hey guys Le Motley here and today I've got a video breaking down the GPT 40 release uh the reason for the spring update from open AI there's some good things and there's some bad things for the AI agency space I wanted to jump on here and give my thoughts now that I've had a few days to Stew it and really think about what the implications are for us so GPT 40 the good and the bad for AI agencies of course what's new if you've been living under a rock we've had the open AI spring update recently and they released a bunch of things but mainly uh the new flagship model GPT 40 um o meaning Omni and this is a step towards much more natural human and computer interaction according to open aai on their blog post so that's taking in inputs of text audio image and video and it's able to output and text audio and image so massive new capabilities that we've kind of been cobbling together using different services and we've had whisper and we've had these other services that allow us to get this kind of functionality anyway but they've just pulled it all into one and allowed the model to understand not only text input but also audio and video and stream it all into one input as well so very exciting stuff um I was hoping for GPD 4. 5 or gpg 5 um but as we'll see later in this video there's some ramifications of what this incremental Improvement might mean for this bace um but I'll save that for a little later so there's also other stuff of the new chat GPT desktop app which is cool I've just been able to get it downloaded I think I need to update with Mac as many of you guys will as well I think it's only available for Mac uh for the time being um and GPT 40 is now available to free users um not just plus subscribers and therefore the GPT store is open to over a 100 million users so if you guys are interested in uh jumping on gpts I know I've made a video before that I think a lot of you found my Channel Through the GPT store is now available to 100 million users or more um which is great great news for you if you're hoping to be a GPT developer um we have yet to see anything about the monetization of gpts so I'd be curious to see what they have in store for that cuz that's kind of they attracted us all they honey potted us into this gpts uh and building gpts on the store thinking it's going to be an app store moment saying yeah monetization this monetization that um and we haven't seen seen anything of at least I have haven't seen anything about about the gbt store monetization so other stuff I don't think it's that relevant for us as as AI agency owners okay I want to start off with the good points first then we'll get on to the bad a little later um here's Sam Alman counting up all the all the money he's making from these recent updates um firstly new modalities it is a good advancement we are getting a ton of new solution opportunities opening up for us as we're able to take in different types of inputs from our end users and then give them different types of outputs back and really there's there's it's not a massive Leap Forward cuz we've been able to do this one of the examples on screen that openai has provided shows them asking a question and then providing an audio file as part of the input and then it's able to answer questions and reason off the back of that as well but prior to that it's not we were able to do that with transcription anyway we just transcribe it and then put it into the give it to the model to to reason over so not a massive leap we've been able to do a lot of these things but really what it is is just a simplification of our workflow and of the systems we need to build for our clients um less fiddling around with multiple different apis which is easier to get the results that we want for our clients and I think this is great for many of you who are not so technically inclined and I know a lot of you have been brought into this opportunity and still struggle with some of the technical parts of it but it's a clear Trend that we're seeing towards this simplification but there's still a level of complexity of how can this actually be implemented into the business so you're getting easier to do um but you still you're getting more power essentially in your hands that you can provide to your clients as well and because we're going to be using fewer apis this is probably going to decrease our cost as well because we're not having to use transcription and then generate an answer and then use text to speech if we're using these kind of systems which we'll get on to next which I think is pretty exciting The Voice AI systems and providers I think are going to win big here um because once audio inputs and outputs become available via the GPT 40 API the response times can be reduced by up to 60% um based off the numbers that open AI is provided which is between 200 and 300 milliseconds for responses at least that's what we saw of the chat GPT in the demo and as you can see here on platforms like vapy um even on the fastest and lowest intelligence model uh we've got a 650 millisecond response time and this is purely because they're having to stack up so many models that when your voice comes in over the phone they have to transcribe that then they have to generate an answer and text and then they need to turn that text into speech and then they need to send it off to you as well so this 650 millisecond latency which was fine it was fast enough we're now going to get a potentially 60% reduction of that as well so I think soon as these guys are able to access vapy and and Bland Etc are able to access GPT 40 via API and send and receive audio inputs and outputs um we could see a a continued boom of the voice AI space which is something I've been talking about a lot on the channel here if you guys are just starting with your AI agency you know looking for a good place to start or specialize in then voice AI is a great place to look into next we have a quick win for us as AI agency owners which is GPT 40 apis being twice as fast and 50% cheaper than GPT 4 Turbo it's always great on these big updates from open AI cuz we can kind of expect these reductions um and it's good to see that they're continuing to do this over time so we can expect it in future and an interesting thing to point out is that we're getting much closer to this GPT 3. 5 Turbo cost which is is basically free this thing is so cheap it's it barely cost you a dime to do anything um but here we can see that we've got input of $5 and 50 for gbt 3.
