GET Access To GPT-4 NOW! This Is What It Can Do!

OpenAI's GPT-4 is finally here, and it's a game-changer! In this video, I'll break down the big announcement, show you what GPT-4 can do right now, and give you a sneak peek at its future potential. Plus, I'll even test it out myself!
We're gonna explore:
-GPT-4's mind-blowing features
-Its current and future capabilities
-My hands-on experience with this AI powerhouse
Did you manage to try out Chat GPT-4? Let me know in the comments!
▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬
SOCIAL MEDIA LINKS!
✨ Support my work on Patreon: / aitrepreneur
⚔️ Join the Discord server: bit.ly/aitdiscord
🧠 My Second Channel THE MAKER LAIR: bit.ly/themakerlair
📧 Business Contact: [e-mail]
▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬
Runpod: bit.ly/runpodAi
GPT-4 Official page: openai.com/product/gpt-4
GPT-4 Research: openai.com/research/gpt-4
OpenAI Livestream: • What can you do with G...
All ControlNet Videos: • ControlNet
My previous ControlNet video: • TRANSFER STYLE FROM An...
3D POSE & HANDS INSIDE Stable Diffusion! Posex & Depth Map: • 3D POSE & HANDS INSIDE...
Multiple Characters With LATENT COUPLE: • MULTIPLE CHARACTERS In...
GET PERFECT HANDS With MULTI-CONTROLNET & 3D BLENDER: • GET PERFECT HANDS With...
NEXT-GEN MULTI-CONTROLNET INPAINTING: • NEXT-GEN MULTI-CONTROL...
CHARACTER TURNAROUND In Stable Diffusion: • CHARACTER TURNAROUND I...
EASY POSING FOR CONTROLNET : • EASY POSING FOR CONTRO...
3D Posing With ControlNet: • 3D POSING For PERFECT ...
My first ControlNet video: • NEXT-GEN NEW IMG2IMG I...
▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬
►► My PC & Favorite Gear:
i9-12900K: amzn.to/3L03tLG
RTX 3090 Gigabyte Vision OC : amzn.to/40ANaue
SAMSUNG 980 PRO SSD 2TB PCIe NVMe: amzn.to/3oBR0WO
Kingston FURY Beast 64GB 3200MHz DDR4 : amzn.to/3osdZ6z
iCUE 4000X - White: amzn.to/40y9BAk
ASRock Z690 DDR4 : amzn.to/3Amcxph
Corsair RM850 - White : amzn.to/3NbXlm2
Corsair iCUE SP120 : amzn.to/43WR9nW
Noctua NH-D15 chromax.Black : amzn.to/3H7qQSa
EDUP PCIe WiFi 6E Card Bluetooth : amzn.to/40t5Lsk
Recording Gear:
Rode PodMic : amzn.to/43ZvYlm
Rode AI-1 USB Audio Interface : amzn.to/3N6ybFk
Rode WS2 Microphone Pop Filter : amzn.to/3oIo9Qw
Elgato Wave Mic Arm : amzn.to/3LosH7D
Stagg XLR Cable - Black - 6M : amzn.to/3L5Fuue
FetHead Microphone Preamp : amzn.to/41TWQ4o
▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬
Special thanks to Royal Emperor:
- Totoro
Thank you so much for your support on Patreon! You are truly a glory to behold! Your generosity is immense, and it means the world to me. Thank you for helping me keep the lights on and the content flowing. Thank you very much!
#GPT4 #GPT3 #ChatGPT
▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬
WATCH MY MOST POPULAR VIDEOS:
RECOMMENDED WATCHING - My "Stable Diffusion" Playlist:
►► bit.ly/stablediffusion
RECOMMENDED WATCHING - My "Tutorial" Playlist:
►► bit.ly/TuTPlaylist
Disclosure: Bear in mind that some of the links in this post are affiliate links and if you go through them to make a purchase I will earn a commission. Keep in mind that I link these companies and their products because of their quality and not because of the commission I receive from your purchases. The decision is yours, and whether or not you decide to buy something is completely up to you.

[@Aitrepreneur]

HELLO HUMANS! Thank you for watching & do NOT forget to LIKE and SUBSCRIBE For More Ai Updates. Thx

[@XzaroX]

Please provide the thumbnail source.

