#configuration ChatGPT

[jumoyz]

Comment Utiliser ChatGPT? Guide pour 2024

En 2024, ChatGPT, le chatbot révolutionnaire développé par OpenAI, continue de transformer la manière dont nous interagissons avec l’intelligence artificielle. Que ce soit pour rédiger des dissertations, écrire des e-mails, générer du code, ou encore trouver des titres d’articles, ChatGPT s’est imposé comme un outil indispensable pour les professionnels et les étudiants. Ce guide vous montrera…

[pillowfort-social]

ChatGPT Bot Block

Hey Pillowfolks!

We know many of you are still waiting on our official stance regarding AI-Generated Images (also referred to as “AI Art”) being posted to Pillowfort.  We are still deliberating internally on the best approach for our community as well as how to properly moderate AI-Generated Images based on the stance we ultimately decide on. We’re also in contact with our Legal Team for guidance regarding additions to the Terms of Service we will need to include regarding AI-Generated Images. This is a highly divisive issue that continues to evolve at a rapid pace, so while we know many of you are anxious to receive a decision, we want to make sure we carefully consider the options before deciding. Thank you for your patience as we work on this. 

As of today, 9/5/2023, we have blocked the ChatGPT bot from scraping Pillowfort. This means any writings you post to Pillowfort cannot be retrieved for use in ChatGPT’s Dataset. 

Our team is still looking for ways to provide the same protection for images uploaded to the site, but keeping scrapers from accessing images seems to be less straightforward than for text content. The biggest AI generators such as StableDiffusion use datasets such as LAION, and as far as our team has been able to discern, it is not known what means those datasets use to scrape images or how to prevent them from doing so. Some sources say that websites can add metadata tags to images to prevent the img2dataset bot (which is apparently used by many generative image tools) from scraping images, but it is unclear which AI image generators use this bot vs. a different bot or technology. The bot can also be configured to simply disregard these directives, so it is unknown which scrapers would obey the restriction if it was added.

For artists looking to protect their art from AI image scrapers you may want to look into Glaze, a tool designed by the University of Chicago, to protect human artworks from being processed by generative AI. 

We are continuing to monitor this topic and encourage our users to let us know if you have any information that can help our team decide the best approach to managing AI-Generated Images and Generative AI going forward. Again, we appreciate your patience, and we are working to have a decision on the issue of moderating AI-Generated Images soon.

Best, Pillowfort.social Staff

[janmisali]

the "what number comes next?" problem

here's a sequence of numbers:

3198, 11, 734, 11, 1115, 11, 1440...

can you guess what number comes next?

that's right, it's another eleven! but what comes after that?

well, in this case it happens to be 1936. can you figure out the pattern? how far into the sequence could you predict?

this is a pretty hard problem, and in the most general form (I give you any sequence of numbers and ask you what number comes next) there's literally no way to solve it perfectly. the possibilities are endless!

but just because something is impossible to do perfectly doesn't mean you can't try. in the sequence at the start of the post, even though there was no guarantee that there's another eleven next, there's a sense in which that's the most "sensible" continuation, the one that follows the pattern most closely.

here's another sequence:

24794, 24794, 24794, 24794, 24794, 24794...

wouldn't it be reasonable to assume that the next number is another 24794? you might even be able to use some fancy statistical model to specify exactly how confident you are that the next number is another 24794.

but of course, without knowing the underlying meta-pattern that generates these patterns, there's no real reason to assume that any one pattern will be more likely than another.

so what if instead of just showing you one sequence of numbers, I showed you a lot of sequences of numbers? is there is a method of discovering this meta-pattern, assuming one exists?

the answer is, sorta! it depends on how many examples you're given and on how chaotic the underlying meta-pattern is.

so, you know how when you have a bunch of data plotted on a graph you can find the "best-fit line"? the closer the data is to being a line in the first place, the better this approximation will work.

defining a line only uses two parameters (mx+b), but the thing is that if you make a more complicated function with more parameters to play around with (as long as you're clever about it), you can define a relatively-simple mathematical equation that gets arbitrarily close to any data, no matter how messy that data is! you can just keep adding more parameters until you have enough to model the complexity of the given data.

