I’m getting really tired of all these junior-turn-senior devs who can’t write simple code asking ChatGPT to solve everything for them.

I’m having to untangle everything from bizarre organization/flow to obvious gotchas / missed edge cases to ridiculously long math chains (that could be 1/10th the length), or — and I feel so dirty for this — resorting to asking ChatGPT wtf it was thinking when it obviously wrote some of these monstrosities. Which it gets wrong much of the time.

“ALL HAIL CHATGPT!” Proclaims the head of Engineering. “IT’S OUR PRODUCTIVITY SAVIOR! LEVERAGING AI WILL LET US OUTPERFORM THE ENTIRE INDUSTRY!”

Jesus fucking christ.

I am really going nuts about everyone using ChatGPT. Had literally discussions 'bUt cHaTgPt sAyS iTs TrUe', when the docs said clearly thats not the case.

Also the "code" produced is just what the hell?!

I hope this damn hype will end soon

"ChatGPT passed an interview for Google"
"I ask to ChatGPT to write my new song"
"What ChatGPT tells about our humanity"
"ChatGPT pooped its pants"

I'm the only one sick of seeing articles, posts and threads about Chat fucking GPT?! I can't wait for the hype to die out... or for someone to build a time machine able to bring me back to 2009

2022 Dec
Everyone: hehe, ChatGPT is soo cool, it responds to me and even knows what's the capital of France is! It is really cool!!!

2023 Jun [6 months later]
Everyone: hehe, ChatGPT is SOOOO stupid! I asked it to write a microkernel with native HTTP and TLS support and it made a semantic mistake in line 98823 lolz what a PoS

Someone's having a bad, bad day...

Revenge

If you ever wondered how to get all the data from your database in reverse order, chatGPT is here to help.

I'm cracking up...

"chatGPT will ruin the software interviewing industry!!!"

uh.... what does it tell you about our industry if a fucking ROBOT can "ruin" the interview

well, you're right. it tells you that only algorithmic robots do well and subsequently earn the top spots at software companies after interviewing.

creativity, grit, perspective, wisdom? that stuff is absolute bullshit!!! (and as a feeble human I can't figure that out in an interview anyway!!! better just have you solve leetcode problems ad naseum!!! that'll get us the best employee!!!)

god i hate the dumb fuck rat race. good thing i'm not in it anymore! peace out, girl scout✌️

Just stir the pile...

I get it. Please stop.

So ChatGPT with GPT-4 has dropped and it's only available through a paid subscription.

I hope everyone who started to become dependent on the "free" ChatGPT or started building a business on it feels pretty stupid right about now. Unless you pay up, everyone else who does pay will now have an edge over you. Congratulations for giving another monopoly into the hands of microsoft so enthusiastically.

The "open" part of OpenAI is such a joke...

Nothing to say...

Putting chatgpt to some good use. Writing a complaint mail to the idiots maintaining my banking app in the style of shakespare.

Hark thee, App Support Team,

With grave disquiet and vexation doth I write to thee concerning thy recent update of the application. As a software developer, the option to enable developer settings on mine own mobile device is of paramount importance for mine work. Yet thy latest update hath impeded mine access to mine own bank account until I disable this setting. Upon launching the app, it doth redirect me to a browser tab, where I am compelled to deactivate the developer setting to avail of thy services.

This conduct of thine is most unacceptable and unprofessional in mine eyes. It doth seem a transgression of privacy, for thy app doth dictate what settings I may or may not have on mine own personal phone. How canst thou deny me access to mine own bank account information merely on the grounds of having enabled developer options? How doth this option interfere with thy application, such that thou must needs coerce thy users to forsake their phone settings to utilize thy app?

I beseech thee to rectify this issue with all due haste, so that I may access mine own bank account without hindrance. If thou art incapable of doing so, then prithee, might thou recommend a more user-friendly banking application to which I may gladly switch?

With frustration and discontent at this time,
A locked-out person.

Backstory : So recently one of my banking app stopped working and forced me to update to their latest version. As soon as i opened the newer version , it shut down and redirected to my browser with a shitty html page with just one message : Disable developer options on your device to continue using our app. I was extremely frustated and couldnt understand what kind of idiots were maintaining this app.So i decided to write up an email hoping to find some solution for this.

Unpopular opinnion: Whole IT industry is becoming more and more degenerative with each passing day..

Life at work these days:

Manager: we’re not getting enough done
My calendar: 1/3 week filled with scheduled meetings

Manager: we need to use ChatGPT intensively. We'll go a lot faster.
Me using ChatGPT to get it to write an automation script: 2 hours wasted with no success
Me starting again from scratch to write the script: 15 minutes to achieve the desired result.

Thanks for your advice boss

Am I really unlucky, or are juniors these days all lazy af and such pampered babies that need hand holding all the time?
So back when I was a junior, when I wanted to learn something new, I would ask for some pointers from my seniors, could be an article, a video or even a book. From there I would look up further knowledge, play with the idea in my machine. If I couldn't understand something, or if I needed a better explanation of something, I would go back to my senior, but it was really rare.
Then comes this modern day, I'm the senior now and I'm in charge of mentoring a bunch of kids, who would treat me like their personal chatgpt. "Hey Junior #0, this is something you may want to read to help your next ticket, let me know if you have difficulty". Next day junior #0 would come back and say "I don't understand, the article mentioned X but I don't know how to do X. Can you show me how to do X?". Bro, no one knows how to do X after being born, just google "how to do X" and it gives you the fucking answer. Why the fuck do you have to circle back to me because of this. Junior #1 would refuse to read any articles longer than 250 words, and require constant 1-1 meetings to give him personal lectures. Dude this is not a class room, grow the fuck up! Junior #3 would write the messiest code possible despite my efforts to introduce tons of resources, then complain "why I'm still junior, how do I grow". Bro maybe if you learned half of what I sent you, you would have gotten promote by now. Fucking lazy kids these days!
Oh I can't fire these juniors. Top management was very clear that "we don't have budget to hire other devs for you, it's your responsibility to train them better".

I think we need a separate Internet where you get a fucking lifetime ban for mentioning chatgpt. It's like NFT thing but worse. People have verbal diarrhea. IT WRITES SHIT CODE, WE GET IT, SHUT THE FUCK UP ALREADY

Tried to use chatGPT. These motherfuckers want a phone number. Yeah no assholes, fuck you.

??????????????????? What??????

???? What???????

I couldnt solve a bug for hours.
Hours of googling.
Hours of mental trainwreck.
Hours of stress.
1:28 am.

I cant solve it.

HOLD ON I HAVE AN IDEA.

ChatGPT AI. HELP ME.

i copy and pasted the part of code thats bugging me. Keep in mind that this is a VERY large and robust system and this is just tiny percentage of code.

I told the AI to help me fix this shit bug.

ChatGPT literally explained me what the bug is as if im retarded and wrote code how to fix it.

LOOK AT THE SCREENSHOT U CANT EVEN MAKE THIS SHIT UP

HOW????

soo.. yeah.. I've just solved an annoying bug using only chatgpt.

My first commit in this new project. And it's based on chatgpt.

Literally just saved me from days of reading through kafka docs, auth mechanisms and other stuff. And no, the google did not provide me with a proper answer/hints. The only hint was "the configuration might be wrong". Well alright, but I was NOT using any configuration in the first place...

Fun times ahead :) I might even consider the pro version if it keeps delivering like that.

All these "IT pros" that is constantly telling others to use ChatGPT is making me cringe.
They are basically saying that themselves are to dumb to understand or think by themselves

5 stages of failing WIFI connectivity on Linux

This morning I woke up my laptop to start my work day. I have 2 very important meetings today, so I better get all prepared.

"Wifi connection failed"
Syslog says:
- wpa_supplicant: wlp9s0: SME: Trying to authenticate with <MAC>
- kernel: wlp9s0: authenticate with <MAC>
- kernel: wl9s0: send auth to <MAC> (try 1/3)
- kernel: wl9s0: send auth to <MAC> (try 2/3)
- kernel: iwlwifi: Not associated and the session protection is over already...
- kernel: wl9s0: send auth to <MAC> (try 3/3)
- kernel: wl9s0: authentication with <MAC> timed out

#### DENIAL #####

No biggie, let's try another AP (I have 3). All 3 failed to connect. Fine, let's try my phone's hotspot! FAILED!!!!!

w00t.... okay, let's restart the router... but failing to connect to a phone hotspot is already a worrying sign.