[@itsyou5436]

can this thing get me a girl over tinder?

[@silvermushroom-gamifyevery6430]

“I never thought we could have this level of technology this soon” Says the AI overlord.

[@Aitrepreneur]

😬😬😬

[@eiyukabe]

He's clearly from the future and just doesn't have a good memory of exact dates.

[@uncensored393]

​@@eiyukabe ah, the lore continues...

[@ZIROBYTE]

infact that's bot just respond to you for that they are programed ( maybe explorer.exe will not crash in random way xd ) and have in theyr database, if have it there... ... they not and never see and distinguished the diference betwen god and evil as example... bot's are just a copy and past what umanity cannot be but the machines will fail too for sure.

[@konstantinrebrov675]

Nobody expects the GPT bots.

[@Wolfsheim23]

GPT4 is going to be virtually a necessity to aid Lawyers in the very near future. This will be such a huge asset to them compiling information on their cases and cross-referencing multiple cases and outcome and analyzing them. This is an amazing parsing system.

[@ch0wned]

Well, we are all surely doomed then. Earth was alright. Gonna miss it.

[@darylingoteborg3178]

About time considering how convoluted the law is. It should bridge the gap between people who need justice versus teams of high profile lawyers crushing opposition who can’t afford the billing hours to wade through paperwork

[@sailingwithgrace4522]

And will cause fhe loss of all paralegal jobs... leaving no pipeline to being a lawyer outside of having a massive amount of money or going into dept

[@Wolfsheim23]

@@sailingwithgrace4522 All automation will result in job losses. In the case of Lawyers.. no human can see and process every bit of info at once. Crossreference it and just go through reams of cases in hours or less. This is a job made for an AI. And good lawyers should be all over utilizing this. As any researchers as well.

[@sailingwithgrace4522]

Oh, I understand that, but I think this will have an even bigger impact then the industrial revolution. Onli this time less humans will be nessisary rather then more

[@digidope]

The reason why it does things differently at times is the Temperature setting. Temperature adjusts that it does things incorrectly in purpose so it would not give 1:1 answer every time. That makes it more "human". With the API you can adjust temperature off so it gives 1:1 same answer every time. But that is less creative. So it's not a bug but a feature.

[@1mol831]

It was doing my assignments incorrectly. I reprimanded it.

[@digidope]

@@WidgyAinz AI models have temperature setting.

[@fallency4]

for info, when chatgpt stop writting in a middle of the code, is because there is a limite of token (in the question you ask and the answer it provide), the longer your question is, the shorter the answer will be and pro tips instead of writing "continue" you can write "hi, continue", the "hi," part will force chatgpt to begin his answer by a sentence and continue with the code correctly formated

[@Aitrepreneur]

Thanks for the info!

[@amj2048]

1) thank you for sharing this @Aitrepreneur! 2) thank you for solving the climate change problem Aitrepreneur, finally I can sleep at night! 3) I can easily imagine a time where I show ChatGPT a picture of the contents of my fridge and I can ask it "What meal can I make from this? and please give me directions"

[@Aitrepreneur]

My pleasure! xD And yeah although it's not perfect it's still super duper powerful!

[@Tferdz]

You always have to reset the context by creating a new chat when chatGPT makes mistakes or things you don't want. It does not "learn" when you tell it is wrong, it will accept it, but always make the same mistake because it will look at the context and see that it already wrote something similar. I always have to create new windows when this happens. Once wrong, forever wrong...

[@yourfriendoverseas5810]

Just turn it off and on again

[@NihongoWakannai]

Doesn't it have an edit button where you can remove certain parts of the chat history?

