supports a large number of the more recent SRFIs (Scheme language extensions), it has very good compatibility with the R7RS language standard, and the Gauche REPL gosh is pretty easy to use as well.
I always run the REPL in R7RS mode, so gosh -r7 -I. -i . Scheme people, give it a try sometime if you haven’t already!
https://practical-scheme.net/gauche/
#tech #software #Scheme #R7RS #Lisp #SchemeLang #FunctionalProgramming #Gauche
10
1
3
0
Open post
Replying to on gamerplus.org
@screwlisp@gamerplus.org computing numbers that are not in the set of {integers modulo 2^64} is really difficult. I have always appreciated how Common Lisp and Scheme both show a great deal of respect to the field of numerical computation by providing programmers with a proper numerical tower.
Re. monads: a struct is a part of a monad, the other part is the binding function that unfreezes the thunk and binds it’s result to an argument of the next thunk. I think the let* binding is a good example of the Identity monad, which is a trivial monad that does nothing but bind variables. You understand how the let* binding can be macro expanded to lambdas, right?
(let*((a (f1 "hello"))
(b (f2 "world")))
(concat a b))
…can be written as…
(funcall
(lambda (a)
(funcall
(lambda (b) (concat a b))
(f2 "world")))
(f1 "hello"))
This is an example of how funcall performs the bind operation of the monad. You can introduce a struct or a record type to chain the lambda functions together and evaluate the monad using a function that runs funcall but also performs some other action before it binds the result of the funcall to the next lambda in the chain. Then if you write a macro like monadic-do where you can build up your chain of lambdas in your record types, it looks like an ordinary procedure but with different semantics than ordinary let*. This is essentially what a monad is.
(monadic-do 'my-stateful-funcall
(a := f1 "hello")
(b := t2 "world")
(concat a b))
Monads are useful to Haskell because this allows you to express procedures using only the minimal semantics of System-F that Haskell provides to you, and can do so in a way that the type system can absolutely prove to be correct.
But I don’t see much use for monads in Common Lisp or Scheme because these languages already provide you with procedures, macros, and other means of changing the semantics of your procedures.
5
10
2
0
Open post
River is a Wayland compositor written in Zig that defines a window manager protocol so that, like with X11, write your own window management software. As with X11, the protocol is language-agnostic, so you can define your window manager using any programming language at all.
I have been gradually becoming more interested in Zig as a potential replacement for C in systems programming. Rust is another interesting potential replacement.
#tech #software #Zig #ZigLang #Linux #Wayland #WindowManager #WaylandCompositor #RiverCompositor #UnixPorn
7
2
6
0
Open post
Replying to
@mdhughes@appdot.net @flyingsaceur@ioc.exchange @flyingsaceur why can’t we pick more than one?
realistically: FreeBSD on RISC-V netbook in a Shenzhen studio apartmentalso possible: A/UX on a Quadra 700 in a bog hutrealistic second preference: Lua on an ESP32 terminal emulator on Inception Beach, although I would probably try to get Gauche Scheme on it, instead if Lua. (Lua is nice, though.)
1
1
0
0
Open post
Replying to
@pixelate@tweesecake.social
I'm not visually impaired, but if there isn't a UEFI screen reader that you can flash onto firmware, there ought to be, and a Linux driver to use the UEFI screen reader ought to exist as well, if it doesn't already.
7
2
3
0
Open post
Replying to
@cdegroot I think you might have a factual error on page 174 in the section called “Emacs Lisp.” You write this:
In a nutshell, Emacs Lisp (“elisp”) is a classical Lisp-1 with dynamic variable scoping. As you may recall, a Lisp-1 has a shared namespaces for symbols depicting variables and symbols depicting functions, while a Lisp-2 has two such namespaces; Scheme is a Lisp-1, and Common Lisp a Lisp-2. The debate on which one is better will likely never be settled, so it was probably either a coin-toss for Stallman or simpy a matter of expediency, if he thought a Lisp-1 would be simpler to implement.
I believe Emacs Lisp is a Lisp-2 system, unless I am misunderstanding some finer point about the difference between Lisp-1 and Lisp-2 systems. For example, Emacs Lisp has defvar, setq, defun, and funcall. You can use setq to set the value of a variable f and you can also do something like (defun f () (princ (format f))). So unless I am mistaken, that makes Emacs Lisp a Lisp-2.