now, the more parameters you have the more expensive it is computationally to find what configuration of all the parameters gets your fancy approximation as close to the data as possible (there are some linear algebra and calculus tricks that help!), but depending on the application, you can use this sort of method to get good-enough approximations to whatever data it is that you want to model!

however, this doesn't mean you've actually found the underlying meta-pattern. it's just a fancier version of looking at data you've already seen and drawing lines between points to interpolate between things. any approximation made this way will be in a sense "smoother" than what it's modeling, always predicting (for this application) that the next number in the sequence is the "average" in some sense of all the things that could come next, gravitating towards the most default and "boring" patterns.

anyway that's how chatgpt works

[carsthatnevermadeitetc]

Volkswagen ID.3 Harlequin Edition, April 1, 2024. The Harlequin paint made famous in the 1990s with the Volkswagen Polo is to return as an option for the electric ID.3 with an electroluminescent twist. The ID.3 Harlequin’s panels will light up and pulse in response to beats from the car’s sound system and integrate with the new ChatGPT functionality being rolled out across the Volkswagen range. The limited-edition ID.3 Harlequin is available from April 1; designs can be viewed on Volkswagen’s online configurator using the code ‘JEST 1’.

[smoqueen]

very connor o malley esque dream about scientists and doctors using ChatGPT technology to teach their bodies how to regrow skin in any configuration, allowing them to change their faces completely at will or appear young continually. i was at the dream where their bodies were starting to “catch up” to them, and my dream was Out of budget, so they were being represented by marker-hand-puppets in impossible and grotesque combinations

[mentelydrunk]

I'm so angry right now

I just found a 'creator' on YouTube who 'makes' transformers inspired music.

They have rock/metal songs for different characters and I was so hyped cause the songs sounded good and the lyrics fit each character/characters journey really well. I noticed pretty quickly that the art in the videos was off but didn't pay it much notice cause I was enjoying the music. I blasted through about 10 songs before I decided to finally check the description to see if my suspicion that the art was AI art was true....

.....and I found out that not only was the art AI generated, the entire songs were. The music, the vocals, the configuration or the lyrics and part of the lyrics themselves....all AI! The creator said they write the lyrics and run them through chatgpt to adjust and configure them to 'fit their vision'

I was really excited to find this music cause I love transformers and metal/rock is my favorite genre but now I'm just upset.

Their channel is called 'the automatic singer' by the way.

[vaingod]

as someone that regularly programs with chatgpt and i use ai configuration for databases and learning cloud computing i cant help but laugh everytime I see someone getting red in the face about people using chatgpt or any ai, its absolute clown nonsense to expect people not to use the software that magically gives them answers right or wrong in an age where people don't know how to store their files or what a file even is, but also seeing how abysmal some people are at using search engines in general and applying this incorrect takeaway to ai- especially when the best way to use chatgpt is to have it run many instances of a given code or data until it meets a framework you can work with ... its just so laughable how personally offended some people are by it they act like it is the end of the world to have a slightly more refined model of the same technology we've been using for the past 2 - 3 decades

[notational]

new Tiles Processing API experiment Back to playing with grids and tiles. In this case two 3 X 3 grids are put side by side and tiles are swapped in different configurations. The thing that makes me happiest about this program is that with some help from #chatgpt  I was able to scrape all images from my #Are.na channels using the Are.na #api and I could save them in higher resolution.....though this video drops that resolution back down again. You win some you lose some. Made with #processing and some help from #chatgpt. I’d be very interested in suggestions, feedback, and comments. Please subscribe for more. If you’re interested in the code, I will eventually post it on Github and Open Processing.  #creativecoding #generativedesign #videoart #remix #programming #vlog #vlogger #inprogress

[geneticallymodifiedidiot]

uggghhhh my god

so in band since we had a concert a few days ago we aren’t doing much so a lot of kids were studying for the chemistry test that a lot of classes have over the next few days

and a couple of kids asked me for help with studying and material because apparently they think im smart

so of course I went to help them and they were specifically asking about electron configuration

and the thing is that the way you write electron configuration can vary based on a whole bunch of things

and at one point i was grabbing some notes to review from and one of them suggested asking ChatGPT

the unbridled rage I felt omg

I was just immediately like “no. Do not ask ChatGPT. Do not do that.”

