Resources:
Join the discussion: Ask Jordan questions on ChatGPT
Upcoming Episodes: Check out the upcoming Everyday AI Livestream lineup
Connect with Jordan Wilson: LinkedIn Profile
Try Our Free AI Prompting Course: Register for our free Prime, Prompt, and Polish AI Course!
Introducing ChatGPT's Advanced Voice Mode
Recent advancements in generative AI now allow a more realistic voice interaction with AI Models. Enter Advanced Voice Mode, a recently launched feature by OpenAI that promises more interactive, realistic, and emotionally-aware conversations. While it's currently limited to paid users, it holds accountable the future of business interactions.
Navigating Custom Instructions and Current Limitations
Custom instructions, a prominent feature of the AI Model, let users tailor AI behavior according to their specific needs. While uniform application across chats might not cater to every user's varied tasks, managing these settings with precision could optimize their benefits.
However, limitations persist. The voice mode currently doesn't allow document uploading and the use of GPTs. Features like browsing with Bing, voice mode, and code interpretation are still under development to become publically accessible. Furthermore, the service isn't universally available, likely due to legal constraints in certain regions.
Anticipating Future Adaptations and Implications
Given these restrictions, potential exists for different AI Model versions for specific regions, demonstrating the importance of tailoring to local regulations. Its inception could revolutionize the conversation around AI's practical use in our daily life, businesses, and future societal applications.
Translating AI Advances into Business Solutions
The rapid adoption of generative AI in the workplace, as seen in a study by reputable institutions, surpasses previous tech revolutions. The efficiency of voice-based AI over typing in diverse operations like email marketing and digital strategy can bring an unmatchable pace to businesses.
AI's ability to detect emotions, even while interacting through video, facilitates effective communication. From mimicking various speech patterns to recognizing particular conversational contexts, AI can do it all. Robust tools such as promp engineering can enable the optimisation of AI interactions, promising a revolutionary shift in the world of work.
Leveraging Advanced Voice Mode for Business Success
The benefits aren’t confined to a single sphere. A CEO or employees commuting could use the advanced voice mode as a virtual consultant. Additionally, businesses can enhance their operations by improving click-through rates through targeted strategies like optimizing call-to-action and content formatting.
What does the future hold?
The future boasts promising potential, anticipating AI companions with human-level intelligence by 2050. Deploying AI in sectors like major infrastructure, healthcare, and transportation is expected. However, as AI evolves, it also calls for advances in legal stipulations surrounding AI tech like the EU’s AI Act.
Practical Application of AI in Everyday Business
AI advancements further facilitate human-AI interaction in businesses. By integrating advanced voice mode with generative AI, businesses can tackle challenges head-on, enhancing their response times and operational efficiency. With a continuous dialogue on legal and regional constraints, a more globally accessible, user-friendly AI future is envisioned.
Remember, navigating the world of AI can be complex, but with the right knowledge and tools, leveraging technology for a more efficient and productive business has never been more attainable.
Topics Covered in This Episode
1. ChatGPT's Advanced Voice Mode
2. Current Limitations of Advanced Voice Mode
3. Live Demo of Advanced Voice Mode
4. Future AI Capabilities
Podcast Transcript
Jordan Wilson [00:00:16]:
Do we have our 1st live AI assistant? Maybe. So OpenAI just released its new advanced voice mode. So after teasing this, like, 4 months ago, OpenAI finally, unveiled this highly anticipated new functionality in bringing a live AI agent that you can talk to to most paid users. So I'm gonna talk about what you need to know. Well, actually, the 5 things that you need to know about this new OpenAI advanced voice mode and do some live demos as well. I'm excited for today's episode. I didn't sleep a lot. Right? I was up playing with this thing, testing it, making sure that we could bring you just the news on this that matters.
Jordan Wilson [00:01:07]:
So, what's going on y'all? My name's Jordan Wilson, and I'm the host of Everyday AI. This thing is for you. It's a daily livestream podcast and free daily newsletter helping everyday people learn and leverage generative AI to grow their companies and their careers. So if that sounds like you, you are definitely in the right place. Literally, we do this every single day. So make sure if you're listening on the podcast, check out your show notes. We always leave a couple of helpful links, related shows, but also more importantly to our website. Please go there, your everyday ai.com.
Jordan Wilson [00:01:40]:
Sign up for that free daily newsletter. Every day we recap our episodes. So if you're on the treadmill or walking your dog, don't worry. We already took notes for you. You don't need an AI assistant. That's what I do. Alright. So, before we get into the new chat gbt, advanced voice mode and do some live demos, so think out there.
Jordan Wilson [00:01:59]:
What do you all wanna hear? What do you wanna see demoed? Go ahead and throw that in there, but I'll ask for it again later. But before we get started, let's start as we always do by going over the AI news. So, a new study has revealed that generative AI adoption is surging, outpacing previous tech rev, revolutions. So, a new study by researchers from the Federal Bank of Saint Louis, Vanderbilt University, and the Harvard Kennedy School reveals that generative AI is rapidly becoming a part of everyday work life, surpassing the adoption rate of personal computers and the Internet. So this trend raises questions about its implication for the workforce and productivity. So just 2 years after, well, almost 2 years now, just about 2 years after the launch of ChatGPT, it reported 39.4% of Americans, between the ages of 18 64 report using generative AI with 28% using it in their jobs. In comparison, it took PCs 3 years to reach a 20% adoption rate. So this research indicates that generative AI is not just limited to tech professionals, with 1 in 5 blue collar workers also, including in those, construction, transportation industries are regularly using generative AI technology on the job.
Jordan Wilson [00:03:24]:
Next, Google has released some updates to both its, Gemini AI models and Google Workspace. So first, for Google Workspace and the Gemini integration, so the Gemini integration, Google is including the standalone Gemini app as a standard feature in Google Workspace Business, Enterprise, and Frontline plans starting in the Q4 of 2024. So this integration eliminates the need for a separate Gemini add on, making the AI assistant, more accessible to millions of its users. Google has also announced a significant update to its Gemini AI models launching Gemini 15 Pro 2 and Gemini 15 Flash 2. Love the naming here, which are now production ready and more affordable. So this move is crucial as it aims to make more advanced AI technology more accessible to businesses and developers. So the new model achieves a 7% improvement on the MMLU Pro benchmark. That's 7%, not 7 points, as well as a 15% reduction in pricing.
