241 What OpenAI’s New GPT-4o Means for Accountants

Mom, new model just dropped from OpenAI. Is it a big deal? I don't know. We'll run through the aspects of it that matter for accountants and how. I actually think the bigger thing is it's a really important stepping stone to super impactful stuff. Like the really big stuff that will change how we work probably in the next 12 months. We're unpacking all that today in a, in a entirely not nerdy way that's hopefully understandable even if you're just not a total AI dweeb like me. Kay, come on in. Let's get up to speed on what's, what's new. So OpenAI had a live stream conference today, the day before Google does their big event, which is kind of interesting and it was exciting. It's funny, back in the day, like Apple events for example, you know, announcing a new iPhone, stuff like that, man, those events were fire. Like they were a big deal. It was so exciting. But over time they just kind of been like man, they've gotten less and less exciting. Like if I go back to my childhood, there were like video game conferences where like they would announce really cool new things and it was these really exciting moments where you never knew what was going to be around the corner. And it was before the days when everything got leaked. And maybe that's part of the problem now is anytime you've got an event, it's often spoiled by leaks and you know exactly what you're going to get before the event really happens. So there's like nothing that's really a cool surprise anymore. And most of those events just aren't exciting anymore. They're not that interesting. Like the new showing off a new iPhone every year. Like eh, kind of interesting, but in no way what it used to be, where Steve Jobs would take the stage and be like man, what wild stuff are they going to announce here? And honestly the only version of that that I think is left is live OpenAI events. Like they have launched the most impactful software definitely of the last decade, maybe of all time. And so anytime they're doing a live event, it's exciting, it's interesting because you never quite know what they're going to drop. They did say ahead of time it wasn't going to be GPT5, but we still got an entirely new model which was like a pleasant surprise. So it's called GPT4O and the O stands for Omni. Omni meaning like multimodal is the technical framing of it, but basically that it works with images and audio and video. It can even like generate audio files. Does some wild stuff that we don't really get into much today and isn't particularly helpful for the ways that we use it. But I'll run you through just the bits that I think are most relevant to us and the things that we do. One I shared in the past a AI assistant called PI PI. And the novelty of that assistant. When was that? That was maybe last summer, last fall. The novelty of that assistant was you could just, you had on your smartphone, you could just talk with it back and forth and it was seamless and it felt like an actual conversation with a thing. And the model behind it wasn' all that good. Like it wasn't that powerful. It was just the fact that you could have a real time conversation with it where you're like, oh, dang. Like, this is different. This is very different than having to go to chat GPT and hammer a message into a chat box. There's actually a lot of friction and having to describe over text through your keyboard what exactly it is that you need as opposed to just talking. And if you haven't done that yet, if you haven't had a really compelling, like, voice conversation with an AI yet, I encourage you to give this a try. The new Voice stuff in ChatGPT is rolling out over the next few weeks. Right now, if you open up ChatGPT, you probably have access to the new model, but not the new speech stuff, like the new voice interface yet where you talk with it. But the first time you do this, it sort of unlocks something in your brain where you're like, oh man, this is different and like better in many ways because it's the same way that you interact with humans. And it's almost like there's these pathways in your brain that already exist where that's just an easier way to kind of work through a problem. And it sort of fires up a different part of your brain having that conversation. And so the first really big update here, besides it being a new model, the first big update here is the speech stuff. The voice support, how you talk with it is wildly better than it was before. Before they were using like a combination of different technologies to take the things that you say and make it text and then get a text response and, and then create like an audio file for the text response and it gives you that audio file back. Now all of that is just being done natively by the AI model. It can hear, it can talk, all that stuff. A bunch of those steps have got removed and the latency or how quickly it does it is like near instant. So rather than this awkward pause, you say your thing, you wait. Awkward pause. It says it thing. Its thing. And you have to, like, hit a button if you want to interrupt. Now. Now you can just talk back and forth and literally interrupt it mid sentence and it'll stop and it'll hear what you're saying. Interesting. The whole interrupting thing throughout this whole. I'll put a link to The Live. The OpenAI event. It's about a half hour. Put a link to it in the show notes. As they were doing all these demos over and over again, they're interrupting it. And every time you do, like, when you hear it, it's like a little bit uncomfortable in the same way that talking over another person would be uncomfortable. And it's just funny how we think of these things as humans more than they really are. When it's like, no, it should be completely fine to talk over a robot. It's just your computer. So I'm going to play you a few specific excerpts from that session and then a couple clips from additional videos that they played that just show you, like, how unbelievably human and interactive this is now, which I think changes, like how we use it. This was the first thing they showed off that was like the tada. Like, holy cow, this is different. Sort of impressive moment in the. In the demo. Barrett here, he's been having trouble sleeping lately and I want you to tell him a bedtime story about robots and love. So he's talking to his phone, the ChatGPT app on his phone.

Bedtime story about robots and love. I got you covered. Gather round, Barrett. Once upon a time, in a world not too different from ours, there was a robot named Bite. Bite was a curious job.

But I want a little bit more emotion in your voice, a little bit more drama.

Got it? Let's add some drama. Once upon a time, in a world not too different from ours, there was a robot named.

No, no, Chachi. Pt. I really want maximal emotion, like maximal expressiveness, much more than you were doing before.

Understood. Let's amplify the drama. Once upon a time, in a world not too different from ours, there was a robot named Bite. Bite.

Can you do this in a robotic voice now?

Initiating dramatic robotic voice. Once upon a time, in a world not too different from ours.