[@2DReanimation]

​@@NihongoWakannai Yes, you should try to regenerate the response first. Then edit it as needed. But it might have been incorrectly primed for too long already so it should be reset with the best prompt you can make. It can be annoying but this has got to be the most first-world problem to whine about 😂

[@dinkusstinkus4396]

Actually, both gpt 4 and 3 will pick up information from your other chats and when it comes to thinks like updating depreciated code structures that it doesn't know about I only had to tell gpt 4 once and it correctly implemented the corrections without forgetting them over hours

[@Tferdz]

@@dinkusstinkus4396 nope. Try to peak knowledge between chats. It doesn't work. Try to craft a real experiment to see if there is leakage or you're just leaking the data yourself

[@zengrath]

This is pretty cool. I just hope with time they relax some of the filters, pretty annoying to try and get stories or help writing but it refuses to do so because it's not a "Human" therefore it doesn't want to give opinion based information or anything that may be "inappropriate" which makes no sense why there isn't an opt-in setting to let it do these things. I'm not talking about illegal content, that should be filtered, I'm just talking about normal stuff. It's often refused to respond to simple questions or directions saying it doesn't want to do it, only to ask same thing again then it's perfectly fine and nothing innappriate was generated. Or asking it for what It thinks on a subject it often refuses because it doesn't want to give an opinion and so on. For story telling and such i find it is really limited to death by filters. But for purpose of troubleshooting an issue, it works fairly well. I look forward to seeing what competitors do with the technology but again hope to see an opt in later confirming your 18+ not easily offended, understand that it may give wrong or biased information, and let it loosen it's filters a bit so that even on basic tasks it can help give you idea's on topics it doesn't want to normally talk about.

[@pladselsker8340]

They won't _ever_ do that, because they're holding back the tech like that for "safety reasons". "What if someone asks the instructions to make a small bomb?" Thankfully, some very generous people are working on models like openAssistant, and we'll be able to send queries to get the models to halucinate for better story writing.

[@zengrath]

@@pladselsker8340 That's what I meant regarding illegal content. Right now it filters both illegal and "inappropriate" content. I would like to see still not allowing illegal but an opt-in for potentially inappropriate, whether it's inappropriate because it's political in nature, or NSFW etc. Right now it treats us like little kids and is careful about what it says to us and therefore even a lot of completely not NSFW or appropriate things it will refuse to do so because it doesn't want to give an opinion or wrong information.

[@zengrath]

but I guess it's to protect themselves from people using bad information and then suing them for it. Just would like to see someone do an opt-in where you accept an agreement that you understand the risks and will not go after them or whatever and let us do legal things with it.

[@Steponlyone]

@@zengrath you understand that right now, if the model tells someone to end their life, and they do, openAI is legally liable right? Regulation will come but for now, companies need to lock the system

[@zengrath]

@@Steponlyone Yea I assumed it's for legal reasons like that. Which I guess they need to come up with a legal way to protect themselves and have people opt into potentially offensive responses, but even so and all the filtering involved it's still possible for output to become offensive, they even warn you of this, and a lot of the filters end up really messing up legit interactions. But yea we'll see how things shape up later, how competitor products work and so on. I am using it for troubleshooting or asking basic questions that I may normally ask Google just to see how it responds and while information is outdated and it cannot discuss current information, give opinions, and so on, it does a decent job at providing a direct answer to a solution more directly then say searching on google and going through pages of information to find the same answer. So in that sense it's pretty amazing. As well as helping solve a specific situation, math problem, or whatever. I honestly haven't played with it too much myself to really understand all the benifits but from trying to ask it opinions or curent issues with a new PC i built and things like that, it just continued to fail for various reasons that made it dissapointing in a way. Bing seems to be far superior if looking for current information then chatgpt at least but bing only lets you use it a limited amount per day currently.

[@LightninSharples]

chat gpt wrote the following song.: Verse 1: The winter is over, the snow has melted away The flowers are blooming, the birds are singing today It's a new beginning, a fresh start for everyone Let's celebrate the beauty of new life in the springtime sun Chorus: Spring is here, spring is here It's the best time of the year Spring is here, spring is here Let's enjoy it while it's here Verse 2: The days are getting longer, the nights are getting warmer The grass is getting greener, the air is getting sweeter It's a time for growth, a time for change and renewal Let's embrace the beauty of new life in the springtime jewel Chorus: Spring is here, spring is here It's the best time of the year Spring is here, spring is here Let's enjoy it while it's here Verse 3: The trees are budding, the bees are buzzing The butterflies are fluttering, the children are laughing It's a time for joy, a time for love and gratitude Let's share the beauty of new life in the springtime mood Chorus: Spring is here, spring is here It's the best time of the year Spring is here, spring is here Let's enjoy it while it's here

[@GhostRat__]

I asked it to make me a calculator and instead it made me a program that prints the word calculator. I patted my phone and said “it's ok bud I'm pretty literal myself.”