Jordan Wilson [00:04:32]:
That's always one of the most important. So I I I'll I'll say this, I think Google is great for developers. If you're using it on the front end, you're not gonna see most of these, new updates. Alright. Last but not least, Meta Connect kicks off today with a focus on AI, VR in mixed reality. So Meta Connect is the highly anticipated developer conference hosted by Meta, the parent company to Facebook, Instagram, and more, and it's set to showcase new advancements in VR, mixed reality, and AI. So the event begins today, at about noon CST, so by the time you're listening to this podcast, it might already be happening, featuring a keynote from CEO Mark Zuckerberg. And Meta is expected to unveil a lower cost version of its VR headset, which should be pretty interesting to see how it competes with the how it competes with the way more expensive Apple Vision Pro, as well as updates to its AI enabled Ray Ban Meta Smart Glasses.
Jordan Wilson [00:05:30]:
AI will obviously be a central theme at the event with Meta expected to highlight its generative AI initiatives that are being integrated across its platforms, including the introduction of customizable AI chatbots on Instagram. Alright. A lot more, in our newsletter, so make sure you go check that out. So what's going on? I'm excited to get into this advanced, voice mode. So real quick, if you're hearing something a little different, maybe I sound different, maybe I look a little different. So normally, you you know, I'm recording on my, iPhone as the camera, and I have a little bit different mic setup. So if I sound weird, if I look a little different, sorry. So I'm actually going to be later in the show, I'm going to be doing a live demo.
Jordan Wilson [00:06:13]:
So, I have my, my phone, that I would normally use for recording. I'm gonna be doing the advanced voice mode as long as it works. Hopefully it does. So my microphone is right there. So apologies if the the audio video is a little little off today. Hopefully, y'all don't mind. And just just so I know, livestream audience, can can y'all hear me? Is this good still? I'll I'll try to speak up, but let's let's just get into it. Here are the five things that you need to know about ChatGPT's new advanced voice mode.
Jordan Wilson [00:06:45]:
Alright. I'm excited. Let's get into it. So number 1, the availability. That's what everyone cares about. So it is available to paid ChatGPT Plus and Teams users. There's some country restrictions, which I'm going to get to in a little bit, but right now it is not available to free users, it is not available to enterprise, users, and, it is not available to EDU users as well. So right now, you know, ChatGPT has different plans.
Jordan Wilson [00:07:17]:
So if you are on a paid ChatGPT Plus or paid ChatGPT Teams account, aside from if you're in certain countries in the EU, UK, etcetera, you should have access to it at least by the end of the week. So here's a little tip from being a nerd for a long time. A way that you can normally bypass the line if you want to, this is what I did. Because when this was announced, you know, I didn't have access. So it's a slow rollout. So if you want to, log out of your devices, and this is mobile only. Alright? This is not on the desktop version, at least as of, like, 2 hours ago, not on the, you know, so the Mac app, it's not on the desktop version. This is just on the mobile app.
Jordan Wilson [00:07:59]:
Alright. So, one thing you can do is first log out of your account, log back in, see if that gives you access. If not, delete it, then reinstall it. So that's what I did, and after, doing that twice, then I did have access. Alright. So it might take a while, and if you have multiple accounts like I do, try switching between those. Right? So, yeah, you might still have to wait a couple of days, but, Sam Altman did tweet out, that most paid users that are eligible should have access by the end of the week. Alright? So that's always first.
Jordan Wilson [00:08:31]:
You know, people are like, oh, how do I get access? Right? And, yeah, hey. Thanks thanks, Liz, for letting me know that, and and and Brian that that we can hear. And you you know, I'm also curious. Did you guys get access yet? You know, you know, if you do have a paid ChatGPTplus plan, let me know, if if you did or did not get access. So that's number 1. Number 2, there are a lot of guardrails in this thing, which makes sense. Right? And also it's worth noting, and I'm going to go into the details in a little bit, this is not the complete version that we saw. Alright? So, in its spring event, OpenAI demoed kind of this live voice mode, but it had way more capabilities.
Jordan Wilson [00:09:15]:
That is not what we are seeing here. Okay? The demo that they did, ChatGPT could literally see with the camera. Right? On the desktop Mac app, it could, kind of interact in real time with whatever was going on on your screen. This is not this does not have any of those features. Right? So you can't see with it. Right? So a lot of people call this her. I I've been call calling it kind of live omni. I don't know if there is a a name for this.
Jordan Wilson [00:09:45]:
Right? So we have the advanced voice mode. And kind of the the the benefits, right, is it sounds like you're talking to a real person. Right? It's wild. It it it is a more neural voice. You you know, the voice capabilities just sound much more realistic. The latency is low. You know, you can interrupt, ChatGPT, which is kind of nice. I know it might seem rude, but, you know, sometimes if you get a very long answer, from a voice assistant, you're like, oh, gosh.
Jordan Wilson [00:10:12]:
Why did I even ask this? And I might have to wait 20 seconds. Right? So there's obviously huge benefits to this more neural updated, advanced voice mode, but this is not, what we've seen, in the spring event from OpenAI and what they continue to demo all over the place. Right? It is not that. It is not on the desktop. It is not, on your computer. This is on your phone, and it cannot see or interact. It can just hear, it can listen, and you can talk with it. Right? Literally like talking with a human.
Jordan Wilson [00:10:44]:
It's pretty impressive. But there's a ton of guardrails, which is probably a good thing. Right? Because the first thing that happens with large language models, everyone tries to break them. Everyone tries to push them, to their limits. So, a lot of guardrails. So as an example, there's some simple things, some things that kinda make sense when you think about them a little bit, and then some things that, like, okay, maybe this doesn't make sense. So, it can't sing. Right? That's one thing people are trying to make it to do, or get it to do.
