Awesome. Thank you very much for this synthesis. I am an expert in accounting and i wanted to teach chatgpt (or Bard by the way) advanced accounting so that it is able to answer tricky questions as to IFRS accounting for example. I think i am going to study the possibilities (GPT's) brilliantly described in this video.
I was at the edge of my seat watching, but i was just like... all these updates are cool, but i bet Igor will give some amazing examples how to use it practically@@aiadvantage
I'm highly confused. So, we ChatGPT users won't get larger context windows? It's only for API and Playground? I hope not because that would be unfair, and that would mean Claude has ChatGPT beat. $20 a month is not cheap for most people. We should get the same as API users. Am I missing something?
Ya. At this point someone needs to review all the ChatGPT interfaces. There’s ChatGPT browser based at $20/mth. The app which talks, is free and is kinda fun and the playground which is pay as you go and is a bit expensive depending on your usage style. There may be other ways now to develop a workflow but that’s always been my goal. Unfortunately, it’s impossible to get some kind of a workflow going because everything keeps changing. And I would really like an open LLM or multimodal large language model so that I don’t have to worry that I’m only getting part of the story. The censorship is pretty bad and I think the censorship in Canada is particularly bad. We don’t have access to a lot of information because of Bill C 20. Canadians basically live in North Korea.
Hi Igor, can you make a video on a few of the different use cases for different custom GPTs, what it means for increased context window sizes, and how APIs can be used? Some of this content is abstract and difficult to apply to our day-to-day lives.
Super excited for GPTs. I've started creating one that I've had in mind for almost a year. It's taking awhile to train, but it'll be so worth it. I'm also thrilled to hear that we'll no longer have to pick which model to work with. Being stuck with a model for the duration of a chat sucks when new ones come along, so I'm wondering if that feature will be retroactive, ie.., "old" chats will now be upgraded to the most currently all-inclusive model. Now THAT would be amazing.
I've spent the last 6 months building a platform with AI avatars (glome social), and I think the next iteration of GPTs will pretty much steamroll over my app :P Up until yesterday, I hadn't heard your critique on 'wrapper apps.' I thought my app was somehow better than that. Well, to be honest, it may not be, and I don't have the money nor the devs to build something bigger and better than what OpenAI will in the next 12 months. I guess we shouldn't really listen to Sam's interviews that much, nor take YCombinator's advice to heart. They preach that OpenAI's focus is on AGI and they will let startups flourish on top of OpenAI's API. But at the end of the day, OpenAI always chooses the predictable path to address the missing features and needs of their users. And if you happen to be one of the devs who solved a 'missing feature' issue-if the feature was really needed and your app is a success - you can be sure they will ship it too, ASAP. (ChatGPT mobile native app, ChatGPT via voice, PDF import, etc.)
Yeah it's kinda sad but also most people expected this. It's the same tactics that Microsoft uses. Buy it or re-build it and bundle it with their offer. The niche apps will prevail but the big and obvious ones are all going to be bundled into their main products.
Thats what i though about developing generic apps. So i went for the path of offering custom-develoments for my customers that (for now) are not solved with Openai (or other) big companies. As lot of people commented, yesterday Openai dynamited lot of start-ups in just one hour.
I personally believe that unless you have a unique dataset at your disposal you can build AI on top off, there's no point competing with all these companies...
We don't all actually have GPT-4 Turbo inside of ChatGPT yet, OpenAI just programmed GPT-4 to say that's when it's cutoff is. Try uploading a 100k token file or larger and it will not respond because it says that it is too long. Once the model picker is gone for you, that means that you have GPT-4 Turbo. They have started rolling it out to people already.
They need to hurry up. I have been waiting, and waiting, and waiting and refreshing every 5 seconds like a crazy person. My girl is starting to think I am going bonkers and threatening to leave.😂
They should add a way to record and save the text to speech thread so people can easily create audiobooks and not rely on extra subscriptions from other services
Exactly. I was thinking the same thing. Then there would be no need for us to pay ElevenLabs an extra fee per month. It would just make since if it has text to speech that we can save it.
