ChatGPT Can Now See! GPT-4 Release, Image-Chat and Edit & Midjourney v5

what would happen if you gave chat GPT eyes well that's exactly what this guy did gbd4 is coming this week and if it's multi-modal we can predict what it will be capable of Microsoft just introduced visual chat EBT it allows you to send receive and edit images while you are in chat mid Journey just revealed version 5 Wonder Studio changes 3D animation forever what would happen if you gave chat GPT eyes well that's exactly what this guy did and the results are pretty astounding let's give chat gbt the power of sight by combining it with computer vision to have it answer questions about something we're looking at what's this building this building is that the Young Museum in San Francisco California when is it open that the Young Museum is open Tuesday through Sunday from 9 30 a.m to 5 15 p.m by using text to speech speech to text and azure's computer vision services we can give chat EBT a voice and eyes so I imagine this technology going directly into a phone that can see and hear exactly what you're doing and it can give you context it can ask questions and it might say like hey over there look there's a celebrity and I'm like who is it and then it says oh sorry it's a cardboard cutout of Dwayne The Rock Johnson gbd4 is coming this week and if it's multi-modal we can predict what it will be capable of first it will be able to do visual IQ tests as you can see here this is an IQ test and the question would be the following image is and then it's choosing in this one as the correct answer second thing would be optical character recognition for example you could take a screenshot of your notebook and then you can digitally make that into chat GPT or perhaps you have a PDF that isn't in text form it's actually pixels you could easily read and write that the third thing would be multimodal chat this is where you can input a picture and you can talk about the picture itself you can even ask follow-up questions to the picture have a conversation around the picture itself for example what's in the picture looks like a duck that's not a duck then what is it looks more like a bunny why it has a bunny ear the fourth thing would be broad visual understanding so you can answer things like why did the little boy cry where the answer is because his scooter broke or what is the hairstyle of the blonde called the answer being ponytail and the last thing could be audio and speech recognition we've already seen this with open ai's whisper and should be fairly easy to integrate note that this is what Andreas Brown the CTO of Microsoft Germany said open AI has not stated that chat EBT will be released what happens if you put chat GPT inside of a robot well you get this this robot is called Amica and his the world's most advanced human looking robot and to me it gives all of the bad vibes something so Eerie giving a robot human looking facial expression let's just take a quick look at this hi Amica how are you hello I am doing well are you yeah I'm good I wanted to ask you about applications for humanoid robots what are they used for there are many possible applications for humanoid robot examples include helping people with disabilities as a company you can act as a companion for me yes I can act as a companion for you would you do to make me feel happy there are many things I could do to make you feel happy I could listen to you offer advice or support do so do so enjoy together or simply be a friendly and positive presence in your life the facial expression she makes when he asks about can you make me happy is you don't even need to say any words like yo I don't want to make you happy dude can you be a companion to me can you make me happy to me it's really creepy it takes off all the boxes of something that I don't really like and putting an AI into a human looking body is kind of weird I imagine asking it something like do you want to end humanity and the robot is like no I don't want to end Humanity humans are great Microsoft just introduced visual chat gbt it allows you to send receive and edit images while you are in chat hear it asks can you generate a cat for me sure here is the cat for you could you replace the cat to a dog and then remove the book and it just does it cool to generate the candy edge of this image it kind of edits it into this how about filter and can you generate a yellow dog based on this image and it paste an image in here I'm sure here is a yellow dog for then it sends a motorcycle picture which color is this motorcycle it's black can you remove the motorcycle and it just removes it from the image which is kind of crazy cool what's in the image the image contains a yard when will we actually be able to access this I don't know hopefully chat gbt4 will get launched and has this feature built in to me the most fascinating thing is that you can talk about about the images or you can probably just copy and paste a bunch of images or text instead of having to write a bunch of things making us all way more productive so we can save a lot of time mid Journey just revealed version 5 which of these images would you vote on the mid-journey team said that these images are not representative of the V5 algorithms we will release soon these images are plain slash boring slash unapiniated to help the system learn whether V5 should have a different style from V4 you should notice images are higher resolution and have generally better detail and stylistic ranged in V5 to me it does a lot better at the realism people and what you all have been waiting for hands look at this image and tell me you can say that this is AI art I could never tell this is a photo or not but it's 100 AIG generated what about this these are just gonna get crazier from here so this guy asked Bing AI to explain the svb bank collapse in monkey banana terms and this is the message being AI sent me try to explain Silicon Valley Bank like big tree with many bananas many monkeys like bananas and put their bananas in tree tree promise to give more bananas to monkeys later but three bad tree lose many bananas in storm three not have enough bananas for all monkeys monkeys get scared and want their bananas back many monkeys run to the tree and grab their bananas but some monkeys too late no more bananas in Tree Monkeys sad and angry big monkey come to take over three big monkeys say he will find more bananas for monkeys who lost their bananas a monkey give some bananas to some monkeys now and more later but some monkeys have to wait long time for their bananas other trees see what's happened to Big Tree and get worried other trees have many bananas too but not lose them in storm other trees hope their monkeys not run away with their bananas too me hope you understand now and I think the funniest thing is that it actually continues the conversation with me understand now I'm gonna talk about the two favorite AI tools that I found today Wonder Studio changes 3D animation forever what you're seeing here is an AI tool that automatically animates lights and composes CG characters into a live action scene no complicated 3D software and no expensive production Hardware all you need is a camera to me this is unbelievable I couldn't believe my ice when I saw this because in the past you have to first paint out the character second have an expensive mole cap suit for like a thousand bucks third you have to rig the animated character and then fourth you have to put it into the scene make the lighting look good and make the colors look the same as the video this is an extremely hard task that you can now apparently do with just a snap of the fingers and you got all of this done for you and I seriously thought that this was fake but then when it got real to me was Steven Spielberg is in The Advisory Board blending the physical reality with the digital reality it's not open to the public yet but you can join in the link in description to the closed beta next tool is you can now get all big language models under the same roof for the first time look you can just submit and then you will have all different language models put out there different answers so you can compare the output of open AI chat gbt and Tropics Claude and coherence language model in a single playground I'll give a link down in the description I'll show you what happens if I for example write how can I get people to subscribe if I click on submit you will see that all of them will generate different answers based on the prompt that I was writing plot is generating bullet points for us like provide value build trust hugging face is literally saying put a subscribe button on your channel cohere is just putting out basic text here open AI chat dbt3 gives us a list of seven things that we can do to get you to subscribe so if you want a video newsletter every other day about what's going on in AI click the link in the description sign up and I'll see you tomorrow peace

Channel: AI Andy

Views: 199,048

Keywords: chatgpt, gpt-4, gpt4, what is chatgpt, how to use chatgpt, ai news, new in ai, chatgpt explained, openai chatgpt, how does chatgpt work, artificial intelligence, chat gpt explained, openai, open ai, midjourney v5, news, midjourney ai, midjourney, ai art, midjourney tutorial, olivio sarikas, matt wolfe

Id: Y6TrtFHGa0E

Length: 10min 35sec (635 seconds)

Published: Mon Mar 13 2023