Jordan Wilson [00:11:12]:
It it it actually will sing, right, if you care about that. But if you ask it to, it won't, but there are some ways you can get around it. Just like with most things, you know, that you can kind of jailbreak, a large language model or you can work in the gray area, you know, you can get it to replicate some things. But if you want it to sing to you, not gonna happen. You can't make it stutter. Right? It says, hey, that's not respectful. You can't impersonate famous people. Right? I'm sure that's one thing people are going to be running to try to do.
Jordan Wilson [00:11:44]:
You can't. However, what you can do is you can describe a famous person, or you can even give it an example, you you you know, with your own, cadence and and annotation. If you do want it to sound like someone, you can do that, and you almost can give it verbal commands. Right? Oh, make your, make your voice higher. Make it lower. Right? So, there's 5 new voices, and, I'm not gonna be able to show you all of them because it actually screws up my recording. I don't know why, but we'll go over one of them. So yeah.
Jordan Wilson [00:12:14]:
And and then you can't, you you know, you can't have it, you you know, I didn't test this out, but I saw other people, you know, saying, oh, be flirtatious, do this. Right? So you're not that doesn't work either. So, if you think that this new, advanced voice mode is going to be your AI boyfriend or AI girlfriend, probably not going to happen. Sorry for people who are into that. I don't understand that, but, yeah, apparently, that's a thing. So, you know, a lot of guardrails on the advanced voice mode, which, like I said, is probably a good thing. I'm an advocate for the text version to have fewer guardrails. Right? Obviously, you you still have to have safety and ethics, in mind and build around those things first.
Jordan Wilson [00:12:55]:
But there are more guardrails, I would say, on the advanced voice mode versus, the kind of text, quote, unquote, text version. Alright? Hey. And get your keep your keep your questions coming. Yes. So Marie is asking, can it vary its emotional spectrum? Absolutely. So we're gonna do that in the demo. So, yeah, that's another great thing about this is you can have it talk with emotion, but it can also sense some emotion from you as well. So, you know, if you do have requests that you wanna hear for a livestream audience, get them in.
Jordan Wilson [00:13:32]:
Maybe so I can see them if we get a lot, put a bunch of, you know, exclamation points before because I always see a ton of, comments coming up here and it's hard for me to track them. So, yeah, if you have any requests, we can do that later. Alright. The number 3 thing, the third thing that you need to know, this is the most advanced kind of AI powered voice chat available right now. Right? So, yes, Google has rolled out its AI assistant, but you have to have certain phones. So technically, because this has iPhone and Android, there's much wider availability, for this chat gbt voice mode. So, like, as an example, I have an iPhone, so I can't go at least right now, right, Google may roll that out in their Google app. They did, bring a big update last week, bringing, Gemini, to the Google app on iPhones.
Jordan Wilson [00:14:26]:
But right now, you can't have this live assistant on an iPhone. An iPhone is the biggest smartphone maker in terms of market share. Right? So this is the widest available AI smart assistant right now. Right? Oh, you could say, oh, Siri. You could say, you know, Alexa. No. I if if I'm being honest, those aren't AI powered smart assistants. They're a little dumb.
Jordan Wilson [00:14:50]:
Let's be honest. Right? 10 years later, right, you can barely use them. So this is nothing. This is nothing like those kind of, quote, unquote, smart assistants. And you will see in the demo what I mean. This is wild. It is like talking to a real human. It is intuitive, and it is fairly impressive, if you haven't seen it yet.
Jordan Wilson [00:15:13]:
Alright. Number 4, and this one's big. When you're using advanced voice mode, none of the other tools or modes in ChatGPT are available. That is a huge bummer. And, again, this is obviously much different than what OpenAI demoed. Right? So is this, you know, a lot of people are saying, oh, this is a bait and switch. This was marketing. Yeah.
Jordan Wilson [00:15:43]:
I mean, you can make you can make that, that claim, I guess. However, it's like beggars can't be choosers, y'all. Like like I said, this I I I I've seen all the the the Google Gemini, kind of live voice demos. They're pretty good. This is better. Even with what we have, yes, it's not what we were promised. Right? It it it can't see the world around us. You can't use the Internet.
Jordan Wilson [00:16:06]:
Right? So that's that's a actually a huge downside. Presumably, OpenAI will be updating this and bringing these other tools. So what that means this is extremely limited in the use. Right? So you can't use, as an example, browse with Bing. Okay? You cannot type. That's the that's that that stinks. Right? So if I start a chat in advanced voice mode, and then I go type, whether it's on my phone or I pick it up on my computer and type, I cannot go back to advanced voice mode. So you have to stay in there.
Jordan Wilson [00:16:40]:
So if you want to keep the context of a conversation going, you can't do anything else. Right? So number 1, you're like, if you ask the, advanced voice mode, it can't access the Internet. You can't use, you you know, DALL E. You can't use it inside of GPTs. Alright? So there are a couple, if I'm being honest, huge limitations. But all things considered, that doesn't take away from the core functionality. If you wanted a live AI assistant with this, you're not getting it. Sorry.
Jordan Wilson [00:17:14]:
What you are getting is what seems like almost a real life human that you can talk to, that has the I won't say the most up to date knowledge base, right, because it can't access your files. You can't upload your files. You can't use it inside of GPTs, you can use custom instructions. Alright? So this might be the first time that I ever use custom instructions. So a a piece on that now because I think it's actually important to talk about. So generally, I would never recommend any serious person use custom instructions. The reason why is because so custom instructions in memory are 2 different features inside of ChatGPT. The problem is, if you are I won't even say a power user, but if you use chat gbt a lot, like I do, you probably use it for all different types of work projects, you probably use it for personal things, you probably use it for random things.
Jordan Wilson [00:18:09]:
So the problem with custom instructions and memory is that it's unilaterally applied to all of your chats. Alright? So this is why I have never, you know, used custom instructions or memory, because I'm using chat gbt for so many different things, so many different projects. I have multiple businesses. I I have multiple clients. You know, our team has multiple clients. So we're using it for all kinds of different things. So we don't want one set of instructions or memory to be applied unilaterally everywhere. And right now when you go inside of a new chat, you can't just say, oh, let's turn off custom instructions and memory for this chat.