FYI there is a failure of direct retrieval with GPT-4 using the new OpenAI Assistant API. GPT tokenizes text and creates its own vector embeddings based on its specific training data. The new terms and sequences may not connect well to the pretrained knowledge in GPT's weight tensors. There was no semantic similarity between the new API terms and GPT's existing vector space. This is a fundamental issue with retrieval augmentation systems like Rag - external knowledge is not truly integrated into the model's learned weights. Adding more vector stores cannot solve this core problem. The solution is to have multiple learned "knowledge planes" with trained weight tensors for specific tasks that can be switched in. This is better than just retrieving separate vector representations.
exactly and did you see the write up for the proposed AI regulations submitted as executive order so can be undone with new office but pretty much BS and gateway to the government potentially having access to the data link: www.whitehouse.gov/briefing-room/presidential-actions/2023/10/30/executive-order-on-the-safe-secure-and-trustworthy-development-and-use-of-artificial-intelligence/
For me currently, it says GPT4 has DALLE3, data analysis, and web browsing but I guess it's not turned on yet or something because it says it can't generate images or browse the web
Same here. We all have been waiting on pins and needles, and still no update. They said we were going to have it the same day, but here we are still waiting and pulling our freakin hair out by the root.🤬😱
Were you able to fit your entire character prompt description into the AI assistant instructions? I would have assumed there's some sort of a character limit
Thank you for this! This was insane. Super excited! BTW, when i saw the thumbnail for this video it was next to another thumbnail for a different RUclipsr that looked exactly like you. It’s your YT thumbnail doppelgänger! I saved a screenshot if you would like me to share it with you. Just let me know where to send it. I can send it to your newsletter email if you like. LMK.
Igor - Do you have a template for an Assistant for a Recruiter? I would primarily be interested in the assistant who would be able to help with sourcing, using Boolean, as well as other techniques to identify candidates.
Hey Igor, Im a chat gpt plus member in Canada, but I do not see the custom gpt option available in my chat gpt. Do i need a special developer key? how do I access the features?
Where is the GPT turbo I was asking the GPT until when is knowledge so it tell me until 2021 november amd Im a member of the 'plus' so what can i do to get this ?
Great video! But when I go to Playground, I don't see where they offer GPT 4 Turbo. I only see them offer various versions of GPT 3.5 Turbo. (And yet I am a GPT Pro user.) Where on the page do you see that they offer GPT 4 Turbo?
@@aiadvantage I'm frustrated. I can't see GPT4-1106-Preview or any GPT4 on Playground. Only various GPT 3.5. Am I looking in the wrong place? I'm on Playground, I clicked the "+" to "Create an Assistant" and then on the column on the left there is a pulldown menu for "Model." And I'm only offered GPTs 3.5. Any idea why I don't see GPT 4 even though I'm a GPT Pro user? @SebtorDude
Could you please explain what I can do with OpenAI as if I’m a confused grandfather? This is very confusing. I barely understand what you are talking about. I want to use this for my hotel business, and to help me write a book. Thank you 😅
I want a gptv included in SD's latent or regional prompt with control net line art. and linked to my photo editor. and now when I draw skitches and I click a button on my keyboard it saves the current image gives it to the gptv, gptv tries to make sense of it and for each position in the image using regional prompt it will put a word to what that is and after it finishes the SD extensions will run the SD will generate 4 sets of the pic given the previous info. and so I can work as an artist and at the same time see my little assistant's generations with my favorite model. and why not with my favorite loras too. the prompting will be done by gptv not me. I only draw. :D
@@aiadvantage Are you 100% sure the large context window is only for API? I heard it was for us ChatGPT Plus users too. If not I really think that is unfair seeing that we pay $20 per month.
01.03 bro, please use a pop filter with your mic. Also, this vocals are so bad; there is too much bass and muddy. Please eq your vocals before uploading.