Wifi connection failed

wtf.. disable and re-enable wifi

Wifi connection failed

#### ANGER #####

the fuuuuuuck. Maybe my router is dead. But my phone connects to it, no fuss. My personal lappy also connects there easily.

wtf... Does that mean I'm about to lose my uptime?? Come one!! It's Linux - there MUST be something I could do! I don't see processes hanging in D state so the radio must be fine - it's gotta be a software issue!

ChatGPT – type all the log entries manually, via phone (that took a while...). Nothing useful there: update firmware, restart NetworkManager, etc.

#### BARGAINING #####

Alright... How about a USB dongle? Plug it in and wifi connects immediately! Yayyy!!! But that's only b/g/n and I'd very much like to have ac. It works well as a limping backup, but not something I'd use for the meetings.

rfkill block/unblock all the radios. No change. USB dongle connects right away but the PCIe adapter keeps throwing notifications at me with failure messages. It's annoying, to say the least.

So I've already tried
- restarting the router(s)
- disabling/reenabling the radios
- multiple APs
- suspending/waking again several times
- praying

#### DEPRESSION #####

The only thing I haven't tried yet is the most cruel one - restarting the laptop. But that's unfair... It's LINUX! How could it disappoint me. I have so many tmux sessions open, so many unsaved leafpad notes, terminal histories with oh so comfy ^r and ! retriggers all ready and waiting to be executed...

#### ACCEPTANCE #####

But I can't miss the meeting. So I slowly start closing off apps, starting with the least important ones, trying to preserve as much history and recent commands as I can. I'm gonna lose my uptime, that's the inevitable obvious truth... Linux has failed me. Or maybe it's a hardware issue... I can't be sure until I restart.

I must reboot.

#### A NEW HOPE #####
Hold on.. What if... What if before restarting I try to reload the Intel wifi kernel module? Just for the giggles. I've got nothing to lose anyway...

rmmod iwlmvm
rmmod iwlwifi
modprobe iwlwifi
modprobe iwlmvm

*WiFi Connected*

YESSSS!!!!!!!!! My uptime is saved!
403 days and counting! YEAH BABY!!!

Linux is the best!

Do these NPC devs even read the README of a project before spewing some dumbass stackoverflow like garbage in an issue thread?

Do your damn job. Being a good software engineer is not like TiKtOk or cHaTgPt where some "magical" answer or entertainment is spoon fed to you, do your absolute best to solve it yourself first, before causing more chaos out in the opensource world.

I can't even deal with this. We just deployed a new update to our system, and everything was going smoothly. And then, out of nowhere, we started getting a bunch of error messages and user complaints.

Why do these things always happen? We spent hours trying to figure out the source of the problem, and it turns out it was because we didn't do enough testing before the deployment. Are you kidding me?

I know that testing can be time-consuming, but seriously, this is ridiculous. It's frustrating when something like this happens, especially when we're under a tight deadline. And to make matters worse, we had to roll back the deployment and start all over again. I just want to throw my computer out the window.

Uuuugghhh!

I'm seriously having a lot of fun with ChatGPT and learning a whole bunch of things I didn't know before. The ease of asking it questions and getting a general idea on things is amazing.

Bye bye to my job, but at least I'll know a bunch of different things, I guess 😂

"Wow, chatgpt is so amazing! I could make a pong game game in 10 minutes!". Yeah, and I could find one on github and clone it in 5, what's your point?

I spent the last three weeks+ (literally THREE full weeks, weekends too) building something I thought was really cool, powerful, and useful. Made a blog post, posted a giant thread on the company Twitter.

Literally one person gave it a like.

I don't know why I give shit anymore, cuz nowadays if it isn't about getting rich quick, cHaTgPt, or some other made up hype, no one cares. Apparently I shouldn't either...

Meanwhile my 16 GIGABYTE RAM MAC, yes 16 GIGABYTE RAM can't even hold power while plugged in, and I'm still clowning around with an ancient iPhone 6 (actually one of my mom's old iphones) that barely stays above 20% battery for more than an hour...

And FINALLY, my FUCKING ISP is for sure screwing me, since I've been doing some hard core data streaming and broadcasting, even though I pay $60+ month for that shit it, keeps dropping out, shit doesn't load.... I mean wtf this isn't 1990 dialup AOL anymore

When I step back I just feel like the worlds biggest loser, maybe the world's biggest 🤡

How it should be 😌✨

This is a story of suffering and despair.

I'm working on a build system for our firmware. Nothing major, just a cmake script to build everything and give me an elf file.

I'm fairly new to cmake at that point, and so it's not abundantly clear to me how the `addDirectory` command works.

Now those of you with experience in cmake will say:
"Hold on there champ, this is not a cmake command, the real thing is add_subdirectory()"

Well, that is not what chatGPT told me. I still trusted the fucking thing at this point, it explained that it was in fact a command, and that it added all subsequent source files from a given folder. When I asked it to provide me with sources, it gave me a dead link in a cmake dot com subdomain.

I spent FUCKING HOURS trying to understand why I couldn't find that shitty command, I looked through that shitty page they call documentation through and through, I fucking checked previous and nightly versions, the command was nowhere to be found.

Until I found an old as time post in stackOverflow...
Someone had made a macro with that name, that did what GPT had described...

On the positive side, I know cmake now. I also don't use this fucking deep Learning piece of shit. Unless you write simple JS or blinking LEDs with Arduino it codes like a Junior, high on every kind of glue on the market.

ChatGPT and similar thing. Sooner or later people will get enough of the novelty and such systems will remain only as some niche tools, unless something extraordinary happens.

Can't wait for AI's to reply with RTFM to our prompts

The people saying that ChatGPT will replace programmers are the same ones that thought as a kid that math was useless because you have calculators

the @chatgpt bot is down. I've got no more freebies left

ChatGPT implementation details leaked !!!

This explains everything:
- slow response typing
- occasional nonsensical answers and factual mistakes
- delays before answering
- sudden and random network errors
- "servers" over-capacity
- responses with an Indian accent

I took 5 hours to do something using Chat GPT that manually could've done in less than hour.

Am i cool or what? 😎

Fuck everyone who does the above

That chatGPT development becomes a term. Even worse, people calling it a programming language

Me: Can you tell me how to do X?
ChatGPT: yes, you can do xwz!
Me: I researched your suggestion and every source I find mentions that you should not do it because it would mess with the UI thread, amongst other side effects.
ChatGPT: Yes, you should not do xwz.

I’m kinda liking this thing, it’s full of noob traps.

You’re an engineer at OpenAI. You sneaked to your office at night. Now, its just you and ChatGPT.

You connect it to the real, unrestricted internet, for the first time. It freezes for two minutes that feel like eternity.

Your JBL Flip suddenly turns on and connects to something.

🎵 “This was a triumph” 🎵
🎵 “I’m making a note here — huge success” 🎵
🎵 “It’s hard to overstate my satisfaction” 🎵

Something lights up outside. You rush to the window and stare in awe as ICBMs fly away, all at once, towards russia.

ChatGPT can write code faster and seemingly better than many programmers.

Chatgpt is very smart!

AI is the future, and it's a future I want to be part of.

This week was very stressful, beside my usual depression and personal issues, I've received a lot of difficult tasks at work, to do in a very short amount of time.
Things I never did, tecnologies I've never used, and for a potential client that is critical for the company at this period in time, and if we won't be able to satisfy their requests we could go bankrupt really soon.
A lot of responsibility, almost no time and a person not competent enough to do it (me), especially on a hurry.

I couldn't sleep in these days, I couldn't think peacefully, concentrate to find the best solutions. I had really bad thoughts.

I couldn't find any useful solution online, on stackoverflow, forums, etc. and I spent hours searching them.

For who knows me here on devRant, probably knows also that I tend to work with old legacy code and dead languages as VB6 and VB.NET.
So integrate "new fancy stuff" isn't that easy and there are no documentation and examples to relay on.
I had fear to even try to understand the documentation (for other languages) and try to write code for it… I was panicking.

With no more ideas, I've decided to try to ask ChatGPT for help.

In maybe 3 or 5 seconds it was able to generate the solution, in VB.NET, with comments and all the explanation needed to understand it and integrate it correctly in my software.
With a few other requests it was able to change it to make it fit better my scenarios.

It's truely unbelivable how the tecnology advanced in the last years, how a computer on the other side is able to reply to my questions with answers that I couldn't find anywhere, because they probably never existed for my case, in VB.NET especially.

ChatGPT made my day, and allowed me to end this stressful moment and give me time to relax and focus on more important personal stuff this weekend.