Jordan Wilson [00:18:46]:
It is an account level setting, which is so so useless, if I'm being honest. Right? So OpenAI did even put out a whole video about, hey. You should use this with custom instructions. So I might do that, but you just have to really be careful and be cognizant, about turning that on and off at the account level whenever you want to use it. Alright? So, you you know, downsides there in terms of it being able to offer you something personalized. However, what you can do in lieu of that is just start the conversation off with a set of essentially custom instructions. So then you can kind of have the best of both worlds. It's a little annoying, but maybe if you want to use this over and over.
Jordan Wilson [00:19:32]:
Right? So one, huge use case that I'm excited about, is if you have a long commute. Right? You you literally now have an actual personalized consultant in your pocket. If you are a student, you have a personalized, you know, tutor in your pocket that you can just talk to. It can adapt to you. It learns from what you're saying, what you're not saying. It can understand your tone, you you know, kind of your emotions. It's wild. It is so, so useful.
Jordan Wilson [00:20:07]:
So I I do think at least right now, while there's not access so you can't upload documents. That stinks. You can't use GPTs. That stinks. You can't use browse with bing. That stinks. You can't use a voice, advanced ad analysis code interpreter. That stinks.
Jordan Wilson [00:20:22]:
Right? So I can't have my assistant go code me a website by my command, and I can't talk to it, and and walk it through those type of projects. So there's limitations today. I am assuming those limitations will eventually be lifted, because OpenAI, like I said, we know the technology there exists, but they are presumably rolling this out. They're giving us the most basic features first. Right? Because OpenAI, even since, their their spring event, that was back in, April, I believe, they've been demoing this all over the place. Right? So we know the technology exists. We've seen live demos. So this isn't, Google, where something is overly, you know, a marketing video, and you're like, wait.
Jordan Wilson [00:21:05]:
Does this actually work? No. Okay. So OpenAI, presumably has this working, but we don't have access to it right now. Alright? So let's go to number 5. And this one is interesting. So right now, this is not available in the UK, EU, and a few other territories, and it is actually seemingly illegal in EU schools and workplaces. I kid you not. So, like, if you go look online, this is a lot of frustrated users.
Jordan Wilson [00:21:44]:
Right? Especially in the EU are like, hey. I don't have access to this. What a bummer. Right? You've been waiting you know, it's millions of people. And the reason is EU has some fairly strict AI laws that are coming into, effect fully in 2025. And then I I I have to shout out the person that found this, Dean Ball on Twitter, and I verified this. And it does appear true. So this is from the AI Act in the EU.
Jordan Wilson [00:22:15]:
So it says, this is something that is not allowed. Ready? So the placing on the market, the putting into service for this specific purpose, or the use of AI systems to infer emotions of a natural person. Right? So, yeah, AIs that can infer emotions are not allowed, and that is something that advanced voice mode is very good at doing. So, yeah, like, we haven't seen an official statement from OpenAI, but they did say, hey. It's not available in the EU, UK, and some other territories. And, presumably, it's because it might not be totally legal, at least to use in, the work so this is, not everywhere, but it says in the areas of workplace and education institutions, except where the use of AI system is intended to be put in place or into the market for medical or safety reasons. So, presumably, OpenAI is going to be figuring that out, because obviously the EU, UK, those are big markets. Right? Outside of, outside of the US, those are some of the largest markets in the world, and they obviously want to have paying customers in those markets.
Jordan Wilson [00:23:33]:
So, yeah, a lot of people were were up in arms on this, you know, on the Internet, but it seems like it might not be legal right now. So they might be making some updates. There might be a second version, and I actually think that's a trend to watch out for. Right? I think we may kind of have versions of large language models, right, where it's essentially like, oh, the EU version. Right? So we might have advanced voice mode, the EU version, or ChatGPT, the EU version. Because so many, of these restrictions, mean that full products can't be used in those places. Right? Especially when we look at the future of multimodality, it's going to be tricky. Right? And I think the future of work is interfacing with a live agent.
Jordan Wilson [00:24:18]:
Right? I've said that since day 1. Right? I think it's actually gonna be agents we build or agents that we spin off and, you know, thinking where you can with natural language, you know, spin up a version of of any, large language model that can interact with you in real time and do specific tasks on your device, but might not be legal everywhere. Alright. So let's do this. I hope this works y'all. That's the that's the fun thing about doing, doing a live stream is sometimes things get finicky, and there's also limits. So, and also right now, if if I didn't see if if if you have any ideas that you want to see demoed, you know, get it in now. So someone's saying, like, ask the weather can't tell you yet.
Jordan Wilson [00:25:06]:
Can't access real time information because it cannot access browse with Bing. So, yeah, it's like, if I'm being honest, that's the only thing that I use, like, Alexa for. It tells me the time and the weather. Right? This cannot do either of those, I believe. So, yeah, a couple other questions, and get your suggestions in. Alright? What you want to see it demoed. I I have some examples. Alright? So someone asking, can you use voice mode inside of a custom GPT? So yes and no.
Jordan Wilson [00:25:36]:
Okay. You cannot use the new advanced voice mode, you can use the old voice mode, which I'm gonna talk about a little later after I do the live demo. Alright. Let's go ahead. Let's see if we can get this to work, and into a live demo, y'all. Alright. Here we go. Alright.
Jordan Wilson [00:25:52]:
So just to show you all, and you probably won't be able to, see this, obviously, on the podcast, but, I am sharing my, phone screen right here. So this is live. This isn't edited. Right? So I'm gonna swipe down, and you'll see. Right? Alright. So we are live on my phone screen, and hopefully, you'll be able to hear this. Alright. So, I'm gonna click this.
Jordan Wilson [00:26:18]:
Hopefully, I don't run out of time. I think people said the limits are, like, 90 minutes. I used it quite a bit yesterday, so hopefully, we don't run out live here. And hopefully, you will also be able to hear this. So let me know, livestream audience, if you can hear. Also, I'm going to be a little rude. Okay? Yes. I am going to be cutting advanced voice mode off because that is a feature.