Yeah don't think so. Most of them built on the OpenAI API and they are re-building all of this from the ground up and not just stealing.. Niche applications will prevail but the big and obvious ones are just going to be added as a feature. Microsoft is a master at copying competitors ideas and bundling them. OpenAI is just copying a proven corporate playbook.
In that same conference, OpenAI announced Copyright Shield to protect customers and users. OpenAI themselves will defend customers and pay the costs incurred if you face any legal claims around copyright. Right now that only applies to ChatGPT Enterprise and the developer platform, but they're making some big moves to protect against the very thing you say will 'crush' them.
THIS was a perfect, hi-speed summary of all these new developments shared today! your LIVE today was incredible too! THANK YOU one question re: the new update: Is gpt4 trainable now like 3.5 was -- can we "fine tune" the model? or is that irrelevant now with the individual GPTs being created? PS: your new product release = Freakin Mind-Blowing!!!! THANK YOU!
Appreciate it :) Yes they announced GPT-4 fine tuning but it's only coming to enterprise users at first from what I have heard so we will have to wait for this. It would be super relevant for the GPT's as you could use fine tuned models as a base which is superior then just using RAG (combining knowledge documents with chat outputs) which is what they introduced now. So as per usual they are taking it step by step. First giving us individual agents "assistants" and the ability to add external data to custom charbots "GPT's" and then over time they will add multi-agent and gpt-4 fine tuning. PS: Fantastic emoji use right there
Thank you, Igor, for the magnificent update that takes us to the next level!
Time to explore it all :)
hi please reply do i have to pay for tokens when someone uses my gpt in the gpt store@@aiadvantage
One thing I will say is those GPTs are going to be a game-changer for us fiction writers. We can create one for each novel or story.
wait.......i saw you comment on OpenAI's Channel
@@aashsyed1277 Yeah, probably so. lol🤣. I've been obsessed with this new update, but unfortunately, I still don't have it yet!😣
I’ve been using ChatGPT to help with the structure for my fiction story and it’s so annoying when it hallucinates and adds stuff that never happened
Awesome. Thank you very much for this synthesis. I am an expert in accounting and i wanted to teach chatgpt (or Bard by the way) advanced accounting so that it is able to answer tricky questions as to IFRS accounting for example. I think i am going to study the possibilities (GPT's) brilliantly described in this video.
Sounds like a great plan as it could be a real benefit to your clients.
Igor, always a pleasure!!! You absolutely keep me in the know all the time! 🙏🏾.
Glad to hear it's appreciated
Amazing summary, thanks Igor!
your videos are just as great and have a unique charm to them. no i am not a bot
Thanks but I bet you watched all of it live anyway :D
I was at the edge of my seat watching, but i was just like... all these updates are cool, but i bet Igor will give some amazing examples how to use it practically@@aiadvantage
I'm highly confused. So, we ChatGPT users won't get larger context windows? It's only for API and Playground? I hope not because that would be unfair, and that would mean Claude has ChatGPT beat. $20 a month is not cheap for most people. We should get the same as API users. Am I missing something?
Ya. At this point someone needs to review all the ChatGPT interfaces. There’s ChatGPT browser based at $20/mth. The app which talks, is free and is kinda fun and the playground which is pay as you go and is a bit expensive depending on your usage style. There may be other ways now to develop a workflow but that’s always been my goal. Unfortunately, it’s impossible to get some kind of a workflow going because everything keeps changing. And I would really like an open LLM or multimodal large language model so that I don’t have to worry that I’m only getting part of the story. The censorship is pretty bad and I think the censorship in Canada is particularly bad. We don’t have access to a lot of information because of Bill C 20. Canadians basically live in North Korea.
Hi Igor, can you make a video on a few of the different use cases for different custom GPTs, what it means for increased context window sizes, and how APIs can be used? Some of this content is abstract and difficult to apply to our day-to-day lives.