@screwlisp#tech #software #Lisp
6
3
2
0
Open post
Replying to
@mxchara@seattle.pink if I were head of the MPAA, I’d have a rule that all autocratic leaders in every movie must be as repulsive as Imortan Joe from Mad Max, or it is a violation of “best standards and practices.”
1
2
0
0
Open post
@screwlisp@gamerplus.org @kentpitman@climatejustice.social I’m just reading up on the MIT-Scheme condition system. Recent efforts to standardize this are defined in SRFI-255: “Restarting conditions”.
An older standards condition systems in Scheme was defined in SRFI-35: “Conditions”. And #Guile users can use the Guile implementation of SRFI-35 to make use of it.
I wish I had known about this two weeks ago when we first started talking about it on the #LispyGopherClimate show, but better late than never, I guess.
#tech #software #Lisp #CommonLisp #Scheme #SchemeLang #R7RS #MITScheme #Guile #GuileScheme
7
1
3
0
Open post
Replying to
@hellomiakoda@pdx.social @foone@digipres.club I mean, I love solid state, especially for how robust they are. But I also love clackity keyboards and the feeling of things like modules and cartridges sliding into place.
2
0
0
0
Open post
Replying to on woof.tech
@momo@woof.tech cool!
Well then, please have a listen to the episodes of the #LispyGopherClimate podcast that me, @kentpitman@climatejustice.social and @screwlisp@gamerplus.org did these past few weeks:
Kent Pitman presents his lisp condition system implemented for pythonKent Pitman, Scott Zimmermann, Ramin Honary, Screwlisp: Lisp Conditions in PythonKent Pitman with Common Lisp condition handling loreKent Pitman, Common Lisp condition system, code review of my programming exampleCommon Lisp condition handling w/ Kent Pitman, Ramin Honary
#tech #software #Lisp #CommonLisp #Scheme #SchemeLang #R7RS #MITScheme #Guile #GuileScheme
6
3
1
0
Open post
Replying to
@wingo@mastodon.social my god, do people really talk like this now?
I guess I am pretty picky about with whom I choose to socialize because (thank goodness) I haven’t ever heard anyone say these things in real life. Hopefully the bubble will burst before I ever do meet anyone who thinks like this.
1
0
0
0
Open post
Replying to
@freediverx
in Apple's case, it was switching to Swift as their application level programmung labguage that probably caused those performance issues.
Microsoft's .NET, on the other hand, is solid as a rock. So Windows performace issues are mostly to do with spending 50% or more of every computer's resources on spying on people.
@gullevek @nikitonsky
2
4
0
0
Open post
Open post
Replying to
@ben
I think the entrenched powers know that if peopke are desparate they are easier to control: they settle for less, they are placated by cheap hand-outs, they are afraid to challenge authority lest they lose what little they already have. In that sense, the cruelty is the point, because its yet another way for them to maintain power over everyone else.
1
0
0
0
Open post
Replying to
@worldwidewerner my solution is to copy all the stuff I really like into a folder on my computer’s hard disk. I’ve been doing it for things I read on the Internet too, lately, since search engines are all really bad at searching now.
1
0
0
0
Open post
Replying to
@PhoenixSerenity @CTGT
It can't be a "Golden-crowned Kinglet," because what it looks like is a "Poofie Ball Birb-n-floofy yellow hat." That's what it ought to be called.
1
0
1
0
Open post
Replying to
@m3tti@functional.cafe thanks for recommending “Lisp Hackers,” I haven’t read that one yet.
@screwlisp@gamerplus.org @cdegroot@mstdn.ca
1
2
0
0
Open post
Replying to
@m3tti@functional.cafe oh, I just checked the website and it is a free PDF download! I guess I can read that one next!
@screwlisp@gamerplus.org @cdegroot@mstdn.ca
1
0
0
0
Open post
Replying to
@cosmiclibrarian on the one hand, there was a military target near by, the missile could have malfunctioned.
EDIT: it was apparently a double-tap strike. So this was 100% a deliberate attack on a school.