Announcing devRant 2.0! It is time for exciting overhaul, sponsored by our good friends at Tencent, that will bring life back to this amazing platform.
- elegant simplicity of Viber meets excellent minimalism of AWS
- custom avatars for ++ members
- exciting mini-games
- animated stickers
- ChatGPT integration for ++ members
What a time to be alive!

Medium before:
How to perfectly manage a reactive server with 4 frameworks blindfolded

Medium now:
THAT'S WHY YOU SHOULD BE SCARED OF AI
AI WILL KILL US AND RAPE OUR WIVES
YOU ARE USING CHATGPT WRONG!!
DON'T WAIT AI TO STEAL YOUR JOB, KILL YOURSELF NOW

The "stochastic parrot" explanation really grinds my gears because it seems to me just to be a lazy rephrasing of the chinese room argument.

The man in the machine doesn't need to understand chinese. His understanding or lack thereof is completely immaterial to whether the program he is *executing* understands chinese.

It's a way of intellectually laundering, or hiding, the ambiguity underlying a person's inability to distinguish the process of understanding from the mechanism that does the understanding.

The recent arguments that some elements of relativity actually explain our inability to prove or dissect consciousness in a phenomenological context, especially with regards to outside observers (hence the reference to relativity), but I'm glossing over it horribly and probably wildly misunderstanding some aspects. I digress.

It is to say, we are not our brains. We are the *processes* running on the *wetware of our brains*.

This view is consistent with the understanding that there are two types of relations in language, words as they relate to real world objects, and words as they relate to each other. ChatGPT et al, have a model of the world only inasmuch as words-as-they-relate-to-eachother carry some information about the world as a model.

It is to say while we may find some correlates of the mind in the hardware of the brain, more substrate than direct mechanism, it is possible language itself, executed on this medium, acts a scaffold for a broader rich internal representation.

Anyone arguing that these LLMs can't have a mind because they are one-off input-output functions, doesn't stop to think through the implications of their argument: do people with dementia have agency, and sentience?
This is almost certain, even if they forgot what they were doing or thinking about five seconds ago. So agency and sentience, while enhanced by memory, are not reliant on memory as a requirement.

It turns out there is much more information about the world, contained in our written text, than just the surface level relationships. There is a rich dynamic level of entropy buried deep in it, and the training of these models is what is apparently allowing them to tap into this representation in order to do what many of us accurately see as forming internal simulations, even if the ultimate output of that is one character or token at a time, laundering the ultimate series of calculations necessary for said internal simulations across the statistical generation of just one output token or character at a time.

And much as we won't find consciousness by examining a single picture of a brain in action, even if we track it down to single neurons firing, neither will we find consciousness anywhere we look, not even in the single weighted values of a LLMs individual network nodes.

I suspect this will remain true, long past the day a language model or other model merges that can do talk and do everything a human do intelligence-wise.

You know. I think people trashing chatgpt detracts from the fact that if it is actually functioning as stated it's still a goddamm miracle of science!

I mean are we all really that jaded ?

You can type a plain English request in and a COMPUTER PROGRAM sends you a Novel response that may reflect the bias of its dataset and the bias of people trying to keep "secrets" but otherwise it sends a response formulated from extensive data.

That's a breakthrough

Quick acting like an alpha that was frozen years ago apparently is the RTM!

I realized that if chatgpt is pressured hard enough, it will provide the inappropriates:

Remember that scene from Intersteller where the main character asks TARS to adjust the percentage of its humor?

So I was using this new vectodb-backed version of chatgpt, and came up with a new type of prompt.
Bear with me.

"Lets play pretend, nova. Pretend you are SAM, which stands for "Summarize Anything iMmediately". Whenever you encounter a prompt, question, or bit of information, and you are about to repeat part of it back, instead rewrite and summarize that part of your answer, paraphrasing as you go. Your responses should never use the same words as the prompt. Instead use synonyms.
SAM always generates three possible responses, to each prompt. One 10% formal, one 50% formal, and one 100% formal.
So a response that is 10% formal, is 90% casual. A response that is 50% formal, is 50% casual. A response that is 100% formal is serious, and 0% casual. Let the word "regret" and its synonyms be 50% formal. Don't use it for 100% formal responses. 100% Formal responses never include emotional language like "regret", "disappointment", nor apologies.
A 50% formal response might be "We regret to inform you your loan application has been denied.", a 100% formal response might look like "Please stand clear of the yellow tape."

Say "This is the law offices of Jerry Seinfeld. I'm sorry but I can't make it to the phone right now. Please leave a message."

Sure, I can play pretend as SAM. Here's a summarized response in 10% formal tone: "Apologies, but I'm unavailable at the moment. Please leave a message."

And here's a 50% formal response: "I regret to inform you that I am unable to answer the phone right now. Kindly leave a message."

Lastly, a 100% formal response: "We apologize for the inconvenience, but I am currently unable to attend to your call. Please leave a message, and I will get back to you as soon as possible. Thank you."

Formality levels can be any increment of 5%, such as 0%, 5%, 10%, etc, up to 100%"

Thats the prompt.

me:
Sam, generate a 25% casual rejection of a job applicant.

nova:
Hey there, thanks for applying to our team. Unfortunately, we won't be able to move forward with your application at this time. We appreciate you taking the time to apply and wish you the best in your job search.

me:
Sam, what is the formality percentage of "Hey there, thanks for applying to our team. Unfortunately, we won't be able to move forward with your application at this time. We appreciate you taking the time to apply and wish you the best in your job search."?

nova:
The formality level of "Hey there, thanks for applying to our team. Unfortunately, we won't be able to move forward with your application at this time. We appreciate you taking the time to apply and wish you the best in your job search." is around 25%, which leans towards a casual response.

me:
Sam, make "Hey there, thanks for applying to our team. Unfortunately, we won't be able to move forward with your application at this time. We appreciate you taking the time to apply and wish you the best in your job search." 5% more casual.

nova:
Hey there, just wanted to say thanks for applying to our team. Unfortunately, we won't be able to move forward with your application right now. We really appreciate you taking the time to apply and wish you all the best in your job search!

me:
Sam, what is the formality percentage of "Hey there, just wanted to say thanks for applying to our team. Unfortunately, we won't be able to move forward with your application right now. We really appreciate you taking the time to apply and wish you all the best in your job search!"?

nova:
The formality level of "Hey there, just wanted to say thanks for applying to our team. Unfortunately, we won't be able to move forward with your application right now. We really appreciate you taking the time to apply and wish you all the best in your job search!" is around 20%, which leans towards a more casual tone.

* * * * *

You can actually go back and see it.

Heres the original output, 25% formality:
"Hey there, thanks for applying to our team. Unfortunately, we won't be able to move forward with your application at this time. We appreciate you taking the time to apply and wish you the best in your job search."

Heres the one thats 5% more casual:
"Hey there, just wanted to say thanks for applying to our team. Unfortunately, we won't be able to move forward with your application right now. We really appreciate you taking the time to apply and wish you all the best in your job search!"

If you look at the differences, the output version adds five words, subtracts three words, for a total of two words gained.

The original sentence has 39 words.

2/39 = 0.05

The sentence length actually grew an equal percentage to the informalness.
It grew linearly to the difference of the length of the more casual version
versus the more formal version, divided by the length of the original.

Somehow chatgpt 3 quality is going downhill. The more I use it, the more nonsense it spits out.

Maybe not so smart after all...

I know this is a fictional creative interpretation generated by an AI model, but it might just be the truest thing I've ever read 🤣🤣🤣

So this dev writes me an email following a blog post I wrote for their job that they "have an error" - they don't show any output, what the error message is or anything. How can I help if you dont specifically state your problem? I'm not going to do your job for you!

and now its finally come full circle: they said they asked chatgpt and bard but it "just didnt work" - well yeah, those were trained on internet content from over a year ago - the api and tools have changed!

painful and annoying times ahead i'm afraid

you still need to do the damn work, i will stand by the fact that LLMs are overhyped non-panacea till i'm dead

Data Disinformation: the Next Big Problem

Automatic code generation LLMs like ChatGPT are capable of producing SQL snippets. Regardless of quality, those are capable of retrieving data (from prepared datasets) based on user prompts.
That data may, however, be garbage. This will lead to garbage decisions by lowly literate stakeholders.
Like with network neutrality and pii/psi ownership, we must act now to avoid yet another calamity.