Super excited for GPTs. I've started creating one that I've had in mind for almost a year. It's taking awhile to train, but it'll be so worth it. I'm also thrilled to hear that we'll no longer have to pick which model to work with. Being stuck with a model for the duration of a chat sucks when new ones come along, so I'm wondering if that feature will be retroactive, ie.., "old" chats will now be upgraded to the most currently all-inclusive model. Now THAT would be amazing.
I've spent the last 6 months building a platform with AI avatars (glome social), and I think the next iteration of GPTs will pretty much steamroll over my app :P Up until yesterday, I hadn't heard your critique on 'wrapper apps.' I thought my app was somehow better than that. Well, to be honest, it may not be, and I don't have the money nor the devs to build something bigger and better than what OpenAI will in the next 12 months.
I guess we shouldn't really listen to Sam's interviews that much, nor take YCombinator's advice to heart. They preach that OpenAI's focus is on AGI and they will let startups flourish on top of OpenAI's API. But at the end of the day, OpenAI always chooses the predictable path to address the missing features and needs of their users. And if you happen to be one of the devs who solved a 'missing feature' issue-if the feature was really needed and your app is a success - you can be sure they will ship it too, ASAP. (ChatGPT mobile native app, ChatGPT via voice, PDF import, etc.)
Yeah it's kinda sad but also most people expected this. It's the same tactics that Microsoft uses. Buy it or re-build it and bundle it with their offer. The niche apps will prevail but the big and obvious ones are all going to be bundled into their main products.
Thats what i though about developing generic apps. So i went for the path of offering custom-develoments for my customers that (for now) are not solved with Openai (or other) big companies.
As lot of people commented, yesterday Openai dynamited lot of start-ups in just one hour.
Does it make sense to build anything outside of something in the GPT store?
Damn. I just went and surfed glome social. Looks like OpenAI took the idea from you. Appearence looks the same.
I personally believe that unless you have a unique dataset at your disposal you can build AI on top off, there's no point competing with all these companies...
We don't all actually have GPT-4 Turbo inside of ChatGPT yet, OpenAI just programmed GPT-4 to say that's when it's cutoff is. Try uploading a 100k token file or larger and it will not respond because it says that it is too long. Once the model picker is gone for you, that means that you have GPT-4 Turbo. They have started rolling it out to people already.
They need to hurry up. I have been waiting, and waiting, and waiting and refreshing every 5 seconds like a crazy person. My girl is starting to think I am going bonkers and threatening to leave.😂
@@BionicAnimations Broo 🤣
@@GlobalWorkings 😂
They should add a way to record and save the text to speech thread so people can easily create audiobooks and not rely on extra subscriptions from other services
Exactly. I was thinking the same thing. Then there would be no need for us to pay ElevenLabs an extra fee per month. It would just make since if it has text to speech that we can save it.
Thanks much Igor for summing up this announcement from OpenAI. More power to your channel.👍🏼👍🏼
Absolutely mmmassive! They've raised the bar once again, as if being at the top wasn't enough!"
This sums it up well
We all know that this is a new era my boys
This is the times of legends
We will forever be remembered
FYI there is a failure of direct retrieval with GPT-4 using the new OpenAI Assistant API. GPT tokenizes text and creates its own vector embeddings based on its specific training data. The new terms and sequences may not connect well to the pretrained knowledge in GPT's weight tensors.
There was no semantic similarity between the new API terms and GPT's existing vector space. This is a fundamental issue with retrieval augmentation systems like Rag - external knowledge is not truly integrated into the model's learned weights. Adding more vector stores cannot solve this core problem.
The solution is to have multiple learned "knowledge planes" with trained weight tensors for specific tasks that can be switched in. This is better than just retrieving separate vector representations.
The major concern remains: Privacy.