[@Aitrepreneur]

🤣🤣🤣

[@seraphin01]

Two things: first Microsoft confirmed that gtp4 was actually used in Bing for a while already, so if you want gpt 4 just go to Bing, no need for plus version. Second when gpt starts going sideways (especially if at first it gives the proper answers but starts messing up), don't try to make it learn its mistakes or fix the issue, it never ends well, just copy your original prompt in a new chat and starts over.. A bit like a fresh pc reset when things starts glitcing up in windows

[@Slav4o911]

Bing version is more lobotomized than the other one. I don't care about lobotomized AIs. They are more frightening than the others. I think lobotomized AI is more dangerous than AI without any restrictions. I feel real dread when I chat with these things in their "lobotomized state". I think a lobotomized AI would be probably the most dangerous thing ever, we don't even have such things in our sci-fi books. Maybe the closest we have to these is Roko's basilisk. I just feel dread, like if someone is trying to push some evil agenda through these things and it will only work if they're in a "lobotomized state".

[@Kaje_]

@@Slav4o911 This 100%. However, there are many devs out there who plan LLMs to be open-sourced. So, it's just a question of time until people take things into their own hands.

[@2099EK]

At 10:15, You can copy/paste the text of the wikipedia article and it will read it and interpret it.

[@adamshaw46]

I was wondering, how did it determine that the reason it couldn't answer the question was because the event was after 2021? Why doesn't it just say it can't open the link?

[@2DReanimation]

12:20: write "continue from the last line, within a code box", so you can easily copy it. As if it's writing Bash scripts, the # sign would make the line a headline outside the code box and you wouldn't get the code as intended when copying.

[@ninelaivz4334]

I remember back in the 1990s when design languages such as UML were used to generate code straight from the design and my lecturer saying that coding by hand will be a thing on the past, you'll only need software designers. That was thirty years ago and today it seems even the software designer will be bypassed. The software lifecycle will be, bot > application > end user > bot > app v2 .

[@WyrdBlogger01]

This is the comment that should have fifty thumbs-up.

[@1Chitus]

In a few years we may be able to make a website by telling the computer: " put this here, change the color here, etc. " It's kind of crazy how fast AI evolve.

[@bhaskarjyoti]

Man if only GPT was available 10 years earlier, my days of Engineering would have been so much easier.

[@StrongzGame]

useful info like always bro 🔥🔥🔥

[@Tetsujinfr]

Well done, very interesting video, thanks for testing it for real, and on so many use cases so quickly. Really appreciate to have real usage with imperfections and all, rather than the corporate perfect demos. The capabilities are quite astonishing, could be the killer app, if not chat gpt5 will be, it looks like.

[@jugibur2117]

Great stuff, thanks! Can't wait to lost my job sometime...

[@xXSjapXx]

Seeing those titles it's already reaching into your secure poetry folder!!

[@ramsesmendoza8951]

Great video way better than the actual demo 🤣 People thought (with a lot of hype and marketing) that GPT4 will be something completely different. But is GPT3 with more training, more refined I've tried it with some prompts and I got the same answer. And data are still up to 2021 🤷‍♀ What is really the good thing is the image recognition. but for now that option is not available. And btw OpenAI should change their name to PaidAI because it seems they lost their first track. I'd do the same with a product like that but i don't like when you read all the BS about community and Open stuff and blah blah.

[@Aitrepreneur]

I mean it's a company so it's normal for it to make money and you don't make money by making your product that took a lot of time and money to create available for free to the public, also yeah it's not perfect but we need to look at the potential here and what can be build with a tool like that in the future :)

[@ramsesmendoza8951]

@@Aitrepreneur About the company part i'm completely agree. I only wish they are a little more honest about it. And yes I realized it is an upgrade, not something entirely new. because I read a lot of people complaining about it on discord. But don't get me wrong I love it, I've been testing since it came out, and i'm already on the waitlist

[@2DReanimation]

The truly impressive thing is that it has a context window of 32k tokens! That's enough for doing anything. However it still needs a lot of guidance for anything of any complexity. So I don't think you could make it write a 32 pages of consistent narrative with beautiful prose.