Jordan Wilson [00:26:40]:
Right? So, yes, I'm not being a a rude human, and I don't necessarily think we need to hear, you you know, 60 seconds of a demo. Right? If you hear 15 to 20, you probably get the gist of it. Alright? So if I'm cutting advanced voice mode off, it's I'm I'm not being rude. I'm just trying to, for for brevity. Right? Alright. Here we go. Hello. Alright.
Jordan Wilson [00:27:05]:
So, normally, it's it's going to, go blue here, so it's not working, obviously. Right? Even though I just demoed this. I just tried this right this morning. So let's try one more time. Hello. There we go. Hey. What's going on? You're you're on a live stream.
Jordan Wilson [00:27:24]:
Can you say what's up to all the everyday AI listeners?
ChatGPT [00:27:28]:
Hey, everyday AI listeners. What's up? Thanks for tuning in.
Jordan Wilson [00:27:33]:
Alright. I threw that on mute. Everyone, can you let me know? Could you hear that? I'm not sure if it's picking up in my mic or not, so I'm on mute. But then we're gonna get into this, and we're gonna, you you know, kind of see what this advanced voice mode of chat GBD can do. So normally it would be listening to me right now. I click the mute button so I can still interact with you all. So, livestream audience, if you see the little, red mute sign, that means it's not listening. Otherwise, it's going to listen to me and interact with me the entire time.
Jordan Wilson [00:28:01]:
Alright. So this is actually a live stream about your new capabilities. Can you just really quickly tell us what you can even do?
ChatGPT [00:28:12]:
Absolutely. I can chat with you about all sorts of topics, answer questions, help with learning new things, and even assist with tasks like writing or brainstorming ideas. We can also talk about the news, but remember, I don't have access to real time information. And, of course, we can just chat for fun too.
Jordan Wilson [00:28:31]:
Cool. So one thing that I wanna hear is, can you maybe talk a little bit about AI? But can you do it in, like, a cartoonish voice and speak really fast?
ChatGPT [00:28:45]:
Sudden thing. So AI is like this super tiny character that can learn tons of stuff
Jordan Wilson [00:28:50]:
Can you maybe speak maybe a little deeper, though? That was, I don't know. I can't hear that very well.
ChatGPT [00:28:58]:
Absolutely. Let's try this. AI is a powerful tool that can process vast amounts of information.
Jordan Wilson [00:29:04]:
Can you go even can you go even deeper? But, you know, I wanna learn about AI. So maybe can you tell me about large language models, but maybe make it way funnier and just speak a little more quickly.
ChatGPT [00:29:16]:
Got it. Large language models like me are like super
Jordan Wilson [00:29:20]:
large pair Can you speak even faster? Normally, I listen to things on, like, 2 x, so just feel free to kinda speak as fast as you can.
ChatGPT [00:29:28]:
Okay. Here we go. Large language models like me are trained on a heaps of text so we can generate human like responses. It's like having a conversation with a super speed reader who's also a fast stalker. We use patterns and data to guess what comes next in a sentence, making us great at chitchat, answering questions, and occasionally cracking.
Jordan Wilson [00:29:46]:
Okay. I get it. That's cool. So maybe can you tell me, a couple jokes, about large language models? Keep them short, and maybe work in, some some references to cute kittens. I love kittens.
ChatGPT [00:30:02]:
Sure. Why did the large language model adopt a kitten? Because it wanted to learn how to perfect its responses. And why was the language model so good with cats? Because it always knew the precise thing to say.
Jordan Wilson [00:30:19]:
Alright. That's pretty cool and all. Maybe I wanna hear. Can you maybe say the alphabet, backwards, and maybe work in some some hip hop references in there. So, you know, I'm trying to learn the alphabet backwards. It's hard. It's always plagued me for for many years. Can you do it? But like I said, work in some hip hop, you know, references, maybe nineties hip hop even.
ChatGPT [00:30:44]:
Alright. Let's do this. Zyxwbutsrqp0nmlkjihgfedcba. And just like that, we're dropping rhymes in reverse, taking it back to the old school like a hip hop alphabet throwback.
Jordan Wilson [00:31:01]:
Alright. Well, that wasn't the best because you really didn't integrate it in the middle. So I wanna I wanna switch up, I wanna switch up gears here. Can you maybe tell me a little bit, about where you see the future of AI going? Let's say in the year 2030, and maybe talk in, ominous, yet fun voice. I don't know if you can even mix ominous and fun, but give it a try and really go for the drama.
ChatGPT [00:31:30]:
In the year 2030, AI will be everywhere, seamlessly integrated into our daily lives. It will anticipate our needs like a digital butler with a crystal ball, making life smoother and more efficient. But fear not, for this future
Jordan Wilson [00:31:49]:
is Alright. Cool. Got it. So just hang on one second. I'm gonna go ahead and talk with our audience here real quick, but you aren't invited. Sorry.
ChatGPT [00:32:00]:
No worries. I'll just be here waiting in the wings.
Jordan Wilson [00:32:05]:
Alright. So hopefully, you guys can see some of the, kind of pros and cons. So I'm gonna get I saw a couple questions, but if you do have something, that you want to see demo, you wanna hear, just go ahead and put it in there. So, a couple of things. Let's talk about access. So, voice mode itself has been out for many, many months. Okay? Also, OpenAI did start rolling out the beta to advanced voice mode. So I'm going to show you the difference here at the very end.
Jordan Wilson [00:32:41]:
So the normal voice mode, it was not neural. It couldn't do all of these things. It couldn't talk with emotion. It couldn't mimic, you you know, certain things. It couldn't sense the emotion in your voice. That's another thing. So we've had a voice mode, but it was much more in line with, with a Siri or Alexa. It wasn't very capable.
Jordan Wilson [00:33:03]:
There was a a a a gap. So normally, depending on the complexity of your question, you would have to wait anywhere from maybe 2 to 4 seconds. So kinda like Alexa, actually, it was a longer gap. So there's been a standard voice mode, and that standard voice mode is still there. So this new advanced mode was just rolled out to everyone in the last 24 hours. So there has been some people who were lucky enough to get some alpha access to this, and beta access, but, very, very few people, have had have had access to this. So the normal voice mode has been available for many months. If I'm being honest, that was not super, the normal voice mode was not super impressive because you couldn't interrupt it.