Because seriously. This pissed me off for three reasons

1. ChatGPT is unreliable as a source of information at best. You have a test later today; do you really think this is the most sensible course of action?

2. The way that the specific topic (electron configuration) is approached is approached differently by different people. Even if you get an accurate answer, it could be a different system.

3. I’m right here, taking the same class you are taking, trying to help you, at your own request.

Anyways, I hate AI.

also i got a 96.7% on the test but i digress

[glowing-disciple]

Another thought about AI:

I’ve never used ChatGPT. The last time I looked at it, an account was required and new accounts were not available.

So I ended up looking into Google’s AI, Gemini. You still need a Google account to use it, but I have one of those.

I’ve also been tinkering with LlamaFile, which doesn’t need an account. (I’ve also configured it to hallucinate like crazy, so it’s just entertainment at this point)

So here’s the thing. Gemini cites its sources and even has a self-check button. Hallucinations get flagged if you use the self-check button, and you’re provided with possible search queries to find your answers for yourself. Gemini’s interface also features reminders that it’s not guaranteed to be accurate or reliable, and shouldn’t be used for medical advice.

This has me wondering if a lot of the problems people are seeing are unique to ChatGPT.

Does it have any sort of double check feature?

Or does it promote itself as super duper reliable?

[adi-barda]

Chapter 4 - Gemini API Developer Competition - Fighting game & Android Export

As planned, I spent the last days on adding fighting game capability to the engine and Android exporting feature. The fighting game has much more details in the puzzle for the AI agent to cope with. There are complex animations for the player and for the opponent, they need to constantly look at each other, you need to be able to demo their kick, punch, block animations, the player needs to be able to move in 3D space etc. Overall I'm very pleased with the results so far. The user can speak freely enough with the AI, get instant results and funny reactions. What's more, I've been able to add Android exporting of the game and automatically open it in Android studio. It was challenging because the Java code worked different on PC and on the mobile device specifically handling of Zip files and all kind of Gradle dependency hell. ChatGPT was on my side all the way, assisting me to resolve configuration issues and coding problems such as selecting the best Zip 3rd party library.

This video clip, demonstrates the current status of the project. It shows a complete story from the user perspective - you have a conversation with the AI, a game is created and finally you export it to Android studio for deployment in Google play store or any other market place.

What's next

Better and shorter presentation

Prepare the installation of all the components as well as SceneMax3D dev studio

Get feedback from the community

Prepare documentation for the architectural strategies, entities diagram etc.

So far I'm getting very good vibes from the game dev. community, and friends on various WhatsApp groups.

[e-carlease]

But what is the GTX? To supplement all of the above, a new ID.Buzz GTX is also on route to the UK. This will have two battery / wheelbase configurations much like the models noted above. The GTX is the most powerful series-production Bulli which adds the dual motor configuration and the 4Motion all-wheel drive system. This translates to better trailer weight abilities with a sizeable 1800kg on offer. The GTX is very much performance and towing specific but there are some distinctive technology and aesthetic improvements. A GTX external styling, black air intake grille, black body elements, 20” alloys, IQ Light LED matrix lights plus a new Cherry Red are available. Inside the GTX Buzz, you can enjoy the black headliner, black ArtVelours mircrofleece and red cross-stitching. A bigger 12.9” touchscreen include the IDA voice assistant and ChatGPT integration is also included. As VW say - “Experience tomorrow’s technology today”

GTX - this AWD MPV will have an 79 kWh usable battery which will offer 0 – 62 times of 6.5 seconds, 99mph top speeds and 250kW (or 335hp). Expect a combined winter range of 175 miles with warmer weather allowing for 230 miles – a 205 mile combined. On charging, the 11 kW AC max will allow 8 hour and 30 min 0 – 100% charging times with the 185 kW DC maximum allowing 27 minute 10 – 80% times. A cargo volume of 1121L is available with this car. It has a vehicle fuel equivalent of 105 mpg. You can tow with this EV - 750kg (Unbraked0 and 1800kg (Braked). It also has Bidirectional charging facilities - V2H (vehicle to home) and V2G (vehicle to grid) are set to feature. The Heat Pump is available.