On the other hand, Israel has a pattern of murdering children with their “Project Lavender” and their technology called “Where’s Daddy” which would follow a “terrorist” (ordinary Palestinian person) home then bomb their entire family. They also bombed almost every single elementary school in Gaza, and continue to starve ordinary children to death in Gaza, even though they achieved their goal of “bringing home all the hostages” and claimed to agree to a ceasefire.
So the school could have been deliberately targeted as part of a strategy to kill the entire family of one of their assassination targets.
This is what happens when a nation-state establishes a pattern of violating international law, no one can ever trust what they do is honorable or moral ever again, not until the war criminals are made to face justice.
1
3
0
0
Open post
Replying to
@screwlisp@gamerplus.org I should also say, LLMs also work on non-textual information like images and and sound. That is the advantage of using purely statistical methods, rather than the symbolic method you described. You can easily interface a LLM with a neural network that decomposes an image or sound into quanta that can be mapped onto the vectors that encode some “meaning” of words/tokens in a sentence, so you can prompt an LLM to create an image from text, or to create a textual description of an image.
Statistical methods can also create a “sentiment” vector for passages of text, finding some way to encode the meaning of a textual passage into a number that you can use to compare to other numbers to decide if two passages are similar in meaning. This is useful for quickly searching through a large text corpus. Most modern chat bots do this, it is what they call “Retrieval Augmented Generation.” They use sentiment analysis to find text passages that seem like they may be related to your prompt, and use the “similar in sentiment” text passages to augment your prompt before then constructing an output.
The purely symbolic language models with sparse matrices would probably be much more energy efficient, but the trade-off is that you lose the ability interface with neural networks trained on image or audio signals and also the ability to perform sentiment analysis.
@AmenZwa@mathstodon.xyz @aartaka@merveilles.town @dougmerritt@mathstodon.xyz @kentpitman@climatejustice.social @neauoire@merveilles.town @rwxrwxrwx@mathstodon.xyz
0
2
0
0
Open post
Replying to
@icm@mastodon.sdf.org
That menu reminds me of installing Linux back in the 90s and formatting the hard disk with FDisk and selecting the partition labe from a similar looking menul. Except this is a list of actual functioning operating systems that you can use from which to choose.
0
0
0
0
Open post
Replying to
@xameer@mathstodon.xyz truth is always going to be difficult, I don’t think that is a technological problem.
Modern technological advances will result in new problems in understanding truth and will require new solutions to those problems. But what will never change is that humans will trust each other and will place more weight on evidence from sources they trust.
@srtcd424@mas.to @wizzy@noc.social
0
2
1
0
Open post
Replying to
@brainsmatter@mastodonapp.uk the term “strong-armed“ implies these AI firms were unwilling to earn profits off of surveillance technology. What an absurd thing to believe of an AI company. Was this article written by the propaganda department of one of those AI companies or something?
(Reads disclosure statement.)
“Bianca Baggiarini receives funding from the Australian Defence Strategic Policy Grant program.”
Uh-huh, so yes, more or less.
0
0
0
0
Open post
Replying to
@alice
Are you a chair manufacturer?
I'll lobby congress on your behalf to pass law requiring every AI agent have their own physical chair... for a fee equal to 1 percent of your eminent billion dollar goverment contract.
@crowbriarhexe @cwebber
0
2
0
0
Open post
Replying to on infosec.exchange
@guyjantic@infosec.exchange but this “powder keg” has already exploded. The surge in ICE and CPB is basically the US government saying, “hey, we’ll employ the unemployable into our private army to attack our political enemies.” ICE and CPB are full of conservative and formerly unemployed people with no critical thinking skills who love to take out their anger on the weak and underprivileged.
0
0
0
0
Open post
Replying to
@foone@digipres.club woah, this thing could very well win the #NanoraptorChallenge couldn’t it!
This thing looks to me sort-of like an IBM PC Convertible with a CD-ROM drive.