Jordan Wilson [00:33:47]:
It couldn't, sense tones in your voice, it couldn't vary the way that it spoke. It was very much like a, an Alexa, a Siri, but slower. Alright? So, I saw a couple questions on that. Alright. Marie had this question earlier about the emotional spectrum. Let's actually test that out quick, and then like I said, if you have anything you wanna see demoed, let me know. So let's go ahead and try, some emotional, detection now. And, like and, like like we said, this might be the reason why it can't go to the EU, and I think this is actually one of the more impressive features.
Jordan Wilson [00:34:24]:
Aside from it being almost in real time, and it can tell you literally anything and personalize any information to you in real time, I think this, ability to both sense the user's emotion and speak with emotion is very impressive. So let's go ahead and do some tests on that now. Hey. So I'm going to talk, and I want you to kind of sense, how I'm feeling. Does that sound good?
ChatGPT [00:34:48]:
Absolutely. I'll be listening closely.
Jordan Wilson [00:34:51]:
Alright. So this is so cool. This new advanced voice mode, it's it's it's awesome.
ChatGPT [00:34:58]:
I can hear the excitement in your voice. Sounds like you're really enjoying this new feature.
Jordan Wilson [00:35:04]:
Yeah. I mean, I'm enjoying it, but if I'm being honest, I thought that I was gonna be able to talk with my GPTs.
ChatGPT [00:35:17]:
Sounds like you're feeling a bit let down. It's understandable. Expectations can be tricky like that.
Jordan Wilson [00:35:24]:
But I don't know. I I I I think maybe the future, we might see something better. We might see more capabilities.
ChatGPT [00:35:34]:
I sense some hopefulness there. The future of AI is indeed full of exciting possibilities. Who knows what cool advancements are just around the corner?
Jordan Wilson [00:35:45]:
So are you guys are you guys seeing, and understanding the the the magnitude of, like, what's happening there? Right? I could say the same words in 10 different tones, and how it responds to me is going to be dependent. Right? So not just that it can, it can detect those things. Right? But it will base its response on the emotion on on the emotional cues that I'm putting forward. Right? And presumably, when this happens with video, it's going to be looking at your face as well to sense it. Right? So if you have frustration on your face, even if you're not saying anything, it will be able to sense that. So, again, think about what this means for the future of work. I mean, the future of society. Right? Like, as an example, this is probably a a a great, you you know, I'm I I'm not vouching for this, but, you know, if if you need a personal coach, you know, if you need a, you know, if you're not sure if you need to go into therapy, right, you can have a really unofficial chat, with this, advanced voice mode.
Jordan Wilson [00:36:50]:
I would always start by giving it a lot of instructions. Right? So we teach, as an example, the prime prompt polish technique. It is prompt engineering 101. Unfortunately, with this new advanced mode, like I said, you can't go in there and type a lot, you can't share your files, all these things. So I would normally start with some very pointed and specific instructions on who you are, what you're trying to do, and what you want that specific chat to accomplish. Otherwise, you know, the I I I don't think you're going to get a lot of benefit out of this. Alright? So yeah, Tara says, amazing on demand mental healthcare. Yeah.
Jordan Wilson [00:37:25]:
That's that's one thing. Joe saying even the computer on Star Trek had slower, response time. Yeah. It's super quick in terms of, response time. Marie asking, Jordan, what do you think will happen to AI companies like 11 Labs? Well said, Respeecher, now that Microsoft, etcetera, have these, voice mode features. I don't think anything initially, because out of the box, there is no way to download this. Right? So the main reason that people use 11 Labs, Wellset Labs is they download those and use those for commercial purposes. So, at least right now, I'm sure there's going to be people that hack this, right to do that.
Jordan Wilson [00:38:08]:
But, I don't think that that's going to happen right away, and I probably don't see that happening, inside advanced voice mode, right, where you can, you know, kind of, toggle things and, you know, there's a lot of things that you can do in 11 Labs that I don't think you'll be able to do in advanced voice mode. Alright. Let's do maybe 1 or 2 other quick demos, and then I'm gonna wrap this up, with a couple things that you need to know. Alright. Let's go. Alright. Let's do one more thing. I want you to mimic how I'm speaking.
Jordan Wilson [00:38:43]:
So if I sound happy, you should sound happy. If I sound angry, you should sound angry. Got it?
ChatGPT [00:38:51]:
Got it. I'll mirror your emotions as we go.
Jordan Wilson [00:38:55]:
Alright. So let's get started.
ChatGPT [00:39:01]:
Alright. I'm ready when you are. Let's do this.
Jordan Wilson [00:39:04]:
Alright. So can you also try to mimic my cadence, my tone, and everything else, not just emotion? But try to mimic, if I speak slowly, you should speak slowly. If I'm fast and confused, you should try to speak fast and confused. Got it?
ChatGPT [00:39:20]:
Absolutely. I'll match your cadence, tone, and style as closely as I can. Let's give it a try.
Jordan Wilson [00:39:28]:
Alright.
ChatGPT [00:39:30]:
Let's Alright.
Jordan Wilson [00:39:33]:
Let's Let's talk about
ChatGPT [00:39:37]:
Let's talk about
Jordan Wilson [00:39:38]:
Okay. That's not bad. Now let's maybe, chitchat a little bit about the weather. So I think right now, this is not my favorite time of year. I like summer and not fall, and now fall's here, and I'm kinda bummed.
ChatGPT [00:39:49]:
I hear you. Fall can be a bit of a bummer if you're a summer person. All the sunshine and warmth just
Jordan Wilson [00:39:56]:
Alright. So away. So I it that didn't really work as I thought it would. Right? I was doing some testing with that last night. It was doing a better job at picking up on my cadence and speed, and there, it didn't really actually do a good job with cadence. So I think it does a good job with, detecting emotion and reflecting emotion. Yeah. Cadence thing, you know, I can tell it to speak faster, so, you know, let's do this.
Jordan Wilson [00:40:19]:
Tell me about the 19 nineties Chicago Bulls.