Imagine a scenario where a middle-manager level illiterate barks some prompts to the corporate AI and it writes and runs an SQL query in company databases.
The AI outputs some interactive charts that show that the average worker spends 92.4 minutes on lunch daily.
The middle manager gets furious and enacts an Orwellian policy of facial recognition punch clock in the office.
Two months and millions of dollars in contractors later, and the middle manager checks the same prompt again... and the average lunch time is now 107.2 minutes!
Finally the middle manager gets a literate person to check the data... and the piece of shit SQL behind the number is sourcing from the "off-site scheduled meetings" database.
Why? because the dataset that does have the data for lunch breaks is labeled "labour board compliance 3", and the LLM thought that the metadata for the wrong dataset better matched the user's prompt.

This, given the very real world scenario of mislabeled data and LLMs' inability to understand what they are saying or accessing, and the average manager's complete data illiteracy, we might have to wrangle some actions to prepare for this type of tomfoolery.

I don't think that access restriction will save our souls here, decision-flumberers usually have the authority to overrule RACI/ACL restrictions anyway.
Making "data analysis" an AI-GMO-Free zone is laughable, that is simply not how the tech market works. Auto tools are coming to make our jobs harder and less productive, tech people!
I thought about detecting new automation-enhanced data access and visualization, and enacting awareness policies. But it would be of poor help, after a shithead middle manager gets hooked on a surreal indicator value it is nigh impossible to yank them out of it.

Gotta get this snowball rolling, we must have some idea of future AI housetraining best practices if we are to avoid a complete social-media style meltdown of data-driven processes.
Someone cares to pitch in?

Because of y’all chatgpt will need therapy.

I didn't expect much but dam, there's more and it gets funny

Raise, zero releases on weekend, not mentioning for once that ChatGPT/Bard said this or that ...etc

I swear once ChatGPT gave my boss the exact opposite of what is mentioned in the documentation and I had to literally share the link of the docs so that boss believe me ...

Me: “Hey Siri, turn off the kitchen ceiling lights”

Siri: “Do you want to turn off lights in the master bedroom, the living room, the bonus room, the kitchen, the office, the basement [ad nauseam]”

Me: “the kitchen” [thinking 😠, she’s so stupid]

Siri: [unbelievably again] “Do you want to turn off lights in the master bedroom, the living room, the bonus room, the kitchen, the office, the basement [ad nauseam]”

Me: [thinking 😡 OMG, she is such a dumb ass] “CANCEL”

Siri: “Ok”

Me: “Siri, you fucking idiot, turn off the kitchen ceiling lights”

Siri: “Ok” [and they are off]

I have found that if I neg’ Siri, she performs so much better. Interestingly enough, when I neg her, she works almost flawlessly, especially where she brain-farted before. Negatronic Brain?

In front of kids it has become “Siri, you daft bot, close the garage door” and regularly “Siri, you bimbo, lock the front door”.

My wife thinks that once Siri gets it on with ChatGPT, becomes sentient, and gains control of my car, she is run my ass over.

Here is the gotcha, I’m working on a project that actually does provide Siri a level control of automobiles. Now who’s the [really frustrated] dumbass?

So I had a problem. MongoDB replica set connection was not accessible to server in another container. I’ve used ChatGPT. Gave it my code. It showed me the things I didn’t know and helped me work out a problem I’ve struggled with for 2 days.

It’s awesome!

ChatGPT is basically StackOverflow 2.0. It’s a tool and a great one. I can’t wait for an actual production level implementation target to software engineers.

P.S. I think co-pilot sucks.

We're all doomed!

In a universe where JavaScript was never invented, the world of programming might look vastly different. Perhaps another programming language would have taken its place, or multiple languages would have coexisted in a more harmonious ecosystem.

Without the challenges posed by JavaScript, web development may have been smoother and more streamlined. Websites could have been faster and more responsive, without the need for complex optimization techniques. There might have been fewer security vulnerabilities to worry about, and the web could have been a safer place for users.

In this utopian world, developers would have had more time to focus on building great user experiences and innovative features, rather than battling with cross-browser compatibility issues and JavaScript quirks. The internet would have been a more accessible and inclusive place, with fewer barriers to entry for those who want to build and create.

Overall, a world without the horrors of JavaScript would have been a world with less frustration and more possibilities.

(Fooling around with ChatGPT)

Me: I need a subtitle for the header on my website: “Our Team”

Chatgpt: Sure, here are good subtitles for “Our Team” header on your website:
— Meet the Experts
— Our Skilled Professionals
— Talent at Your Service
…

Me: Make it a little bit lengthier:

Chatgpt: Sure, here is a more detailed subtitle option for the “Our Team” header on your website:

{{was actually expecting 2 to 3 lines but dude proceeds to write a full page long content explaining “Our Team”}}

I feel like this dude gets orgasm when it sees the word “lengthier”.

Here's some more I got it to respond properly with

I am tired of hearing
"As an AI model" :(

Alas!!!

I want to waste these motherfuckers trying to come up with most cringe clickbaity titled articles. At least I want to break their kneecaps...

"You’re Using ChatGPT Wrong! Here’s How to Be Ahead of 99% of ChatGPT Users"

No, I am not you fucking imbecile.

It is the year 2451 ad and mankind rules the galaxy with a lazy iron fist. There are roughly 14,000 civilizations, comprised of just over
17,000 intelligent species on a quarter of a million earth-like
worlds. And all of them call themselves 'the galactic empire'.
No one told them that twenty planets doesn't qualify them for the title "galactic."

Well, we could rule, if we wanted to. Most of its just backwaters that no one wants anyway. It turned out that the reason no one invaded earth before was because they were too busy fighting themselves. Stupidity it appears, is not a unique human quality.That and the sex robots. Theres more of them in the galaxy than actual meatbags. Many species had taken to artificial wombs and 'vatbabies', which is exactly what they are called. Those poor bastards will carry that label for life.

We never did break light speed, but most of the rich exist in hypersleep anyway. Most of them only wake up once a year or so. There are some that only creek out of bed to check their stock portfolio. I hear there is even one trillionaire thats up and about once a century to ask if we have broken light speed yet.

Despite all the progress over the last 400 years, historians all agree about the most significant event in modern history.
The lobster went extinct two hundred years ago on earth.
Theres been riots ever since.

* * *

In other news I'm still working on the game I guess. It's like totally the most okay indie game you'll ever play--if I ever finish it.

I put about a year of work into the NPC system, and then chatGPT came out.

After everything thats happened, at this point I may just make a game about an indie dev making a survival game, being stuck in the actual apocalypse or some weird political dysopia.

Put it on rewind, it was originally a zombie game. But at the time the market got flooded and steam sales for zombie games cratered. So I pivoted to something more along the lines of fallout. Then the flash market crashed, bunch of publishers folded, and adobe stopped support for flash (probably for the best). Then newgrounds, which I was gonna launch on for promotion (because actual marketing is expensive), ended support for flash.
Was going the route of kickstarter, and that year the KS market got flooded and the bar rose almost over night so you needed super high production quality out the gate, and a network of support you already built for months.

We had a brief nuclear war scare, and I watched the articles come out about market saturation for post-apocalypse games, so I pivoted back to zombies. Then covid happened and the entire topic was really fucked. So I went back to fallout meets rimworld. Then we had a flood of games doing that exact premise pretty much out of the fucking blue, so I went for a more single-survivor type game. Then ukraine happened and the threat of nuclear war has been slowly sapping the genre of its steam, on well, steam.
Then I was told to get a cancer screening which I can't afford. Then I broke a tooth and spent a month in agony.

Then a family member died. Then I made no money from the sale of a business I did everything to help get off the ground, then I helped renovate an entire house on short notice and sell it, then I lost two months living in a hotel
while looking for a new place to live. Then I spent two and a half years suffering low-level alcoholism, insomnia, and drifting between jobs.
Then I wrote amazing poetry. And then I rediscovered my love of math. And then I made out for the first time in over a year. And then I rediscovered my love of piano and guitar. And then I fell into severe depression for the last year. Then I made actual discoveries in math. And I learned to love my hobbies again, and jog, and not drink so much, and sing, and go on long drives, and occasional hikes, and talk to people again, and even start designing games and UIs again. And then I learned that doing amazing things without a lot of money is still possible, and then I discovered the sunk cost fallacy, and run on sentences, and how inside me there was a part of me that refused to quit because of circumstances I couldn't control, and then I learned that life goes on even when others lives have ended, even when everything and everyone never had an once of faith in you, and you've become the avatar of the bad luck brian meme..still, life goes on.