[adafruit]

FULL VIDEO - Writing an Arduino driver with OpenAI ChatGPT and PDF parsing 🤖🔧📄

One of the big tasks that Ladyada still has to spend a lot of time on is writing Arduino libraries for all our devices and sensors, particularly all the I2C & SPI chips out there! These ICs use register maps and sub-byte addressing to set dozens of configurable knobs and switches, and a good driver lets folks set and get all of the noodly bits.

However, there is yet to be a standard format to get that configuration map. Instead, you have to pore over datasheets with long lists of binary tables and bit insets to figure out how to convert that into C or Python code.

It is tough. Only a few folks can write an excellent comprehensive library…. but Ladyada can & has! In fact, there are hundreds of Arduino libraries on Adafruit's GitHub https://github.com/orgs/adafruit/repositories, all in the Ladyada 'style,' using Adafruit_BusIO for I2C / SPI register addressing https://github.com/adafruit/Adafruit_BusIO/ and since ChatGPT 4 was trained on all of it, we can ask it to become a mini-Ladyada to write new drivers.

With a PDF parsing plugin, we can even upload the chip's datasheet to extract register names, values, create enum tables, and text for doxygen comments. Here's the chatGPT log for the video https://chat.openai.com/c/f740eb57-17a6-41e3-ae0a-12da959a1f4c - and here's a previous one that is more 'complete' https://chat.openai.com/share/f44dc335-7555-4758-b2f9-487f9409d556. The amount of time it takes for ChatGPT to write a driver is about the same as it would take Ladyada, and you definitely need to be eagle-eyed to redirect the AI if it starts making mistakes… but it can be done even when Ladyada is tired after a full day of baby-care, or at the same time as pumping https://www.youtube.com/watch?v=EpbH-sXRNps - plus there's a lot less continuous typing/mousing so her wrists don't ache the next day!

Do you use ChatGPT for electrical engineering or coding work? Any suggestions on how to make this even better? This is only our 3rd day using this tool, so we're getting started with how to integrate it into our workflow.

[mmopor]

Watch How We Turn TikTok Into Our Personal ATM Using ChatGPT & Make $330/Day From Short Ai Videos!

Brand New  “ChatGPT + TikTok” Software Churns Out Pro-Level Short Videos 24/7 Using Ai & Plasters Them To TikTok!

ZERO Video Creation, Editing, VoiceOvers Required!

Create short videos with 1 click in any niche for any keyword using True Ai

ZERO video creation or editing required: ChatGPT does all the work

Instantly publish to TikTok or schedule for any hour

Use our “algorithm hacking” to get in front of millions of people and get virtually unlimited free traffic

Built-in affiliate campaigns: just copy/paste your link and let ProfitTok SELL for you!

AI based video creator & editor makes it all take only SECONDS

You can simply clone, cut & publish other people’s successful campaigns 

COMMERCIAL LICENSE built in: sell videos to local businesses or provide video making services

Leverage the tools TikTok gives you in their fight against Instagram & Youtube

Virtually unlimited free traffic from the world’s hottest platform

Unlimited free end-to-end SSL encryption protects you, your privacy and personal data

100% Newbie Friendly

No Monthly Fees Ever – Pay once, use forever

Double Your Money Back Guarantee

24/7 support from marketing gurus

HURRY! Price Reverts To Monthly Fee In…

Get Started With This Revolutionary New Software Today!

Click Here To Grab ProfitTok

WARNING: FREE COMMERCIAL LICENSE FOR EARLY ADOPTERS ONLY…

With ProfitTok, You Are Minutes Away From Pro-Level Videos And Millions Of Visitors!

Become The World’s Next Big TikTok Star In Just  3 Simple Steps:

Login To Our ChatGPT-Powered Short Video Creation Software!

(There is NOTHING for you to download, install or configure)

Write Any Keyword Or Paste Your Affiliate Link And Watch ProfitTok Auto-Create, Edit And Publish Your Short Video Using “True Ai” Tech!

(The AI video maker does it all for you so there’s ZERO video editing required on your end)

With 1-Click Send Your Short Video To TikTok And Get Millions Of Views Using Our Next-Gen “Algorithm Hacking”!