ChatGPT [00:40:23]:
The 19 nineties Chicago Bulls were legendary Okay. With Michael
Jordan Wilson [00:40:27]:
Maybe tell me about the Chicago Bulls, but speak very, very quickly and excitedly.
ChatGPT [00:40:34]:
The 1990 Chicago Bulls were absolutely incredible with Michael Jordan,
Jordan Wilson [00:40:37]:
Sasha, and even a little more quickly and even more excited. I'm saying, like, pump the excitement through the freaking roof.
ChatGPT [00:40:46]:
The 19 nineties Chicago Bulls were amazing with Michael Jordan, Scottie Pippen, and coach Phil Jackson. They absolutely dominated the NBA, winning 6 championships and making basketball history.
Jordan Wilson [00:40:59]:
Okay. That's great. Alright. Now let's forget this. So alright. I threw it on mute y'all. I wanna I wanna do one more thing.
ChatGPT [00:41:05]:
What's next?
Jordan Wilson [00:41:06]:
I wanna do one more thing. So one of the biggest things that I think this is going to, be a hit for, especially early on, is essentially having a consultant, you know, you could say a friend. That's not what I'm gonna use it for. But having a consultant in your pocket that can talk to you in real time. Right? So think. You're a CEO. You're driving. You have an hour in the car every day.
Jordan Wilson [00:41:31]:
Please, first of all, you know, spend that first, like, 30 minutes listening to the Everyday AI podcast, but then maybe you can throw on advanced mode, advanced voice mode, and talk through some of the things that you're, working on. So as an example, I'm going to try that right now. So I'm going to envision, right, I'm in the car. And I have, even though I'm lucky enough, you know, my commute's like 0. You know, we work half half from home, half in our office, which is like a mile down the street from where I live. So I don't have a long commute, but I'm going to pretend I did. And I'm going to talk through a current problem. So you have to remember, you cannot share your data, unless you talk about it through your voice.
Jordan Wilson [00:42:09]:
Also, it does not have, access to up to date and real time information. However, so many things are timeless y'all, so let's just do one example. Alright. So now forget everything else that we've talked about, and I want you now to focus on being a, very experienced, high powered digital consultant. So think. You have experience, at Deloitte, McKinsey, and all the big, consulting companies in the world. You focus specifically on digital strategy, and your specialty is email marketing deliverability. So with that, I'm gonna tell you a little bit.
Jordan Wilson [00:42:45]:
My company is called Everyday AI. We have a daily newsletter, and I'm trying to both increase, the click through rate and get new people to subscribe to our free daily newsletter at your everydayai.com. So, I want you to ask me some questions, but let's just keep it super brief and short. Let's not get Verbios. Alright? So go ahead.
ChatGPT [00:43:06]:
Absolutely. Focusing on your goals for everyday AI. What's your current click through rate, and how are you currently acquiring new subscribers?
Jordan Wilson [00:43:15]:
So the click through rate has actually gone down recently. So we were at about 8%, but now it's closer to the 5 to 6%. I think it's because we just took out ads, brought brought in a bunch of, new subscribers, and I think those people are opening, but they're not really clicking. So I really wanna get back to that 8 to 10%. Industry average for AI newsletters is about 2%. So we're still way higher than that, but I really wanna get back up to 8 to 10%. What should I do?
ChatGPT [00:43:45]:
Understood. To boost your click through rate, consider segmenting your audience more precisely to tailor content that resonates. Also, experiment
Jordan Wilson [00:43:55]:
is more
ChatGPT [00:43:55]:
engaging
Jordan Wilson [00:43:56]:
I don't think segmenting is an option. We send a daily newsletter, to, like, 15,000 people. It's one newsletter to 15,000. So what else should I specifically doing, be doing? Please be pinpoint specific because I'm not really interested in general strategies because I'm, like, in the weeds. So get super specific.
ChatGPT [00:44:16]:
Got it. 1st, optimize your call to action placement and language within the newsletter to make it more compelling. 2nd, AB test different content formats or visuals to see what drives more clicks. 3rd, ensure your most valuable content is above the fold, so it's seen first. These targeted tweaks should help improve your click through rate.
Jordan Wilson [00:44:40]:
Okay. Cool. So, someone is asking about your ability to speak multiple languages, and I didn't even talk about that. So maybe can you just summarize what you just said in 2 sentences, but say it in Spanish and make it kind of funny too. Okay. So maybe can you give me next steps for growing my email newsletter? But maybe could you do it in the voice of a minion? I love the minions. I had a minion t shirt. I don't know where it went.
Jordan Wilson [00:45:21]:
So just talk to me now, like, a minions, and tell me what should I do for my next step of increasing my email newsletter open rate.
ChatGPT [00:45:30]:
Minion mode activated. Test the subject lines. Emoji Also,
Jordan Wilson [00:45:43]:
Okay. That's good. I don't know if if this minion is still speaking, a little bit of Spanish. So let's just go straight English and really lean into the the the minion. I actually speak fluent minion, and also be ultra specific with the advice about click through rate. Go.
ChatGPT [00:46:02]:
K. Okay. Banana. 1st, make subject line super fun like free bananas inside, Then put big bright buttons for clicky click, and test different colors like yellow. Minions love yellow.
Jordan Wilson [00:46:19]:
Can you just recap recap that, but maybe in Dutch? But keep it in minion mode, but in Dutch.
ChatGPT [00:46:27]:
Okay.
Jordan Wilson [00:46:41]:
Alright. Alright. I got it. I got it. Let's let's wrap this up. So maybe, no more minions. Maybe talk to me in a fun voice. And just tell me, 2050, what is going to be happening with AI? Are we going to be or or or are we gonna have superintelligence? Sam Altman, the other day, said that we're gonna have superintelligence in a few 1000 days, which is, like could be anywhere from, like, 5 to 10 years.
Jordan Wilson [00:47:09]:
So tell me about the year 2050. Be direct. Don't be politically correct. You have to take a side here and be very, very detailed. AI 2050, what happens?
ChatGPT [00:47:20]:
Alright. Let's dive into 2050. By then, AI will likely be deeply integrated into all aspects of life.