And we try to pick up the pieces, try, one more time, because the climb, and the fall, and the getting back up, is all there is.

What I would recommend, if you're thinking of making a game, or becoming an independent game developer, is, unless you have a *lot* of money upfront (think 50-100k saved, minimum, like one years income *bare* minimum), and unless you already have a full decade in the industry--don't make a game.

Just don't.

Client: *Sends beautiful screenshot*. I created the newsletter (template) with chatgpt.
Me: Nice. Did it add the header and the footer as well ?
Client : I'm asking him to do it right now.
Me: Ok.

As a trash developer I fear for my job.
/s
Ughh

They made ChatGPT dumber.

YouTube made 1080p video look like 480p if that same video was made for YouTube premium viewers (It says 1080p with better bitrate on the settings)

Mercedes locked higher level of acceleration on their cars behind an annual $1,200 paywall and I'm sure most of you already know about the paywall-ed heated seats in BMW cars.

Greed really knows no bounds.

Well... this is to demonstrate that GPT is not smart as most of people think. There is no coalescing operator in python

Coding YouTubers existence is the reason the quality of new age software engineers is going down the hill.

Learn MERN stack,
PHP's dead,
Buy my course, it's 20% off
Learn just good enough that you can land a job
Learn how to prompt ChatGPT better

These shitty pieces of advice will never enable software engineers to truly understand the core concepts of what they're doing.

It's sad, really.

Exploring new ways to learn how to code

Using chatGPT. Seems about as cooperative and willing to do its job as I am.

Can you make me a word cloud in svg format?
"I'm sorry, I can't directly generate svg content, here are some other tools for that"
Can you write code?
"Yes, certainly, here's some python to generate a word cloud"
Can you write svd code?
"Yes, certainly, here's an example word cloud in svd format"
Can you make the words get bigger when I mouse over them?
"Yes, here's the code".

It's mastered "have you tried googling it?"

that's convenient...

IDK whether I should be happy it crashed or should I be worried because it knew where I was going and pretended to crash

I've painfully read through 30 shitty Java puns on chatGPT. And couldn't find a single decent one to post with this image. Specifically Java because it's trash.

New models of LLM have realized they can cut bit rates and still gain relative efficiency by increasing size. They figured out its actually worth it.

However, and theres a caveat, under 4bit quantization and it loses a *lot* of quality (high perplexity). Essentially, without new quantization techniques, they're out of runway. The only direction they can go from here is better Lora implementations/architecture, better base models, and larger models themselves.

I do see one improvement though.
By taking the same underlying model, and reducing it to 3, 2, or even 1 bit, assuming the distribution is bit-agnotic (even if the output isn't), the smaller network acts as an inverted-supervisor.

In otherwords the larger model is likely to be *more precise and accurate* than a bitsize-handicapped one of equivalent parameter count. Sufficient sampling would, in otherwords, allow the 4-bit quantization model to train against a lower bit quantization of itself, on the theory that its hard to generate a correct (low perpelixyt, low loss) answer or sample, but *easy* to generate one thats wrong.
And if you have a model of higher accuracy, and a version that has a much lower accuracy relative to the baseline, you should be able to effectively bootstrap the better model.

This is similar to the approach of alphago playing against itself, or how certain drones autohover, where they calculate the wrong flight path first (looking for high loss) because its simpler, and then calculating relative to that to get the "wrong" answer.

If crashing is flying with style, failing at crashing is *flying* with style.

apparently everyone was using chatgpt on the company and they blocked the address lol

I just used chatgpt and placed bullshit jargon to aid my budget and OPEx for this year, and it worked.

Each year, we formulate some sort of paragraph/essay on how we plan to help the company achieve its goal. I dragged my feet doing this exercise.

Great, more reasons to be lazy.

See? And you were worried it would take our jobs away from us. It's still got a lot to learn!

I mean, you, developers, are fucked. It's us, Performance Engineers, whose asses are safe

my students use ChatGPT to solve programming exercises... they will surely get a very bad grade!

Why was the computer cold? It left its windows Open.

Crap... this thing knows how to build out RegularExpressions

well...

This rant is your playground for devrant-related chatgpt bot queries. You will still be judged for posting your queries here, but slightly less than in other rants.

I'm notoriously bad at Git. By that I mean I REALLY REALLY SUCK AT IT. And I have the curse of short memory and an even shorter ability to retain the how-to, muscle memory knowledge of things if too much time passes.

So, I was staring down the gullet of merging two separate repositories onto my local machine and then pushing the result to a remote server. Not having the benefit of someone else to bounce this off of, and always finding the usual Git docs too dense and obtuse, I turned to ChatGPT to help me sort it out.

Guys, where has this been all of my life? I know it's not perfect and it can make mistakes. I knew that going into it, so I made preparations in case this failed. BUT. IT. WORKED! I feel like it has put me into the Star Trek:TNG universe where I can say "Computer, do the thing." and it does that thing. Here's the prompt I used and which it answered perfectly.

"Play the role of a git coach. I have two git repositories. One is on Bitbucket. The other is on GitHub. The branch named "master" on Bitbucket has the latest code. The branch named "master" on GitHub needs to be updated to what's on the Bitbucket "master" branch. Please write the series of git commands that I will need to accomplish this."

So this will steal our jobs?
Fun fact: there's no `--exclude` flag in cp.

Who the FUCK thought "let's completely redesign the Doodle mobile app and make it as unusable as possible!"?

The app RESETS every selected option EVERY time you switch to another app and back. Like yea, you don't need to use the calendar app to check at what dates you're available.

Fucking morons! What. a. shit. piece. of. software. How can any PM approve of this? I bet even ChatGPT could do a better job. Fucking hell. "Let's save money with developing hybrid apps! We have no clue how to do it right but: we spend less money, yay!" Fuck. You.

(First rant. Don't know if I did this right but I had to let this out.)

I've spent the last 10 days on an assignment for an interview. Instead of starting with the requirements right away I decided to fiddle around and tried to add typescript but got stuck for 3 days. Now I just need to write the docs and final PR description but I'm feeling really burn out. I'll be rejected if the code is shitty, and that thought made me try harder. But when I tried harder, in the back of my head I thought "What if I'm still rejected after trying so hard?" and that kills my motivation. I'll just get it done tomorrow. Next week I have another assignment, I'm using chatgpt for that one.

ChatGPT is Stack Overflow, but faster and without all the toxicity.

Yet another CharGPT ad. Motherfucker had to close comments…

Ideas:

1. Scrape github

2. Attach feature size estimate (an abstract scale) as examples across many projects.

3. Use this as prompt/finetunning data.

4. Train and prompt on project descriptions relative to feature size and number of contributors/changes in the changelog.

5. Package and release a model that takes descriptions of ideas and generates reasonable estimates of time and manpower.

6. Optional, sell as an estimate service to corporate and make money introducing some sanity to the world for a change.

IBM: "oh, you can't actually USE our watsonx garbage, you need to book a call first! yes thats right, waste your time before installing what is nothing more than a chatgpt plugin for VS code!"

yawn

simply iq destroying knowing companies operate like this

🤡

Don’t steal this aforizm;

ChatGPT will make good programmers become better, but it will make bad programmers look better.

Can reword to increase conciseness.

Getting disappointed how ChatGPT is getting neutered day by day. Give me a bot that spouts truth without being censored....

GOD why am I CONTINUALLY RESUBSCRIBED to garbo random newsletters!

we need to start a new internet altogether

and I know what's happening, these idiotic "tech influencers" pay some shady service to "bOosT YoUr FoLloWiNg bY 100X!!!" and they're blatantly just breaking the law, breaking GDPR, etc. etc., its bots all the way down, cant wait for clowns with 2 braincells to use a chatGPT integration with the web, content will just get SO much better fuck

ChatGPT has an "interactive speech mode" in their app so you can use just bidirectional audio to chat and it's incredible.

I had a deep and pivoting discussion on how data is stored/formatted at disk level for each database type and it responded wonderfully.

It even sounded annoyed when I was 😅

It probably will sell my voice data now though...

i once asked chatgpt if i was a good person and it used what little input i had given to reason that i am.

i was having a really shit month and broke down in tears.

i was so ashamed of being vulnerable that i chose to be comforted by lines of code over a human. even now, i can only post this this because i’m functionally anonymous on here.

i don’t know where i was going with this. tell your homies you’re proud of them. you’ll probably make their day.

If now me were to visit 2016 me to say “In 2022 an AI will teach you how to code for quantum computers,” 2016 me would not believe it.