(Leverage the tools TikTok gives you to make them compete against FB, IG, YT, get #1 rankings and visibility and use our “algorithm hacker” to get in front of millions)

Step #3

Step #2

Step #1

ProfitTok Is SO GOOD – It Gives You UNLIMITED Opportunities…

You Can Become The Next TikTok Sensation!

You Can Even SELL These Videos For $1000+ Each Using The Built-In Commercial License!

Feeling STUCK!? 

Let Us Hand You The Biggest $33 BILLION Opportunity On Earth...

You’ve probably seen a LOT of video makers…

​

They look “easy” and “shiny” before you buy.

​

But they’re ill-suited for today’s world…it takes a lot of time to figure out how they work.

​

You need to constantly come up with new ideas for videos.

You need to “record” content yourself (or risk a copyright claim!)

​

And at the end… after all this work you’d still be missing the forest for the trees!

​

You see – all the traffic is now going to short videos.

​

That’s TikTok plain and simple…they’re pulling 5 trillion views a month!

​

Only problem? All existing video makers are not focused on creating short videos, so you’re stuck with paying hundreds to outsource, or trying to record & edit the videos yourself…which frankly is all IMPOSSIBLE for newbies!

FACT: Short Videos Are Pulling In 5 Trillion Views A Month!

BUT No Existing Video Software Leverages ChatGPT Ai Video Creation AND TikTok…Yet!

How would you like to be able to tap into ChatGPT and “force” it to create pro-level short videos for you, auto-edit and publish them to TikTok within the next 2 minutes?

​

Better yet, imagine if minutes from now you could have your very own Ai video maker that enables you to finally create the traffic pulling videos you’ve always wanted… 

​

but without any video creation required... 

​

What if this software also came with commercial license so you can SELL videos to other people…

And It's All...

100% Done For You

100% Newbie Friendly

200% money back guarantee if it doesn’t work for you (it will!)

[sirfrogsworth]

A few people have been reblogging this old post of mine from 2014.

I was a little worried because technical information is often the very opposite of evergreen content. And I feared this would be full of a bunch of outdated advice. But after looking over it, thankfully, all the main points are still accurate. So that is a relief.

The only changes I might make if I were writing it during present day...

You can just leave "Resample" on "Automatic" and it will choose the best method for you. Changing it to bicubic sharper is just an extra thing to remember and not really necessary. The people who need to use something besides automatic already know all of this and don't need this tutorial.

And I have no idea if the native resolution of modern printers has changed in the past 10 years so you will need to check the printer model you are working with to figure out what ppi it is using. It's likely they are still mostly 300 ppi (360 ppi for Epson), but I have not kept up with that information and cannot say for sure.

Remember the advertised dpi is not the metric you need. You have to do some digging to find the native ppi of the printer. You might be able to ChatGPT it.

And if you are having your image professionally printed you should always ask their preferred document specifications. Often they'll have a cheat sheet you can follow or a FAQ on their website.

I did notice a few people tried to add some information to this post and some of that was... not great.

Someone said that if you work in 300 ppi in Photoshop it could slow down your computer. But Photoshop does not care about the ppi, only the number of pixels in your document. Pixels are what is registered in your computer memory, not the print resolution.

So if you are experiencing laggy behavior, you would need to make sure you are decreasing the total pixel count. You might also try consolidating and reducing the number of layers. A lot of times 300 ppi documents have large pixel dimensions because people aren't paying attention when they are configuring their initial specs, and that could make it seem like 300 ppi docs are slower. But ppi is only relevant to printing.

Another person said 72 ppi is "web resolution." And again, with pictures displayed on a screen the ppi is a nonfactor. So I will reiterate... you can create a 1920x1080 72 ppi document and a 1920x1080 300 ppi document and they will appear the exact same size on your screen and on the web. But if someone tries to print your 300 ppi image from a website, they might get a surprise when it outputs very tiny.

As far as what pixel dimensions to do for web output, I would still vastly oversize your art when creating it and then create a resized version for the web when your image is finalized. It's probably best to research the platform and see what their max image dimensions are and resize your final image to match before publishing. Sometimes that can prevent extra compression when uploading.