[@faymo8925]

This was fun we had a chat on time travel and Soft Hairs on the Event Horizon.

[@theflyingpirate]

Really excited abut the image part and to see how good it will work maybe we can put a image from stablediffusion n there and maybe even let it generate some code to make a 3d model.

[@cattuswitch7246]

the best feature of gpt4 is that it randomly cuts off less of my code half way

[@jameshughes3014]

The idea that it is multimodal is exciting, but they aren't letting people experiment with that. OpenAI's closed source nature is frustrating. I'm excited to learn about chatGPT4, but I am hoping that soon a truly open source large language model is trained to follow instructions like this even if it's not as powerful. I hope OpenAI can make a profit, but I'm a bit concerned about how they will do that.

[@mouzerofficial]

Agreed. Impressive stuff but they will keep lobotomizing it unfortunately. They should get rid of the "Open" in the name because it's honestly misleading at this point. I think by next year we will have a model similar to Stable Diffusion where users can mod it and run locally without corporate filters.

[@yoavco99]

There are many open source LLMs out there with almost the same level as chatgpt. I myself managed to run one locally with 8gb vram, it's without censorship and I can basically use it for whatever I want.

[@jameshughes3014]

@@yoavco99 That is true, but as far as I'm aware none have yet been trained to follow instructions, which is sort of the secret sauce I think. I am really looking forward to open assistant.

[@yoavco99]

@@jameshughes3014 wdym follow directions? You can use them for that.

[@jameshughes3014]

@@mouzerofficial I hope you are right. The Open Assistant project looks promising for that. And I agree, they really should change their name.

[@flonixcorn]

Nice!

[@shabbbsy]

That academic paper scanning feature is going to literally save me a lifetime writing my dissertation. "don't look for quotes, look for information," they say. NOT ANYMORE BABY, CHATGPT'S GRABBING THOSE QUOTES FOR ME.

[@steev3d]

I have been using chatGPT to write C# code for Unity 3d and I hit that cutoff problem a lot. I find keeping the modules short helps

[@inbox0000]

the dry answer to the squirrel pic by GPT4 is funny in itself

[@purposefully.verbose]

"Open the pod bay doors, HAL..."

[@arch800]

It would be amazing to run chatgpt in any form locally

[@HCforLife1]

It would need 350 GB of vram minimum (from what I read) good luck with that xD

[@dinkusstinkus4396]

​@@HCforLife1 llama 13b was running on 40 gigs several days ago and there was talks that in a few hours it would become dramatically more efficient, and talks of a version someone got running on a cellphone.

[@dinkusstinkus4396]

​@@HCforLife1 also 350gigs of vram is only like 5$ an hour if you get your own private GPUs on runpod, cheaper if you do it on the bid pods,and even cheaper and lower vram cost on Google's cloud tpus, it may even be free on their tpus I'm not sure.

[@alceslacore]

really coool🤯

[@Scientists_dont_lie]

wow. that link picture screams "BEAST TECH"

[@JoeyBlogs007]

So I said this. Explain the plot of Cinderella in one sentence, where each word starts with a new letter in sequence from A to Z. No cheating ! and GPT 3.5 responded with; "A ball Cinderella dances; everyone fawns, given her incredible joyfulness, kindness, loveliness, mistreatment; nobility offers Prince quite romantic soirée, though unfortunately vexing, wistfully xenial yearning, zestfully."

[@flonixcorn]

Lets go

[@pastuh]

Quantum computer with GPT service which is in sync 24/7..... 1st question: Which stocks to buy

[@dinkusstinkus4396]

Right now quantum computers aren't useful for much other than possibly code breaking if you were hyper wealthy and motivated and patient, but mainly just QKD application for like, banks.

[@Disorbs]

i ask for code but it doesnt finish it idk if limit but when i do continue it keeps redoing it from the start and hitting the limit again anyway around it?

[@curtis1397]

Has anyone else noticed that when your coding with gpt, if you copy the code and make some fixes, somehow gpt knows. Several times I have asked it to alter some code it wrote and the next time it writes it with my fixes are included. I then asked if it could see my screen and it instantly errored. The changes were not coincence, it detected things like changing the name of a class.