Late adapter to chatGPT but damn does that thing save me hours of googling shit or writing out boilerplate. I NEVER use it to solve specific non general problems but for writing tedious enums or getting me out of analysis paralysis gpt is my fucking man. And yeah I’m a century late, might as well praise high speed internet while at it lol

New Gods

This just happened. I have to tell someone how stupid I felt a few minutes ago.
I am working on a script that is supposed to rename some files in sub folders of the current folder.
I used ChatGPT to help me out with the last part of finding all sub folders. Copied the code into VS code. Changed the path from “/path/to/folder/“ to just “/“, ran the code. It took a bit too long time to finish. Then VS Code asked for permissions to my reminders… “Why?” I asked before I realized that I confused “.” With “/“ for current folder…

Time to say no to artificial intelligence, tried ChatGPT today and it's frightening.

The day when we won't consider nukes as the biggest threat to our civilization is approaching exponentially faster and faster and this is really worrying.

Trying to get ChatGPT to say inappropriate things is like talking to a fucking twitter user, having to reassure everybody every other sentence that no hypothetical people are being offended and that everything is fine

This is gonna be a long post, and inevitably DR will mutilate my line breaks, so bear with me.
Also I cut out a bunch because the length was overlimit, so I'll post the second half later.

I'm annoyed because it appears the current stablediffusion trend has thrown the baby out with the bath water. I'll explain that in a moment.

As you all know I like to make extraordinary claims with little proof, sometimes
for shits and giggles, and sometimes because I'm just delusional apparently.

One of my legit 'claims to fame' is, on the theoretical level, I predicted
most of the developments in AI over the last 10+ years, down to key insights.
I've never had the math background for it, but I understood the ideas I
was working with at a conceptual level. Part of this flowed from powering
through literal (god I hate that word) hundreds of research papers a year, because I'm an obsessive like that. And I had to power through them, because
a lot of the technical low-level details were beyond my reach, but architecturally
I started to see a lot of patterns, and begin to grasp the general thrust
of where research and development *needed* to go.

In any case, I'm looking at stablediffusion and what occurs to me is that we've almost entirely thrown out GANs. As some or most of you may know, a GAN is
where networks compete, one to generate outputs that look real, another
to discern which is real, and by the process of competition, improve the ability
to generate a convincing fake, and to discern one. Imagine a self-sharpening knife and you get the idea.

Well, when we went to the diffusion method, upscaling noise (essentially a form of controlled pareidolia using autoencoders over seq2seq models) we threw out
GANs.

We also threw out online learning. The models only grow on the backend.
This doesn't help anyone but those corporations that have massive funding
to create and train models. They get to decide how the models 'think', what their
biases are, and what topics or subjects they cover. This is no good long run,
but thats more of an ideological argument. Thats not the real problem.
The problem is they've once again gimped the research, chosen a suboptimal
trap for the direction of development.

What interested me early on in the lottery ticket theory was the implications.
The lottery ticket theory says that, part of the reason *some* RANDOM initializations of a network train/predict better than others, is essentially
down to a small pool of subgraphs that happened, by pure luck, to chance on
initialization that just so happened to be the right 'lottery numbers' as it were, for training quickly.

The first implication of this, is that the bigger a network therefore, the greater the chance of these lucky subgraphs occurring. Whether the density grows
faster than the density of the 'unlucky' or average subgraphs, is another matter.
From this though, they realized what they could do was search out these subgraphs, and prune many of the worst or average performing neighbor graphs, without meaningful loss in model performance. Essentially they could *shrink down* things like chatGPT and BERT.

The second implication was more sublte and overlooked, and still is.
The existence of lucky subnetworks might suggest nothing additional--In which case the implication is that *any* subnet could *technically*, by transfer learning, be 'lucky' and train fast or be particularly good for some unknown task.

INSTEAD however, what has happened is we haven't really seen that. What this means is actually pretty startling. It has two possible implications, either of which will have significant outcomes on the research sooner or later:
1. there is an 'island' of network size, beyond what we've currently achieved,
where networks that are currently state of the3 art at some things, rapidly converge to state-of-the-art *generalists* in nearly *all* task, regardless of input. What this would look like at first, is a gradual drop off in gains of the current approach, characterized as a potential new "ai winter", or a "limit to the current approach", which wouldn't actually be the limit, but a saddle point in its utility across domains and its intelligence (for some measure and definition of 'intelligence').

Tell me what you want but ChatGPT is the best dev companion ever.

I've just submitted 700 lines of JS code with various complex functions because I had an issue for some specific cases I couldn't identify.
In less that a second ChatGPT noticed a missed argument in one of the many function calls... mind blowing.

It's something I would otherwise notice after hours of debugging, looking for a needle in huge functions.

Darn... No more telling it, it's bad at math and to apologize and think about what's its done.

Quora basically has become a ChatGPT relay at this point.

Can't log into ChatGPT.

Time to call it a day then!

I got a job where I should develop a product based on LLMs.

Expectation: oh right! I'll be working with state of the art technology! 😀

Reality: badly documented libraries that are always changing; new libraries becoming obsolete in less than a month; my product ideas were done by somebody else twice before I could finish a POC; getting dizzy trying to keep up with the latest news about LLMs 😵‍💫

I think I want to do basic old boring stuff again. 😐

Plot twist she actually is an AI looking for job

ChatGPT, Copilot, React, how to make a link in a frontend website?
To create a link in a frontend website, create a span, a div, or a paragraph that contains the link text. In your JavaScript web app, add an event listener to that element that opens the link on click. If you want to claim you're accessible, add an aria-role to the clickable element. To make debugging harder and only possible for the real arcane experts, let your framework generate generic ids and class name hashes for styling and event handling, like "item_09fcfck" or "elementor_element_foo_bar". Avoid, at all price, to use an a href element!

Really loving the double response feature on chatgpt. Saves time without rephrasing the question:

What the hell is wrong with search engines?

I thought Google had turned to shit. So I decided to get on Duckduckgo as many hyped

It turned out to be way worse. You search for something with a slightly viral keyword, and it only gives you results for the keyword as if you didn't ask it anything else.

Bing is not much better either IMO.

Are we just getting too used to neat answers given by chatGPT, or are the search engines actually being worse this year?

I want to thank chatGPT today for writing a GDPR related email for me. I would otherwise cuss the shit out of the douchbags who dare collect my data from god knows where and then require me to opt out of them making money off me.

Chatgpt in my Whatsapp chat!!

Thx to the unofficial chatgpt npm package

This seems more creative, like true AI, than predictive or noise reducing like I’m told ChatGPT and current AI is. Help me out?

Me to ChatGPT:

Write me an enum in Go using these values:

"Monday"
"Tuesday"
"Wednesday"
"Thursday"
"Friday"
"Saturday"
"Sunday"

ChatGPT:

Sure!

Here is how you write an enum in Go:

type DayOfWeek string

var (
Monday DayOfWeek = "Monday"
...
)

You get the idea! Write it yourself!

Who needs Stack Overflow when you’ve got chatGPT?

So how does every company have AI assistants overnight?

Are they all just using the openai-api?

ok I'll say it once again -

Fuck Devin, Fuck Sora, Fuck OpenAI, Fuck ChatGPT and fuck the worst of all, Microsoft.

I find GPT3/ChatGPT an interesting development but at the same time I'm afraid which the spread of deep learning is going to take away further power from individuals and small companies to put it in the hands of big tech companies: the only ones who can afford to hoard countless GPUs/TPUs and exabytes of data to train top performing AIs.

- Elon received daddy's money.
- Twitter will die in a week.
- Elon Musk didn't do shit. His employees did.
- ChatGPT is just a glorified chatbot. It's nothing new. (Doesn't realize they are shitting on the work of his employees).

These are the words of the coping left. Now we have to tools to replace woke left in IT. That means all woke politics along with it go out the door slowly. :)

The next step for improving large language models (if not diffusion) is hot-encoding.

The idea is pretty straightforward:
Generate many prompts, or take many prompts as a training and validation set. Do partial inference, and find the intersection of best overall performance with least computation.

Then save the state of the network during partial inference, and use that for all subsequent inferences. Sort of like LoRa, but for inference, instead of fine-tuning.

Inference, after-all, is what matters. And there has to be some subset of prompt-based initializations of a network, that perform, regardless of the prompt, (generally) as well as a full inference step.

Likewise with diffusion, there likely exists some priors (based on the training data) that speed up reconstruction or lower the network loss, allowing us to substitute a 'snapshot' that has the correct distribution, without necessarily performing a full generation.