How cool would it have been to have a IBM PC Convertible with a built-in LaserDisc player? (Nudge, nudge, @NanoRaptor@bitbang.social )
0
1
0
0
Open post
Replying to
“spent 6 months vibecoding my app and nobody wants to use it ;_;”
@pathunstrom@ngmx.com @cwebber@social.coop I think the idea of “apps” was a stupid Silicon Valley construct — an artificial border drawn around a piece of software functionality for the purposes of rent-seeking.
I am glad the idea of apps is dying off, and I guess we can thank so-called “agentic AI” for this now. You download Pi or #OpenClaw, then rather than download “apps” or “extensions” for it, you just ask the agent to modify itself. Tailor it to your workflow.
And by the way, #Emacs was the original Agentic AI from the first AI bubble back in the 1980s when everyone was calling the #Lisp programming language “AI”. It just that with Emacs no one ever seriously considered using natural language to program it (“The Doctor,” based on ELIZA, was included as a game, not an actual utility). The point is: all that was old is new again.
0
0
0
0
Open post
Replying to
@tiefling@bardicperspiration.club yeah, you’re probably right about it being mostly tech people, and maybe some pixel art people and 8-bit game people. It makes sense that would be the case, nowadays you do need some technical skill to re-live the old creative Internet, since it has been so thoroughly supplanted by the new exploitative/extractive Internet.
@DaimonRainbow@bardicperspiration.club
0
5
0
0
Open post
Replying to
@Flaky@furry.engineer The choice between Linux and Mac/WIndows is fundamentally a choice between privacy and convenience, so it is a little weird to talk as if Linux is a competing product with operating systems like Mac or Windows, like in a Consumer Reports article or something. To me, it just doesn’t make sense at all to think of Linux in that way.
I don’t think anyone should judge you as to whether you need convenience whether you need privacy, it depends on your personal needs, so just choose the OS that suits your needs.
If a “smooth experience” or convenience is your priority, then just don’t switch operating systems at all. You don’t have a choice of products, you just get what came with the computer when you bought it. There is no competition among operating systems here. You just have to sacrifice some of your privacy for convenience.
If, however, your priority is privacy, or getting useful work out of old computers, then as with everything else in life, you are going to have to sacrifice some convenience for your privacy, or sacrifice convenience with for being able to use an old computer, and just deal with the rough patches inherent in using Linux.
0
1
0
0
Open post
Replying to
@tiefling@bardicperspiration.club @DaimonRainbow@bardicperspiration.club I think that is why sites like Neocities exist (where op is hosting their blog, and this article).
To find a more humanist, fun Internet that has not been ruined by capitalism, follow the hashtags #smolweb #smallweb #retrocomputing
0
7
0
0
Open post
0
2
0
0
Open post
Replying to
“I mean tbh that’s a failing ov the “creative” internet then? Unless ppl just want an alt techbro space I guess.”
@tiefling@bardicperspiration.club yeah, I agree, it is a failing. Tech bros, even the anti-fascist variety, have a hard time reaching out to people who aren’t like them. I was just talking about this in a different thread, but if us tech guy’s unique skills are so useful, and if we are really interested in becoming more communitarian, we would offer our apps to more people in our communities.
We just have to be clear about how we communicate, that our apps aren’t going to have the same level of quality or reliability as something from Google or Amazon or Apple, and we can’t provide tech support all the time. It is something we do for fun, and maybe to provide something we can use together with our friends. So as long as no one comes us demanding customer service when it isn’t working right, we’d be happy to share. But we need to learn to temper our speech, and stop talking like a Silicon Valley tech bro about changing the world and better ways of doing things and what not. Stop worrying about “scale” and “features” and “selling points,” because this isn’t about selling anything.
Then we might get more creative non-tech people involved, which would be benefit to everyone.
@DaimonRainbow@bardicperspiration.club
0
2
0
0
Open post
Replying to on gamerplus.org
@screwlisp@gamerplus.org I have been investigating it myself a bit more, and though I am still not completely sure how Emacs is doing it, I think the problem is that (set-default 'var-name ...) is probably looking-up symbols among the set of buffer locals first, then the stack, then the top-level, but (eval 'var-name) or (setq var-name ...) looks-up in the order of stack, buffer-locals, top-level.