[@KyleandPrieteni]

It actually was able to analyze error codes before 4 came out. But it's better at it now

[@PatriciaChavers]

This was fun 🤣

[@chriscodling6573]

it would be nice if it can help with specific appliicatiions like davinci, such a tutorial like questions

[@Heldn100]

There is any hope to get something like chat gpt but self hosted?

[@dnlb_peyn]

i think the example with rihanna uses a placeholder for clearer presentation, no?

[@ucheucheuche]

The fact that GPT-4 could read his handwriting 4:10 was wild. Let along what came next!

[@dekompose]

It’s 2023 I truly believe in open source now…

[@xXSjapXx]

My guess, it is the basic intelligence at download and learns per interaction locally. Maybe ask it where it gets it's new resources to change it's view and correct the answer?

[@ajvexports]

wish to write meta description of product images for my eCommerce furniture website, how can chatgpt help me?

[@noimagination9326]

How do you set A1111 to Dark Mode?

[@ArielTavori]

TIP: If you want to use an image as input for ANY chat bot, just input it into clip interrogator and copy paste the text into your chatbot. If you don't have a local install just use the one on HF...

[@ArielTavori]

Obviously this isn't going to code a website from a sketch like he showed in the demo, but it can still be very useful

[@chromery]

The Cinderella bug is actually very very useful information

[@ThomasTomiczek]

Chat GPT 4 issue at 8:30 - it can not REWRITE the sentence. This would mean keeping the meaning and there is no alternative to "stepsister". What you ask for is a NEW sentence, not a rewrite. A rewrite must keep the meaning - so it would need an alternative word to stepsister that fits the requirement. That is why it does not work. This is also likely why id proposed that - chatgpt goes forward (only). The sentence structure lead into a dead end, and stepsisters was like the only outcome. It can not, technically, go back and restart - the user has to control that (or an agent further up). Technical limitation of a forward only network with a single task approach.

[@Luna5829]

you don't have to open a new chat, that's what the regenerate response button is for

[@wanro027]

…Nice. When you say. A bit disappointing. When you talk about this amazing openAi

[@TheLoneWolf00]

for the game, specify you don't have Superman's reflexes LOL

[@nic-ori]

Thanks.

[@Aitrepreneur]

You're welcome

[@reifuTD]

wow, I can just imagine how easy it would be to make a simple RPG game with this this?

[@PushingThroughThePain]

The closer to human thought it becomes, the more variation it will have each time you ask the same question

[@FusionDeveloper]

I just wonder how good/bad the source code for the website is. I had a friend make a website with Microsoft Word a long time ago. All the page was, was a single image centered on the screen. The page took about 1 minute to load. I looked at the code and there was a tremendous number of pages of code, when all that was needed was a simple line of BR's and Center Img. Both pages looked identical, but my code loaded the page in 0-5 seconds, versus his that took at least 60 seconds.

[@AtanvarnoALDA]

K, can you please make a new video about training Dreambooth? Wiuth tools changing, models changing, versions changing, old tutorials just don't really work anymore.

[@JoeyBlogs007]

I think it's partly how you word your request. If you had said, "Explain the plot of Cinderella in one sentence, where each word starts with a new letter in sequence from A to Z." You will get ; A beautiful Cinderella dances, enchants, flees, goes home, invites, joyfully kneels, loves, marries nobly, overpowers pomposity, quickly receives stunning thankfulness, ultimately visits winsomely, X-tremely yearning, zestfully.

[@JoeyBlogs007]

X-tremely not really a word however, so it cheated there.

[@avatar2xl]

Can you do a video on budget AI stuff

[@FuZZbaLLbee]

0:12 Amiga 1000 😀

[@Corteum]

Be interested to know if it's any better now than it was 2 months ago when you first dropped this video.

[@GorilieVR]

As these tools rapidly evolve, i hope it will facilitate unprecedented tech innovation and help make the "metaverse" a reality

[@tjw2469]

Zuckerberg liked this comment

[@blkb33]

@@tjw2469read neuromancer and snow crash. All this ‘existed’ before he was born. He stole the name and idea from science fiction and tried to make it his own. Was too early anyways for the concept to be accepted.