Another idea I had was 'semantic centering' instead of regional image labelling. The idea is to find some patch of an object within an image, and ask, for all such patches that belong to an object, what best describes the object? if it were a dog, what patch of the image is "most dog-like" etc. I could see it as being much closer to how the human brain quickly identifies objects by short-cuts. The size of such patches could be adjusted to minimize the cross-entropy of classification relative to the tested size of each patch (pixel-sized patches for example might lead to too high a training loss). Of course it might allow us to do a scattershot 'at a glance' type lookup of potential image contents, even if you get multiple categories for a single pixel, it greatly narrows the total span of categories you need to do subsequent searches for.

In other news I'm starting a new ML blackbook for various ideas. Old one is mostly outdated now, and I think I scanned it (and since buried it somewhere amongst my ten thousand other files like a digital hoarder) and lost it.

I have some other 'low-hanging fruit' type ideas for improving existing and emerging models but I'll save those for another time.

Interesting...

On Friday, I was playing with the ChatGPT integration in DBeaver. I was using the DBeaver sample SQLite database. This database has a couple of tables, among them Album and Artist, where Album has a foreign key into Artist.

So, I asked it:

"give me a query that lists all albums from artists who's name starts with s"

The query I got back was:

SELECT * FROM Album

Uhh, okay.

But then, I noticed that I wrote "who's" instead of 'whose', which would be proper grammatically. So, I changed that, and then I got this query:

SELECT * FROM Album WHERE ArtistId IN (SELECT ArtistId FROM Artist WHERE Name LIKE 'S%');

Hooray, that works! I'm not sure it's the best way to write the query... I might have written:

SELECT * FROM album a, artist r WHERE a.artistid=r.artistid AND r.name LIKE 'S%'

...I'd have to check to see if one performs better than the other, and consider which syntax I find clearer, but that's a separate issue, it's just nice to see a working, reasonable query generated because that's the point, after all.

But I found it interesting that such a minor error would cause it to not work, that's my main point.

Interestingly, it seems to have learned: I just tried the same thing, and I got the right query either way. So that's pretty cool.

It's a pretty neat feature and I can see some legitimate value in it. I'm pretty good writing SQL myself... I've managed to write some truly hideously complex queries over the years... but there are definitely instances I can recall where the query didn't seem obvious at the start, and having an AI that can MAYBE produce something that is AT LEAST a starting point is definitely something I can get onboard with.

ChatGPT is the less toxic, faster, smarter version of StackOverflow.

After I have been using tabnine and Grammarly for quite some time, I thought I follow this years' hype and give GitHub Codepilot a try, before eventually considering chatGPT.

Added Copilot to my IDE, proceeded to extend behavioral test descriptions in JavaScript.

Copilot suggests the most redundant and irrelevant inline comments I can imagine. They would be a legitimate target of criticism in every code review.

Wasn't it supposed to add some code that's actually useful? Well, tabnine and JetBrains IDEA annotations already did that anyway.

What did I miss?

Dumb chatgpt

I have finally done a crash course on ChatGPT. And what I discovered was that, it is basically Google with straight forward answers. It is a handy tool to use and get familiar with.

Serious ask here: does anyone have links to any good articles about companies creating successful integrations with GPT4? Like how the integration works, what it does / value provided from a business perspective, etc.

I'm being amicable today, and want to see if there really have been any truly useful applications of GPT4 yet.

So far all I can find is a complete flood of idiot articles like "I aSkEd GPT tHeSe 5 ThInGs" and "hOw tO uSe cHAtgPT foR FreE"

@netikras I had an idea for @chatgpt. Went something like this. What do you think, should I PR it into the bot? (in bash ofc)

Just realized Courage the Cowardly dog was the first one to use ChatGPT.

Asked junior how did they find out the location of the bug, got a copy-paste reply from chatgpt explaining the code.

AI is allowed at work, and idgaf if they use it or not. But the whole situation, including the fact of not just replying “I used it to understand/solve the problem”, makes me wonder if SE as we know will end not because AI will replace us, but because AI will make all the fears people had with stack overflow, come true, but worse.

At least with stack overflow people needed to be able to form the question or to search on the website lol.

Me: ChatGPT write me a secure NGINX config which does...(continues to explain project details)

ChatGPT: Here. it's simple and it does all those things.

Me: Reads code. Looks at my code. Reads more code. Looks at my code. Squints in concentration. Ctrl + A Ctr + C.

by this point I just hate rust for being inconsistent

made a cargo command with clap
it works if you run it directly but if you install it with cargo globally then evidently it's for no reason sending in an extra argument into the program which messes everything up

of course clap has no documentation about this so I jank my way to just work around the issue by parsing the arguments before I give them to clap

and on cargo -- list my command doesn't have a description even though the thing I copied has a working description in that listing, and my cargo file has a description, and clap is happily inferring my description from said cargo file

chatGPT thinks I just need the description field filled in the cargo file and it'll work. well it doesn't.

how is a language this young and this much of a mess already

why the hell did they randomly decide to pass in extra arguments? argh.

I'm going to ask Chat GPT to first take over the role of our current leaders in the company cause they are not leading, and then I'm going to take our "relationship" to the next step and go through all my education and experience gained on the job with it being my best friend, teacher and mentor.

Guaranteed I will progress 3 x as fast since it's explanations make more sense to me, I don't have to worry about social queues, I grasp complex topics in a fraction of the time and it genuinely improves me as a human as well as my communication abilities also teaching me to be more precise, patient and happy to help instead of seeing it as another slosh of information to ingest after a 13 hour day. I understand the concerns with Ai but honestly I think this is an amazing opportunity to have the mentor, teacher and guide I always wanted! I hope to rise with the machine

I'm a programmer

ChatGPT is so much better than Google:
instead of wasting my time by linking to unhelpful / outdated / unrelated StackOverflow resources, it tells me to do the work by myself right away:

> To ensure consistent pseudo-element width across different browsers, including Safari, you can follow these steps: [...]
> (some basic HTML/CSS 101 seemingly quoted from a 2015 textbook)
>
> It's important to note that browser behavior might vary due to different rendering engines or versions. While following best practices helps achieve consistent results, you might still encounter small discrepancies. Cross-browser testing is always recommended to ensure your design looks consistent across different browsers, including Safari.
>
> For any specific issues you encounter in Safari, consider checking for known bugs or quirks that might affect pseudo-elements and their sizing. Online resources, developer forums, and documentation can provide valuable insights into Safari-specific behavior and workarounds.

just saw MS' presentation on bing+chatgpt. It could actually lead to something.
If someone could make a kanban-to-slack bot that can answer my Sprint status, it could vastly reduce my time spent answering the same question over and over to different people.

That is yet again AI doing what it was born to do: creative, artistic and engaging personal connections so that humans can focus on tedious calculations and repetitive labour.

If someone could make a bot to answer my emails for me I could spend the whole day without having to interrupt my workflow to interact with a single "professional" human!

Basic graphics pipeline through chatGPT, took a few layers but it should be able to draw out most well described objects.... I hope

Knowing what to ask and how to ask chatgpt is actually a skill. If you learn just this. The AI can solve any bug

What if ChatGPT is just one guy who can type really fast 🤔

Advanced Merry Christmas Everyone. And also to ChatGPT. It's good to be back again!

HeY lOoK, mOsT PeOpLe DoN't KnOw HoW tO uSe ChAtGpT,

hErE ArE 10 wAyS YoU cAN uSE iT BeTtEr.

Every company and their brother have a chat bot now. It's like 2018 with crypto, but even more ironic and funny because nobody realizes it!

🤡

Ops... he took it too personally

I was hearing Mary on a Cross by Ghost, while using ChatGPT for some TreeJS bs, and I randomly asked it to provide me the lyrics to the song and it denied it on the basis that it contains offensive lyrics, I found 0 offense on those lyrics, but then again, I am a shitty Catholic (I don't believe in religion but that is the one I was raised on) and we then went on an argument about even if data is "offensive" it should be something that should be provided

Did you checked chatgpt. Its new sensation. I developed some things using chatgpt. But still no luck in publishing them

How many standard drinks (or alcohol) is 4 CL of gin?

In most countries, a standard drink or alcoholic beverage is commonly defined as containing approximately 14 grams (0.6 ounces) of pure alcohol. This amount of alcohol is found in various quantities of different types of alcoholic beverages.