In any case, everything else is still current information and hopefully helpful to artists who print things on occasion.

How big should you make your art?

I’ve noticed some digital artists out there who just kind of guess when choosing the dimensions of their artwork. Trying to understand ppi, dpi, print dimensions, and resolution can send you down a rabbit hole of complexity likely to break your brain.  

If you are creating an image only for the web, it is really up to you how big you want to make it. The only relevant dimensions are the pixels. Print size and pixels per inch are of no consequence. A 1920x1080 300ppi image will be the same as a 1920x1080 600ppi image. Your screen only cares about pixel dimensions.

If you plan to print your image, that is where things can get complicated.

Here is the simple version…

First you need to determine the maximum size that your work may be printed. Here are the most commonly used sizes for poster prints. 

Now you need to know the brand of printer that will be used. Typically it will be Epson, HP, or Canon.

For Epson printers…

Input the width and height. 

Input a resolution of 360 pixels per inch.

For Canon, HP, and most other printers…

Input the width and height. 

Input a resolution of 300 pixels per inch.

So let’s say you are in photoshop and you want your art to be printed at 11" x 17" on an Epson printer. This is how you should create your document. 

If you do not know how your work will be printed, I recommend erring on the side of “too big.” If you end up having to enlarge your work, it could result in some quality loss. 

Hypothetical situation…

You don’t know how your work will be printed. So you decide to make your art 24" x 36" at 360 pixels per inch. Your image size dialog box looks like this. 

You find out that the art will be printed at 11" x 17" on a Canon printer. You need to resize your document for optimal sharpness. You’ll want to change the largest dimension value. In this case, 36" is the largest, so you change that to 17". Since it will be a Canon printer, you will need to adjust the resolution to 300 pixels per inch. And because you are making the image smaller, you’ll want to use “bicubic sharper” to maintain the best image quality.  

Now you have to crop off that .333 from the width. That is done in canvas size. 

Hit okay and now you have an 11" x 17" document at 300 pixels per inch ready to print on a Canon printer. 

Getting more technical…

The goal is to size your work so that the printer does not resample your image. Meaning the printer driver doesn’t take your image and make it bigger or smaller. Your image editor is always going to do a better job of resampling than your printer, so if you can set up your document to the native resolution of your printer, it will output with the highest possible sharpness. 

A lot of people see the dpi of printers and think that is its resolution. The manufacturer will even say that is its resolution. It’s not. That is just how many dots it can cram on the page. Many dots make up just one pixel of your image. And the number of pixels your printer can cram on the page is actually its resolution. 

The printer’s resolution is measured in ppi or pixels per inch. You will notice in photoshop it doesn’t actually say dpi. It says pixels per inch. Many, many people use ppi and dpi interchangeably and it is frustrating and confusing for everyone. They aren’t the same thing, but most people say dpi for everything. Even the manufacturers will use dpi incorrectly. To make matters worse, usually printer manuals don’t even list the native pixels per inch. Probably because it is a low number and looks less impressive when marketing. 

Some oddball printers may not use the 300 or 360ppi native resolution. If your images still look soft and you want to be absolutely sure of your printer’s native pixels per inch, the best option is to email the manufacturer about your model. Be very specific about what you are asking. If they give you a crazy high number like 6000 x 12000 dpi, you tell them, “NO, YOU FARTNUGGET. I WANT THE NUMBER OF PIXELS PRINTED IN ONE INCH, NOT THE NUMBER OF DOTS!” If they respond back with something like 400 pixels per inch, then that is the resolution you’ll want to use for your document. 

Epsons typically are 360ppi or 720ppi. Other manufacturers are usually 300ppi or 600ppi. Wide format printers almost always have the lower ppi. The higher ppi’s are typically only used for fine detail vector printouts and you have to set your driver to use that fine detail mode. The nice thing about vector illustrations is that you can alter their dimensions and ppi without any quality loss. You can certainly make your digital paintings at those higher ppi’s, but I’m told there is little quality difference when printing and the files at that resolution can be a lot for your computer to handle. 

Hopefully this was helpful. If you have had problems with images looking soft, I think this might help you get them looking sharp. Take care and make some art.