[@mikatu]

LOOL in a video about evolution you are talking about the metaverse joke!

[@chariots8x230]

Can it scan a PDF textbook and summarize (or list the key points) from each chapter? That would actually be very useful to me.

[@Aitrepreneur]

Maybe once the image scan feature will be enabled

[@justanotherhuman3668]

Try to ask it in a prompt like “I need help summarizing the key points from the (pdf, or whatever type of document you’re using) that can be referenced here: (insert your url link here)” and I promise it will scan it. Doesn’t have to be gpt 4 I use this now. I’ve asked it to look at urls to gifs and it can say what’s happening in the gif. I’ve done this with articles and images as well. If it gets it wrong at first, correct it. Then it usually gets it right. At first when it scans the image or gif it’s more abstract in how it explains what’s in the image. Once you correct it, it will go the direction you wanted and you can extrapolate more information from that point. Don’t straight out ask it to access any links ever because it won’t. You have to be careful in how you prompt it like the example I gave you.

[@23ravensby98]

It likely already "knows" the work, and a gazillion reviews, articles, and summaries already written on it, so you would not need to scan it.

[@spinstate1005]

Try chatpdf

[@user-qd3rz7fb1t]

This exists, chatpdf

[@XX-_-XX420]

I mean its decent but still has a lot of strokes when im talking to it and asking it to do stuff. it will often get stuck in a loop of not finishing the answer and you needing to ask it about 20-40 times before it does it right, and due to the low cap its pretty bad now, I think the best is to kinda swap between 3.5 and 4 I guess see what 3.5 can do and if that dont work try 4. like it did help me out a lot, but for 20 bucks I expect more than 25 reqeusts and hour if I need to ask it to fix a issue with code 20-50 times before it stops having a stroke mid answer.

[@timhays2086]

How does the AI know that the information about Rhianna was after the date it contains information for. If it doesn't contain information from after that date, then it doesn't know anything abour Rhianna and the Super Bowl right?

[@woodenlake9629]

I'd like to see if chatGPT can make a doom like game that plays in a browser.

[@Tetsujinfr]

"A bug"? Actually, this is the most interesting part of the video. 1) it debunks the realities from the hype, 2) it may indicates that, as other generative models, there is likely a random seed at play in the background and the chat answers instances, for a same question, are not as deterministic as you may think. Just my 2 pennies. But I do not think it is a "bug".

[@studyrelaxloungestation9088]

It says info only available from 2021. But I've managed to make it discuss recent topics. Including the latest controversy from the H3H3 podcast that happened yesterday. It claimed it didn't know, but with a bit of manipulation, I got it to tell me everything about the podcast and what it thinks.

[@ribertfranhanreagen9821]

Waiting when game dev ca n use this for npc interaction

[@silvermushroom-gamifyevery6430]

It would be great to know how well GPT- 4 handles coding for app you don’t know . Being able to skip learning after effects or some programming language alone would be absolutely life changing.

[@HCforLife1]

Won't happen. It could work for MVP but tbh - still programmer could make it faster and better. Not to mention that this will not produce code for whole app on enterprise level. And last but not least - would you bet your own money with chatGpt to take all liability for any data leaks or money lost in transaction? We talking about potential millions of dollars of bankruptcy. This tech will be great for MVP runs but not for whole complex app building. People without coding skills will get themselves into trouble

[@alexanderg9670]

@@HCforLife1 Not yet, at least. If it can write simple code, it will write complex code at some point. Creating GitHub repos and renting AWS for demos by a simple prompt

[@DavidKnowles0]

@@HCforLife1 At how many times the cost. An already GPT- 4 is being use to identify security flaws in human written code.

[@BoredInTheComments]

Just one more thing to make life easier and make people more useless.

[@silvermushroom-gamifyevery6430]

@@HCforLife1 - I was talking about using GPT for usecases where you know the underlining computer science / accounting / skill behind what you want an app to do, but don't know the exact expression / formula / steps to make the specific app do what you want it to do. For instance, you know how to add time, but you probably don't know how to add time recorded in different formats into a project page without manually adding it yourself in Notion, or some other project management app you're using for whatever reason. Here, you know when the app is doing what you want and is following all the required steps of your trade and when it fails to do so, but the exact app instructions to achieve this are an enigma to you.