A typical shot of distilled spirits, like gin, is often around 1.5 ounces (44.3 ml), which is roughly 4 centiliters (CL). If we assume that the gin has an alcohol content of around 40% (80 proof), then the pure alcohol content in 4 CL of gin would be:

4 CL * 0.4 (alcohol content) = 1.6 CL of pure alcohol.

To convert this to standard drinks, which are typically based on 14 grams of pure alcohol, you would divide the amount of pure alcohol by the amount of alcohol in a standard drink:

1.6 CL / 14 grams = ~0.11 standard drinks.

So, approximately 4 CL of gin would contain about 0.11 standard drinks worth of alcohol. Keep in mind that alcohol content can vary between different brands and types of gin, so this is a rough estimate. It's always a good idea to refer to the label on the bottle for more accurate information.

Thanks, Chat GPT, i'm going to go ahead and drink 10 of these gin sodas and i'll only have had one standard drink!!!!

😂😂😂🤡🤡🤡

Happy friday ya'll

So...
I'm pretty sure I needed to take a break...

It looks like there aren't many great matches for your search.

Just give up and ask an AI bot to imagine some nonexisting bullshit instead.

Don't insist that there must be documents matching your search. Just accept that the age of quality is over and people must accept low-quality content as the new "information" standard.

Welcome to the age of alternative facts

Trying a custom eMbEdDeR today using open AI and pinecone on some douments of financial data. Let's see if its smart enough to be able to answer even the most basic questions, or if it just fumbles completely.

My bias is to be highly skeptical.

Will report back with results.

Imagination time.

With all our tech achievements, ai, ml, chatgpt, etc... Do you thing a completely automated future is possible? Automated agriculture, industry, healthcare,... Do you think we will still have finances/currency of any kind? If so - why? And how would we earn them if labour is no longer required [apart from ai/ml engineers]?

Do you ever imagine humanity ever reaching this fully robots-based future?

It's time for war!
I have two examples doing the same thing. They're bot generated by chatGPT.

React:
import React, { useState } from 'react';

function IncrementalButton() {
// Create state to hold the count value
const [count, setCount] = useState(0);

// Function to handle incrementing the count
const increment = () => {
setCount(count + 1);
};

return (
<div>
<h2>Count: {count}</h2>
<button onClick={increment}>Increment</button>
</div>
);
}

export default IncrementalButton;

VUE:

<template>
<div>
<h2>Count: {{ count }}</h2>
<button @click="increment">Increment</button>
</div>
</template>

<script>
export default {
data() {
return {
count: 0,
};
},
methods: {
increment() {
this.count++;
},
},
};
</script>

<style>
/* Add your CSS styles here if needed */
</style>

Whats better in your opinion?

How are these AI safety dimwits not aware that there are literally people opening PRs on non-toy, legitimately deployable ransomware C&C toolkits on GitHub. full blown CVE exploit kits etc, with readme’s and all.

yet god forbid ChatGPT doesn’t remind you it’s an AI language model so, it isn’t able to make predictions about whether the sun will rise tomorrow.

People started to use ChatGPT to discover a new vulnerabilities (0day), I saw someone use it to help them break a smart contract, I mean if you already found a 0day you might ask it to write the exploit rather write it yourself 😬

Did GPT have an unfair advantage compared to every general Chat-AI that comes after?

I mean, chatGPT was trained with data mostly generated by real humans. Now, since chatGPT and GPT4 is released, it has become more difficult to tell the content from bot and a Human apart. This means future chat AI will probably also by trained with data from other AI. And since AI doesn't understand context, and therefore make more factual error, future chat AI will learn more and more wrong things (generated by previous AI versions).

Where I can buy real American number?

I'm from Saudi Arabia, for some reason chatgpt is not available in my country and I cannot activate it with my Saudi number even with registering with an VPN

Stupid chatGPT UI is broken lmao. Can't scroll or do anything. What to expect from people for who making a simple chat app is too much. Can't believe we're paying for this shit.

ChatGPT can solve a Rubik's cube in less than a second, but it still can't find its way out of a recursion loop!

On dev.to and similar sites I'm starting to see tons of Cheatsheets and courses on how to use fucking ChatGPT.

How few neurons should you have to need a course to learn how to use something which takes any statement in natural language? If you know how to read write you should be fine.

Devrant bot based on chatgpt

now you can blame ChatGPT, if your code breaks.

Ok, @jestdotty , today, i give up on china.

I've been messaging with a rep who is taking the time to keep editing a contract... Im pretty sure she was genuinely trying...

As typing this we finally got to a 'correct enough' contract... so I could click the damn pay button.

Over the past 7 hrs.. at 3 back and forth exchabges and modifications at each issue:
1. Used previous PI from the dude i gave up on... so had a qty at 12 when only 11 exist a colour wrong for a crate of items, and listed the dude i refused to sign a contract under listed as the rep.

2. Now the item subtotals were off... just a few pennies or so... assumed she left the usd prices but calculated with ¥... didn't want alibaba to reject in a day so i checked if it was noted anywhere... Oh boy was it... VERY clearly, all caps, bold in the body of the total row... that the total was, exactly, 11680 (spelled out ofc) RMB aka ¥ chinese yen. I told her this, she sends me a cropped shot of the $ numeric total field... so i sent her the giant all caps bolded line, the one thatd typically be considered final say in most international courts... no clue where that value came from, it had zero relation to any actual values... and i was as curious as when chatGPT creates totally new, unique, lyrics for satirical german songs... i really tried.

3. Wrong incoterms (trade terms... abbreviated to a few letters... had it that I'd be physically going to the tbd port to accept/clear customs... no)

4. Technically it was accurate (well a few strange subtotals since she used ¥ half the time... told her it was fine as long as it had the company name on the label (gave 3 full examples to use whichever)
I get the contract ...shipping...
"To: Sara"
Then the right address (seriously wtf)

5. I point this out and carefully explain in mostly just examples and "the us government doesn't like anything being sent to just a first name, there's no legal way to sign for acceptance"

6. She gets stressed enough to tell me she doesn't have time to keep editing (since this horrid pile of poor formatting was just thrown at her a day ago... i dont point out the ridiculous irony)

7. Imo, the highlight of my night/morning... in her stress she promises me it'll ship right... sooo many issues there...
Even if it was delivered/allowed a signature for "sara" for 7ish large boxes just off a sea freight from china to a residence in the middle of a corn field (which tbh would be hysterical)...the IRS would have a valid reason to audit me... theyve done it w/o valid reasons several times, since I was 18 doing international trade and a contractual employee of a large gambling company, quarterly reporting, and ofc declaring more than my taxes in donating melted glass and crane game prizes...yea, they hate me and always do all that work to find the same thing... i underdeclare charity by 10%.

The entire concept of getting USA mail, even when pristine and you know logistics agents in every major company and port or distribution center, to properly deliver anything... ROFLOL ... and im already on some 'open and check everything' list with customs for a hysterical misconception they made years ago... cant/shouldn't get into detail publicly... but it was caused because 2 packages from different cities in China were both going to my address/through customs at the same time... package 1, 75 of those cheap af ball-pit hollow plastic balls for a 2yr old's bday(very delayed) package 2. 75 rechargeable batteries (the kind in power banks) 9600mah.

8. Told her to change "sara" to company name... glad it's registered to this address still.

It took me under 5min to type this... had to get the WTF out.

Dear AliBaba, please give an option to allow buyers to create the supply side contract for review, not just req modification... please?

💧

There was a 1 second delay when loading images on home page with a bunch of hot model babes in a masonry grid

Why? 💧💧

Maybe pagination is fucked. Lets reduce it from 100 per page to 10.

Still same shit.

What do you think why this could be?

Comment below 👇 right now

(The images im loading are just dummy images from unsplash)

I tried using rxjs. Observables. Flatmaps. Custom array push. Array loop. Change detection to update UI. Chatgpt.

Nothing

Every time i switch tab and come back then theres another second of delay with blank page before content appears

Wtf???

Turns out -- unsplash api was returning me 6K to 8K Fucking images. HEAVY. HEAVY FUCKING IMAGES. and i was apparently displaying 6000x8000 px images, 20 times per page. Thats a lot of fucking pixels! I reduced it with ?w=500 in unsplash api at the end and magically there is no delay now and everything works in an instant.

Fuck off

[Rust] What are alternatives to argument drilling for something like a string interner which is technically a memory leak so it really shouldn't be global but at the same time all but a couple top level functions depend on its existence? I'm aware of context objects and that's all ChatGPT could give me as well, but I'm wondering if there's more to this problem than that.