Imagine what Microsoft will have access to with all these private files and data uploaded to OpenAI’s servers.
exactly and did you see the write up for the proposed AI regulations submitted as executive order so can be undone with new office but pretty much BS and gateway to the government potentially having access to the data
link: www.whitehouse.gov/briefing-room/presidential-actions/2023/10/30/executive-order-on-the-safe-secure-and-trustworthy-development-and-use-of-artificial-intelligence/
Yup and the censorship. The gates narratives are sus
I’m so glad we are using the term code interpreter again. 😁
For me currently, it says GPT4 has DALLE3, data analysis, and web browsing but I guess it's not turned on yet or something because it says it can't generate images or browse the web
Same here. We all have been waiting on pins and needles, and still no update. They said we were going to have it the same day, but here we are still waiting and pulling our freakin hair out by the root.🤬😱
That chatgpt assistant is going to destroy many jobs
And the rest!
Hard to argue with that statement
No they will enhance them
In ChatGPT, I'm not clear on what the difference is between creating a customized Chatbot versus creating an assistant ?
Hey, will you be having a black friday deal on your business blueprint or should I buy it now?
I'm confused of which subscription to take.Claude or ChatGPT .Any suggestions?
So excited for gpts
I can't see the new updates on my plus account. I'm particularly interested in the custom chatbot, when will this be available in the UK?
I don't have it either. They said the store will be available later this month.
Thanks for summarising everything for me 😊
Great video Igor !
Thanks :)
Here's an idea use the vision api to watch a screen recording to know what action its doing ...
This seems like a really good one. I can't wait to see what people come up with.
Superb presentation thank you very much.
Thanks for the kind comment
Were you able to fit your entire character prompt description into the AI assistant instructions? I would have assumed there's some sort of a character limit
Yep worked great! 32000 character is the limit on it!
@@aiadvantage whoa! That's awesome
Always great updates! THX
Great update. Will look to see if they have reached my account in a couple of months 😂
Can you make a video on how to make a GPT
Thank you for this! This was insane. Super excited! BTW, when i saw the thumbnail for this video it was next to another thumbnail for a different RUclipsr that looked exactly like you. It’s your YT thumbnail doppelgänger! I saved a screenshot if you would like me to share it with you. Just let me know where to send it. I can send it to your newsletter email if you like. LMK.
Sure send it :D
Igor - Do you have a template for an Assistant for a Recruiter? I would primarily be interested in the assistant who would be able to help with sourcing, using Boolean, as well as other techniques to identify candidates.
Hey Igor, Im a chat gpt plus member in Canada, but I do not see the custom gpt option available in my chat gpt. Do i need a special developer key? how do I access the features?
Thanks, this was very useful!
Where is the GPT turbo I was asking the GPT until when is knowledge so it tell me until 2021 november amd Im a member of the 'plus' so what can i do to get this ?
Great video! But when I go to Playground, I don't see where they offer GPT 4 Turbo. I only see them offer various versions of GPT 3.5 Turbo. (And yet I am a GPT Pro user.) Where on the page do you see that they offer GPT 4 Turbo?
Is it the GPT4-1106-Preview? If so, then I can see it on Playground already.
Yes that's the one!
@@aiadvantage I'm frustrated. I can't see GPT4-1106-Preview or any GPT4 on Playground. Only various GPT 3.5. Am I looking in the wrong place? I'm on Playground, I clicked the "+" to "Create an Assistant" and then on the column on the left there is a pulldown menu for "Model." And I'm only offered GPTs 3.5. Any idea why I don't see GPT 4 even though I'm a GPT Pro user? @SebtorDude
Thanks for the video
Are there any changes to chat GPT 3.5 ? As a hobbyist , I'm just using the free version.
Could you please explain what I can do with OpenAI as if I’m a confused grandfather?
This is very confusing. I barely understand what you are talking about.
I want to use this for my hotel business, and to help me write a book.
Thank you 😅
Those gpts are awesome!
I absolutely agree. This is going to be so fun.
I want a gptv included in SD's latent or regional prompt with control net line art. and linked to my photo editor. and now when I draw skitches and I click a button on my keyboard it saves the current image gives it to the gptv, gptv tries to make sense of it and for each position in the image using regional prompt it will put a word to what that is and after it finishes the SD extensions will run the SD will generate 4 sets of the pic given the previous info. and so I can work as an artist and at the same time see my little assistant's generations with my favorite model. and why not with my favorite loras too. the prompting will be done by gptv not me. I only draw. :D
i am unable to find turbo model in playground
Me too.