I understand that Emacs buffers act as a sort of “environment,” or maybe an instance of a “module,” and that buffer-locals are sort-of like protected variables from object oriented languages
0
1
0
0
Open post
Replying to
@tusharhero
@mark
I would only ever call it "Emaxxing" as a trenchant parody of "work maxxers".
It would be really funny if someone made a straight-faced parody video about how to use Emacs in the same over-enthusiastic and overconfident fashion in which people made videos about how to get rich writing apps with OpenClaw or whatever is popular nowadays, emphasizing the false fact about how Emacs had made you and thousands of others multi-millionaires with almost no effort.
Although no one would get the joke I'm sure, what with the state of AI hype being what it is nowadays, parody is completely dead at this point.
0
0
0
0
Open post
Replying to on furry.engineer
@Flaky@furry.engineer yes, well then I guess I am disagreeing with those people who try to treat Linux as a product. I guess I can sort-of see why KDE would do that, they actually hire developers to perfect their desktop environment, maybe there is a profit motive to their work.
It would be so much better if regular people who try to “sell” Linux would instead try to encourage people to learn more about the importance of computer privacy and “upcycling”, that way people are motivated to switch to Linux for the right reasons, not because they expect to find an OS that has “better features” or whatever.
0
0
0
0
Open post
Replying to
@freediverx
Yes, I too pine for the days of System 7.5, the Apple Human Interface Guidelines, and the "Platinum" theme by Susan Kare. You really only needed one mouse button.
I agree, modern MacOS is a trash pile.
Nowadays, on my work laptop (MacBook M1), I avoid the MacOS UI/UX by keeping Chrome and Emacs both full screen and switching between them with the Ctrl-Arrow keys. No dock, everything launched with Spotlight or the Emacs Shell. Actually, I do something similar on Windows as well. Super-Tab to switch from Emacs to Chrome and back.
@gullevek @nikitonsky
0
0
0
0
Open post
Replying to
@dougmerritt the lack of cat videos definitely destroyed Gopher’s commercial chances as well. Gemini should have addressed that issue.
@wrog @cy @screwlisp @mousebot @cdegroot @kentpitman
0
0
0
0
Open post
Replying to
@DeliaChristina@sfba.social if they are saying this, then they have reason to believe that they are untouchable.
0
2
3
0
Open post
Replying to
@wrog my instance has a 5000 character limit, but I rarely type that much. Most of my posts are three or four paragraphs. I could probably make do with 2500 characters, but it is nice to have a little extra space so I don’t feel like I have to watch the character counter so closely. And I have space to write in complete sentences, to include example code, paste long URLs, and quote parts of other posts inline.
@cy @screwlisp @mousebot @cdegroot @kentpitman @dougmerritt
0
0
0
0
Open post
Replying to
@lew
The "horrible massacers" right now are being caused by the US and Israel, not the regime.
A girls school was targetted by Israel yesterday, 110 children were killed. This is not an accident, this is a pattern the Israelis have of murdering whole families --- "project lavender," and "where's daddy" for example were Israeli's bombing their assaination targets in their own homes to maximize collatetal damage. They are going after families of Iranian political leaders, thats why they bomb school children. Netanyahu is wanted for war crimes, of course if he evades justice he will commit more war crimes.
The Iranian people will never accept any regime change if it is forced upon them by the US. This war has made regime change impossible for at least another decade now. Quit drinking cool-aide.
0
0
0
0
Open post
Replying to
@mkerjean
They need to make a transparent bubble to put your laptop in with two holed you can fit your hands through.
@julesh just yesterday I wad thinking about attaching linear servo mechanisms betweer the backplane and frame mount of my mechanical keyboard that would cause the whole keyboard to jump just a tenth of a milimeter every time I press a key, to get that nice, Selectric feel to it.
0
0
0
0
Open post
Replying to
@CppGuy@infosec.space dd == “destroy disk.” But I have used that command hundreds of times, and I have not yet once destroyed a disk.
Only one time I had a scare: the command failed because I specified the input source twice. Had I flipped the input and output, that would have been the first time I destroyed a disk.
But it hasn’t happend (yet).
0
1
0
0
Open post
Replying to
@knowmadd@mastodon.world for a second, I read the question as, “The car is 50 meters away, should I walk or drive?” Then I realized it said “The car wash is 50 meters away,” and I got why this would trick the AI.