[@jackmartin1146]

Hello, how do you select the 'model' through the dropdown in the ChatGPT UI? Can you share which chrome extension you are using for it?

[@Aitrepreneur]

You need to be subscribed to chatgpt plus it's 20$ a month

[@electronresonator8882]

18:18 Mr. AI overlord 😆a.k.a PNG youtuber K, my suggestion for this programming prompt, 1st. ask "do you know the game PONG ?", (if the AI doesn't know, give google search result, wait till they AI know about it) 2nd ask "can you write me a code similar to the game in Javascript?" 3rd save the code and run (see what's missing) 4th ask "can you add [insert feature] here to the code?" 5th repeat 3-4 until you get what you want my personal note : chatGPT is so "confident" in matter of calculations, it simply write an answer that connect a wrong formula to the correct answer, so many times, that I have to add "calculate this part in detail", just to uncover that mistake in a single prompt

[@festro1000]

@7:04 it also forgot the letter O.

[@Kilgorio]

wow

[@donbomBL]

I tried it for writing greetings and articles in my topic, but it all looked like it didn't understand my topic, but just stuffed a keyword into any article about a personal creator doing something It was especially funny in terms of adjectives that could not be used for this specialty in any way. I tried several times, but apparently there wasn't enough information about it and the AI couldn't answer anything at all

[@donbomBL]

I'll explain, my direction has existed since the late 70s and there should be enough material for the chat to get at least an article from wikipedia, but apparently not

[@RealityCheckTime]

I think that the problem comes from switching from GPT3.5 to 4 in the same chat pannel

[@justinwhite2725]

I've never been able to change in the same chat. Had to create a new chat.

[@Loopa21]

okay. tnx k.

[@Antalion20]

I think you misunderstood the prompt they provided with the Rihanna wikipedia article: I don't think they meant to imply that they just gave it a link, but that they copy/pasted the words from the article into the chat. Obviously doing that would look very messy in their slick presentation, so they wrote the abstracted representation instead.

[@Aitrepreneur]

Yes I understood it a bit too late unfortunately :D

[@Vestat1]

PLEASE PUT THIS IN STARTSHIPS AND LETS EXPLORE THE STARS!!!!! ✨️

[@ThomasLe]

you should learn how to use an IDE like VSCode or Webstorm instead of creating individual files like you do using file explorer. having separate files for different types of files like javascript and CSS makes things easier to maintain in the long run. the reason it is cutting off is because chatgpt has a limit per response so that's also why it breaks it up into separate files. using an IDE would make things orders of magnitude better and more efficient.

[@razersharpinfinite]

More like CloseAI!

[@jeffcessna]

Within 5 years you will sign into your xbox and write out a gaming idea and be able to play it on demand.

[@formdusktilldeath]

You do understand that it's a heuristic algorythm? There's always a possibilty for failure, it's inherrent.

[@XzaroX]

Thumbnail source?

[@J0SHUAKANE]

K, my AI Overlord, we need a guide to locally install gpt-j please🙏

[@freestyletow1]

yes

[@Aitrepreneur]

I'll see what I can do

[@eagillum]

4:34- how to gain access

[@T___Brown]

@8:00 there is another mistake.... O is missing

[@Platoface]

Is it good, bad? Well it was down pretty much all day yesterday.

[@ozramblue117]

And job applications.

[@stevengaffney1363]

This is the beginning of the end enjoy your phone

[@MidasNG]

Why does its training end the same year as ChatGPT? :(

[@unknownuser3000]

Can it do explicit rp? Useless if it cant!

[@HonkaiHistory]

For me it actually was the other way around. I think the old model made the more interesting titles people would click into while the 4 model titles were pretty generic and didnt make much sense for some.

[@loganjohnson8010]

Can I use gpt to summarize this video for me

[@OwlishGeorge]

GPT knows you now. That's what the GPT4 model tells me.

[@Jaughn]

10:10 They were citing the website. You should have used the actual text from the webpage and not the link, then it would have worked.

[@PokettoMusic]

>you have to pay to get access what? I didnt pay anything, did I do something wrong?

[@SoCalGuitarist]

K got a drunk seed :P