It's called "GPT4-1106-Preview"
Keep up the good work Igor :)
;)
How about jukebox the audio ai that openai kinda forgot about?
Is chat gpt turbo available only on playground or also in the Plus version?
Both! The large context is exclusive to the API though.
@@aiadvantage Are you 100% sure the large context window is only for API? I heard it was for us ChatGPT Plus users too. If not I really think that is unfair seeing that we pay $20 per month.
@@aiadvantage If us Plus users do not get the larger window, that would mea n Claude has ChatGPT beat.
Is the GPT plus subscription cheaper now?
No, the cost reductions are all tied to the API, not ChatGPT
@@aiadvantage Sounds to me they are favoring API over us ChatGPT Plus users then.
Is “data analyst” even a career path?!
What about Grok? Will you be doing videos on that?
Once I have access to it yes. Love the idea!
welp my midterm worries as a law student has disappeared
exciting
Very
Sorry, didn't the tests indicate that gpt4 had noticeably decreased its performance?
If you have some type of unique data now is your time to give it to Open AI 😉
Why does he say at 3:13 that we're gonna be able to generate images with text in them?
Because Dall-E 3 can do that already
@@aiadvantage Yea, but very POOR results... I could never get it to be accurate with text.
🔥
Im 19 and in computer science in uni. Am I doomed already
fantastic update 7.11.23
120 k token . The best ever
Yeah just matching the competition wasn't enough
Is there anyone who can already use custom GPTs?
Dobrý den 😊
Dobry :)
What mean one token?
For text to speech, do you know if it supports Arabic?
I will not take a deep breath
they clearly "addressed" safety concerns by adding www in and out to it. facepalm
POE did this a long time ago.
Some of it yes and even though I covered it multiple times most people don't bother with anything outside of ChatGPT.
No more free GPT3.5?
Still free inside ChatGPT
01.03 bro, please use a pop filter with your mic.
Also, this vocals are so bad; there is too much bass and muddy. Please eq your vocals before uploading.
So they are eating their creators ideas, that will get them crushed in court, if they aren't paying for the ideas
How? Unless it's backed by copyright or patent, they're out of luck.
Yea sure😂
Lol delusion is a heavy drug ….. openai is backed by Microsoft, and if there’s one man that secretly controls america , u know it’s Bill gates
Yeah don't think so. Most of them built on the OpenAI API and they are re-building all of this from the ground up and not just stealing.. Niche applications will prevail but the big and obvious ones are just going to be added as a feature. Microsoft is a master at copying competitors ideas and bundling them. OpenAI is just copying a proven corporate playbook.
In that same conference, OpenAI announced Copyright Shield to protect customers and users. OpenAI themselves will defend customers and pay the costs incurred if you face any legal claims around copyright.
Right now that only applies to ChatGPT Enterprise and the developer platform, but they're making some big moves to protect against the very thing you say will 'crush' them.
THIS was a perfect, hi-speed summary of all these new developments shared today!
your LIVE today was incredible too! THANK YOU
one question re: the new update:
Is gpt4 trainable now like 3.5 was --
can we "fine tune" the model?
or is that irrelevant now with the individual GPTs being created?
PS: your new product release = Freakin Mind-Blowing!!!! THANK YOU!
Appreciate it :) Yes they announced GPT-4 fine tuning but it's only coming to enterprise users at first from what I have heard so we will have to wait for this. It would be super relevant for the GPT's as you could use fine tuned models as a base which is superior then just using RAG (combining knowledge documents with chat outputs) which is what they introduced now.
So as per usual they are taking it step by step. First giving us individual agents "assistants" and the ability to add external data to custom charbots "GPT's" and then over time they will add multi-agent and gpt-4 fine tuning.
PS: Fantastic emoji use right there