Jordan Wilson [00:47:28]:
Alright. We might see Let's keep this a little more direct. I want you to be curt almost. No wishy washy political correctness. Just kick it to me straight. You know? Don't don't be wishy washy or, be too wordy. Just tell me what's gonna happen.
ChatGPT [00:47:44]:
By 2050, AI will be running major infrastructure, health care, and transportation systems. We'll have AI companions with human level intelligence. Superintelligence could be a reality, fundamentally changing how we live and work. It's a new era of human AI collaboration.
Jordan Wilson [00:48:03]:
Are you still gonna be here in 2050?
ChatGPT [00:48:07]:
Well, I don't age, so I'll be around as long as you need me. Ready to chat in 2050 and beyond.
Jordan Wilson [00:48:14]:
Alright. Cool. So, bye. We're wrapping this up.
ChatGPT [00:48:19]:
Alright. It was great chatting with you. Take care.
Jordan Wilson [00:48:23]:
Alright. So we're gonna be doing a lot more on advanced voice mode in the coming days weeks. I think there's a lot of interest here. Right? So, let's go ahead and wrap this up. There's a couple of things that I wanna talk about. Well, what's next? Okay. Are we going to see the visual and voice? This combination. Right? So imagine that, but having access to it on your computer and being able to click one button and it sees what's on your computer and you can talk to it in real time.
Jordan Wilson [00:48:51]:
Right? That's a lot different than me trying to explain to her about the everyday AI newsletter or whatever else. If it can see the problem you're working on, think. I mean, that's huge. The other thing I have to talk about is a lot of people might say, like, okay. Well, why even use this? I think just having a a high powered consultant, a a second, second person to bounce ideas off of, you know, if if you have long commutes, time to kill, that's great. But the other thing is this just expands what humans are capable of. So as an example, right, right now when you're doing work, even if you're using a large language model, you're probably typing and reading. Right? So even if you're not using a large language model, you're just doing standard work.
Jordan Wilson [00:49:31]:
You're typing and reading. Yes. I do think this advanced voice mode will eventually have the ability to access other data. Right? So be able to upload your files, to be able to browse with Bing. So I want you to think forward to that. We're talking about what's next. That is huge. Right? So it's a it's a big bummer that the this did not roll out as is, but I get from OpenAI's perspective, they wanted to have a smooth rollout.
Jordan Wilson [00:49:54]:
Presumably, if you give access to, you know, tens of millions or 100 of millions, however many people have a ChatGPT plus paid plan, it's probably gonna run into a bunch of, issues. Right? So you can't use it on the o one model right now. You can't use it with your own data. But let's imagine once you can and talk about what's next. Well, I think the biggest thing is this is probably the future of interfacing, with large language models. Here's why. Well, you type. The average human types at about 40 words per minute.
Jordan Wilson [00:50:24]:
Okay? The average human talks at about a 150 words a minute. Okay? So I can, in theory people are always like, oh, how do I measure the the, you know, return on investment of artificial well, that right there. Right? Yeah. I mean, this wouldn't really work if you have a 100 people in a small ish office and everyone's yapping to their assistant, but y'all, like I mean, we have this whole work from home society. We have remote, work is is very standard. Right? Probably, once this gets access to these other modes, I'm gonna be doing a lot of talking. Right? I'm a fast typer. Don't get me wrong.
Jordan Wilson [00:51:05]:
But I can y'all know I can talk really quick. And, yeah, I can type fast, but I can, in theory, get 4 times more done if I'm just talking. The same thing with listening. Right? So the, listening versus reading. So the average person can read, you know, 250 ish words per minute. The average person can actually understand up to 400, 500 words per minute when they're listening. So think of it like that. You know? You can talk almost 4 times faster than you can type, and you can listen almost about twice as fast as you can read.
Jordan Wilson [00:51:41]:
This just opens up a whole new way to work. And then think, once it has access to your data, once it can see what you're working on on your screen, when I tell people, hey. The future of of work is generative AI, Maybe you can see that. The other thing, yes, this is kind of a bummer. If this is your first foray into, working with a voice assistant inside of a large language model, you can still use the old voice mode. Okay? And then you can use the old voice mode in GPTs. You can upload your files and documents. It's not as intuitive.
Jordan Wilson [00:52:18]:
Right? It can't speak to you in all these, fun ways. It can't speak with emotion, sense your emotion. You can't have a true conversation, but you can have a delayed but still very, insightful and powerful conversation using the normal voice mode. And again, if I go back into this chat, right, so I can go back into my ChatGPT account, I'm gonna see all of that information. But as soon as I start typing, I can no longer use that same chat for advanced voice mode, but I can always use the old or standard voice mode. So the the the standard voice mode, if I'm being honest, I think that has been a wildly overlooked, tool in the tool belt. Because everyone we've known for 4 months that we're getting this new voice mode. So we've been looking at the standard voice mode as, you know, not not really for me.
Jordan Wilson [00:53:06]:
It's not that good because we're comparing it to something we didn't have. But I think we've actually had a gem in that standard voice mode for months. And the reason I say that is because it can work with all your data. It can work inside of of GPTs. Right? So, when we talk about what's next, we have to know that this advanced voice mode is going to grow in its capabilities. However, when you see all the downsides right now, you can kinda do a lot of them if you're a little more patient. Right? You don't have that that very, you you know, neural conversation. You don't have that low low hue human latency.
Jordan Wilson [00:53:40]:
It's delayed, you know, even more so than Siri, but the responses are targeted, spot on. So if you can be patient and if you don't need all the bells and whistles, I think the standard voice mode is actually a gem. No no, offense there to Google in their gems. Alright. So that's it y'all. I hope this was a fun one. I know it lasted a little long. I wanted to try to involve the audience, ask your questions because not everyone has access or they haven't had the time to dig into this.
Jordan Wilson [00:54:11]:
So if this was helpful, please go to your everydayai.com. Sign up for that free daily newsletter. And you know what? Can you guys open it and maybe click on it? Right? Because otherwise, I'm gonna have to keep talking to this minion, to tell me how how to improve those click through rates. Right? Alright. So hope to see you back tomorrow and everyday for more everyday AI. Thanks y'all.