LLMs work on the “attention” model to predict what output comes next. It is trained on which parts of the sentence deserve the most focus when predicting the result and generating an answer. If the meaning of a sentence can be changed entirely by just one short word, it is more likely to trip-up an LLM.
0
0
0
0
Open post
Replying to
AI really makes them come up with the weirdest ideas
@Brokar to me, these weird ideas they come up with is evidence that they don't have a plan or a business model to earn profits (apart from selling AI to fascist dictatorships for mass surveillance) and are desperate to recoup the losses in their investments, and prevent the bubble from bursting.
0
1
4
0
Open post
Replying to
@wizzy@noc.social @xameer@mathstodon.xyz I remember reading that comment about a half-year ago during the so-called “12 day war,” which really should have been called the start of the “US-Israeli war of aggression against Iran” or maybe just “World War 3.”
I wonder if opinion polls were taken today they would paint a different picture? A lot of that “economic mismanagement” was about spending an unbelievable percentage of their oil profits on the IRGC and their missile defense program. For example, their aging infrastructure in Tehran has led to a water crisis so severe the government was talking about moving their headquarters out of the city. From what I heard, a lot of those “riots” from last month were motivated by criticism that too much money was spent on defense and not enough on improving the water infrastructure.
Now that the war has escalated, I think a lot of Iranians are starting to feel more thankful that they have these defense systems in place, and enraged about the murder of 160+ children in Minab. At least that might be why we have seen a number of pro-government demonstrations in Tehran since last week. I am guessing that right now people are hopeful that the new government might be less despotic and more pragmatic.
But I don’t live there, I am only going off of what I heard from a few people I know who have a better idea about what is going on, like @faab64@freefree.ps , and it is hard for anyone to get information out of Iran right now.
@srtcd424@mas.to
0
7
2
0
Open post
Replying to
@cosmiclibrarian yes, you are right, and I was wrong.
I am now reading that the attack on the school was a double-tap strike, so there is no way it was a malfunction or a targeting error. They bombed a school, waited for emergency responders to come in, then bombed the emergency responders as well.
The barbaric savagery of the Israeli military is boundless.
0
2
0
0
Open post
Replying to on mathstodon.xyz
@screwlisp@gamerplus.org I remember doing something like this as a grad student when I first started learning about natural language processing. I was just playing around with various ideas in code (I was using Perl at the time).
I remember breaking-up a sentence into tokens and just counting how often one token followed another in a large corpus of data. Then I created a simple weighted directed graph data structure associating every token with every other token weighted by the count, that is, how often any one token X was seen immediately after any other token Y.
The directed graph was really a dictionary data structure where the token X was the key and the value associated with the key was another dictionary associating the token Y with it’s weight, so there was no matrix, or I guess you could say it was a sparse matrix.
I even wrote an “inference” algorithm where you would just perform a random walk through the graph biased by the weights. The output was gibberish, but (to my surprise) mostly grammatically correct. I wanted to scale it up, associating tokens triples with other token triples to see if that improved the grammar, but I lost interest.
This is a little different from how LLMs works though, in one important way: an LLM also encodes the position of the tokens in the prompt and uses that position information as something to be trained on. You could do this with the symbolic process I described above (the “sparse matrix), but you would have to take an input list of tokens X and an output list of tokens Y, and run the “training” process for each element in the list, so X[0] -> Y[0] would have it’s own sparse matrix, X[1] -> Y[1] would have it’s own sparse matrix, and so on. You would create a sparse matrix for each element in your list for some number of tokens (e.g. a thousand). Then it would really be like an LLM.
Or did I just describe exactly what you have done here in Common Lisp?
@AmenZwa@mathstodon.xyz @dougmerritt@mathstodon.xyz @kentpitman@climatejustice.social @rwxrwxrwx@mathstodon.xyz @aartaka@merveilles.town @neauoire@merveilles.town
0
2
0
0
Open post
Replying to
@simonzerafa@infosec.exchange
I am ashamed to admit that I understood this joke, Dr. Noonian Soong.
@NanoRaptor@bitbang.social
0
1
0
0
