GPT 4o - Deep Dive Review - AGI? - ChatGPT massive improvements
Vložit
- čas přidán 17. 05. 2024
- OpenAI shows Chat GPT 4o with massive improvements. Here is a deep dive review with many examples.
Link from the Video: openai.com/index/hello-gpt-4o/
#### Join and Support me ####
Buy me a Coffee: www.buymeacoffee.com/oliviotu...
Join my Facebook Group: / theairevolution
Joint my Discord Group: / discord
AI Newsletter: oliviotutorials.podia.com/new...
Support me on Patreon: / sarikas - Jak na to + styl
It's gonna be exciting few last years !
🤣
We didn’t saw the beginning of all this, but we might see the end 😊
that new voice in the demo isn't released yet, that's why your experience was different
Good to know
Your question about Celsius. Do remember that this is not just restricted to context. I'm pretty sure it understands that the variations in temperature are very much obvious as Celsius. Given that a movement between 0 and 34 is definitely going to be Celsius, and very unlikely to be Fahrenheit. Even if we posit that it is possible that this is an Arctic region, that range of temperatures is still found nowhere on Earth. My guess is that it's simply assumed it based on the travel between the temperature. Or maybe it is in the comments of the code . But I think if it recognizes that July is in the '70s for Fahrenheit if the temperature is Celsius, this makes a lot more sense.
Ok, Love the shirt Olivio!!😻😻
thank you :)
you should try Pi from Inflection AI... really crazy natural voice interaction, way better than the corny ChatGPT-4o voice.
Sounds very monotone to me don't see much difference to the regular gpt4 voices
sounds also very good. i like the P1 voice :)
I hope multi emotive voices are going to be one of the next major features in a few months or sub versions ,with fictional text input understanding.
This will give us narrators and multiple characters to convert text to audio books, which would work for novels ,comic and even dubbing movies or even having multiple virtual people to talk to in things like virtual debates , or prepping for the question part of talk you may be giving or helping people deal with anxiety in group conversations or making a table top game where you need one or two extra characters in your table top .
Another great video. I am looking for a way to combine two images, I have an image of a cheerleader of a football team, and I have a logo, I want to have the cheerleader holding a sign with that logo. do you have a video that explains combining images?
cheers
Absolutely MINDBLOWING!
The stuff that AI can do to process audio and then discern individual voices is one of the most exciting and practical applications. The AI might sometimes hallucinate and add extra people or combine people on the worst recordings (and recordings of twin siblings?) but in general this will be extraordinarily useful to get "usable" transcripts of live meetings and older recordings. Likewise, the AI ability to read written text in an image is exceptionally practical, especially when combined with increasingly better language translation. While it's amazing and fun to see what happens when AI is allowed to "get creative," it's mundane powers like these that excite me the most, because they are meant to "just work" so you don't have to kick the AI repeatedly too get a usable result.
Actual best review of the new product.
Thank you very much 🙂
Yeah, pretty freaky. I'm not sure sure I'd want my AI to pretend to be human, but the emotionality is an interesting addition.
Also, is it just me, or does the talking AI at 5:12 look like the silhouette of HAL?
Is there a way with Comfyui to have ControlNet randomize poses from a batch folder?
Yes, just load the input images from batch (bath load node) and set the render batch the the correct amount of images
@@OlivioSarikas Awesome thank you!
I don’t think the image generation is Dalle anymore. Should be native gpt-4o image creation if I understood this correctly
Love the thumbnail ahahah ❤
thank you :)
I can't find the app anywhere. Is it rolled out to few testers only?
Yes, the chat gpt app has been out for months. I'm in Europe
@@OlivioSarikas I know the normal app, but it has absolutely nothing to do with the one shown in the OpenAI demos 🤷♂️
All the interactivity is not yet implemented in the version that I can get in the app store. No voice, no video, no camera recognition, different design ezc.. It's still just the old chatwindow.
@ProfWho-ut5he, you can change the version by clicking on the three dots in the upper-right corner.
@@SuperQuni No, there I can only switch the models. It doesn't offer the interactive features.
@ProfWho-ut5he For me, it gives me the options to describe images that I took, etc. The full live camera feature is not out for free now. If you click on the headset symbol, you can talk to gpt-4o like in the video shown.
I think the model wants to awnser so fast it's hard to give it multiple task at once before it start handling the first request. That's why he talked so fast and that's why it sounded like it hallucinated at first an he restarted the task. The model responded to all his input in order but cut itself off in the same pattern he gave instructions.
i cut out a bit of waiting time, but it is indeed very fast and very good :)
The current app does not have the speech and audio features of the video.
it has the speech function, but not yet as amazing as they show it and the video function comes later
THIS is how children should be educated - through personal AI, not in drug-infested schools where they are bullied and traumatized for life. time for a revolution!
But all the bullying gave me character and then I became a bully Hunter some of the best days of my life
I don't think the "builds character" argument should be made by anyone under, say, 90 years old.
Hardship does not (necessarily) equal trauma. Anyone who is traumatized is, by definition, not stronger for it.
yes, totally. now we can all have private teachers and it is never too late to learn new things
@augustuslxiii I'm 46 I might not be 90 but it sure as heck gave me character most martial arts build peoples character whether it's on the streets or in a school
I admit at first I thought this o version was mostly a gimmick.. but after trying it, I was really blown away and somehow it didn't feel like chatting with chat gpt. I just can't get the voice control to work on pc yet though despite being on gpt-o as well.. I guess it will take a bit of time to implement
Yes, the update is really mind blowing
It's called Omni not because of its multimodal capabilities but because they made it free it's going to be omnipresent. I think we owe openai some credit here. While everyone's been racing to catch up they've been working on making it free
really? sounds good too. I rather thought omni because of it being omni precent in all forms of media and in any way you need it
Old news by now, seriously. I was expecting your own testing with 4o. I compared it to my prev 4 (not 4o) prompts and 4o is at least 10 human IQ points higher 😊
It may be old news for you but for many others this is an excellent introduction. This video was obviously not made for you.