5 turbo so it's just a 10x price difference considering the the massive increase in intelligence and and modalities that we're going to get from GPT 40 you can't not be happy with that outcome next we have another quick win for us as AI agency owners which might have slipped under the radar for you a little bit better language support for GPT 40 that can handle over 50 different languages now covering 97% of the spoken world and it's also going to decrease a token usage as you can see here the new token compression method is actually reducing the amount of tokens for some of these languages as you can see now this may not seem like a major but this is a question I get all the time my accelerator and on my free community q&as which is should I be selling local or should I be trying to sell in the US or should I be trying to sell in Europe it's mainly people interested in selling in the US um and my answer is always no ideally Go Local um if you're if you're from South America and you're trying to go over to the United States and start selling there you're at a natural disadvantage just by purely being outside of the country you might sound a little different over the phone um you might have a name that doesn't necessarily ring like you're you could be someone's neighbor um and there's nothing wrong with that but it's just the the cold heart facts of it's going to be harder for you you're playing at some kind of disadvantage or debuff versus someone who is is John Smith who lives next door you know if you're in for example the Spanish speaking world I'm sure you've already had Fairly good responses and and good translation capabilities from GPT uh but it's really the smaller areas and these smaller languages that up until now haven't really had the support now you can be the first person into those markets so if you live somewhere that you thought oh no one's ever going to be able use this in my language or I shouldn't bother selling local now is your chance to be the first guy or the first girl in that market to go and start selling these Solutions and you might say that oh but they don't they're not interested in AI don't try to sell it as AI then just sell it as a meaningful difference in their business and now getting into the bad and we may have the rise of e girlfriends sooner or later um but that's not what I want to go into here um it's actually the long road to integration that's the first thing that I'm kind of concerned about here um and and by that what I mean is these new modalities and text and audio and video and image and all this stuff is cool but it's it it doesn't mean anything to us as AI agency owners until we're able to get that to our in customer with all these platforms that we use like make. com and voiceflow um and sending things to WhatsApp and the different solutions we build they are lagging far behind the technology that open AI is providing it really is an issue of trying to get the stuff in the hands and making it useful for our end users um but until these platforms catch up and and allow support for the customers concerned voice notes and they can send photos and and say voice flow allows you to send photos through your web chat widget which I'm not sure why and more so for things like WhatsApp deployments for your your AI agents um being able to send voice notes to the customers and receive it from them and send pictures and get them back um that's I think a long way off and I'm looking forward to seeing how they allow us to build these different modalities into our systems that we sell even for my own platform agentive we now have this question of do we want to integrate audio and video and image and all these different things into our application and into our platform or do we want to just stick with text base and I think this is a conversation many of these platforms are going to be having um it's interesting to see how they play out and moving on from that to something closely related is the lagging consumer Behavior now we can have technology that moves ahead very fast and and early adopters kind of catch up if you've seen my uh my technology adoption curve video which I'll put up here somewhere while technology can race ahead the actual tastes and preferences and and behaviors of the the consumer populace take a lot longer to adjust and e-commerce is an example of this where it took a long time for people to become comfortable with putting the credit card online and now we do it like the thought of putting a credit card and giving it to some some random website was ludicrous back in if you go back far enough it was a completely silly idea and over time it took like decades for them to get to the point where it's okay yeah now now we all buy stuff online this is the same sort of thing with with AI and I think we're going to run into this GPT and and chat GPT may help with people getting used to speaking to these AI assistants and having conversations um but I think there's still a considerable lag in the actual consumer behaviors where if we're trying to sell these Solutions do our end customers actually want to be sending voice notes to WhatsApp do they want to be sending pictures and and giving videos to them and personally if historical preceden are anything to go off I'm not betting on this thing moving too fast next we have more of a technical one that I think could be an issue which is the image and video difficulties that come along with Building Systems around much more complex and varied inputs like text image and audio um in this example here you might have watched my prompt engineering video where I I highlighted the difference between conversational prompting and single shot prompting I'll put the video up there if you haven't watched it highly recommend it's very important for you to know how to do prompt engineering and it's not your regular video it's very very different we had some really good feedback on that there conversational prompting and then there single shot prompting for us as AI agencies in many cases we're working in this single shot range where we need to engineer The Prompt and engineer the system to be reliable and predictable and continue to give the same outputs over and over and over again um so that they can actually be built into a company and operate as a a artificial intelligence task that plugged into their systems and doing that with text only proves to be difficult enough as I'm sure some of you have found out but now we introduce a whole another layer of complexity of images and videos so imagine this example here of of an email classification system where there's a user and they fill out a contact form we get an email then we use prompt the GPT task on something like make.