Isn't that wild? So you're not changing settings of the voice assistant, you're just telling it. Sound this way, sound that way. There's another example where the guy says talk faster, and it just talks faster after that. So it's not like you're using this other machine to make a voice, which is what I do on this podcast. When I create a little AI character to do this silly thing, I'm sending text over to another tool that does its best to, you know, talk in a entertaining or clear way. In the case of ChatGPT, you're just telling it. Here is how I want you to talk. Here's an example of it being sarcastic.

Hey, Chad, are you there?

Hey.

Yep, I'm here.

What's up?

Hey, let's do. Let's have some fun. I'd like you to be super sarcastic. Everything you say from now on is just gonna be dripping in sarcasm. How does that sound?

Oh, that sounds just amazing. Being sarcastic all the time isn't exhausting or anything. I'm so excited for this. Nope, the sarcasm. Let's get this party started or whatever.

Astute sci fi listeners here will notice that it is a spitting image of Scarlett Johansson in the movie her, which honestly feels kind of unnecessary. Like, everybody's afraid of this AI future and how dystopian that all is. Why do we need to make it sound exactly like a character from a movie where a guy falls in love with his AI assistant? Like, is that what we're going for here? And almost takes some of the focus off of just how cool the tech is. If you haven't seen the movie. Her, it's just one word. Her. Highly recommend. Like we're there. Like, we are living that movie right now. This episode is sponsored in part by Cloud Accountant Staffing. Tell me how this resume sounds. Graduated with an undergrad in accounting in 2012. 14 years experience now the last five with US public accounting firms. Software experience, QBO, Xero, Netsuite, bless your heart, Bill.com, stripe, Shopify, Gusto, Justworks. Rippling full cycle accounting. Oversaw the monthly close and review. Onboarded new clients, coached and trained staff. Oversaw a small team of bookkeepers getting the stuff done. Not bad, huh? That is a real resume from Cloud Accountant Staffing. A super talented accountant in the Philippines that could join your team. Not this person specifically, because this ad read might be old, but to find the accountant of your dreams, I got to do get on the horn with Cloud account staffing. Talk about what your needs are. They've got recruiters on the ground in the Philippines. They'll find you somebody great. You get some resumes, you do some interviews. You. You wanna know the wild thing? You don't pay a dime until you decide to hire somebody. Literally. What do you have to lose? The only reason you're gonna end up cutting a check to Cloud Accountant Staffing is if you got somebody that you absolutely love. So why not? I shared before I hired a bunch of people in the Philippines. Completely changed how we approached staffing our onshore firm. Give it a try. Link to Cloud Accountant Staffing in the Show Notes this episode is sponsored in part by liveflow. The spring G2 little award badges that just came out. You know G2 like the software platform review platform? What the software review platform just going to read you the the Spring little award badges that lifelong leader. That's all it says. Just leader. Spring 2024 Next 1 Best Estimated ROI Best Usability Most Implementable Best Relationship and Momentum Leader. All those just from Spring 2024 live flow. It takes your QuickBooks data, syncs it out to Google Sheets or Excel, whichever flavor you prefer. And obviously the kids are loving it cause they're burning down G2. Big news lately is they now support Excel. Sweet score. Huzzah. And shortly before that they released Consolidations, the ability to roll up different company chart of accounts. Like whether they have matching chart of accounts or not. They give you a super straightforward way to consolidate those things. Man, when I was running a firm we had like 20 plus company consolidations that were so much bloody work to roll up. We actually stopped doing it monthly because it was just taking so much time and we said we're only gonna give you the consolidations quarterly cause we can't be bothered to do it. This makes all that way easier and because it will actually sync that data out like live in real time on an ongoing basis. You can even do some cool like near real time consolidated stuff. Pretty wild. Learn more about this one. The link in the show notes, couple other clips to give you an idea of what this this voice stuff can do. And then I'll show you some of the they announced a desktop app. I'll show you how that could impact how we work. This was an example of using two assistants talking with each other who then begin to sing a song back and forth. Yes, it can sing if you just tell it to sing.

Can you just sing a song about what just transpired?

Oh In a room with a stylish feel A person stood in sleek black and like you engage with us and.

Can you just can you alternate lines?

Person with style and speed Person with style and sleep with a playful moment right in the mist in the room.

In a room where modern light speak.

Only one line at a time and then stop.

In a room where modern lights peak.

A surprise guest with a playful streak.

That was not really singing. Do do, do the singing voice again please.

In a room where modern lights peak.

Surprise guests with a playful streak Bunny.

Ears and laughter A swift retreat, a.

Moment of joy in the stylish sweet.

Says two phones sitting next to each other. Phones are listening to each other having a conversation back and forth. Which makes you think how many of these conversations may already be happening somewhere on a server where you've just got AIs talking back and forth. Interestingly, from a technology standpoint, this is how some of the AI tools that we use work is. You will actually have two AIs that will work through a problem together and those two AIs may have access to different information, not unlike org charts within a company. Right, so you've got what the reviewer would say, you've got what the preparer would say. And these different perspectives actually interacting with each other can like, like tackle some cool sort of novel problems. Last thing I'm going to show you here. OpenAI shared some bloopers of the model screwing stuff up as they were shooting these videos. And the reality is it's still not going to be perfect. But this blooper just kind of blew my mind.

Let's root, root, root for the home.

What was that?

Sorry guys, I got carried away. Guilt. What can I say? Sometimes I just can't help myself. Ready for another round?

So it's in the middle of singing a song and then it devolves into like garbled stuff and then it starts speaking French and the people start talking and they interrupt it and they say oh no. And it starts laughing with them and says, sorry, I devolved into speech speaking French there. It happens sometimes. And just the natural way that it like joined the laughter of all the other people and then apologize and then tried to jump right back in. It's like, it kind of makes you think that you can almost give it grace for not being perfect. Because the way it recovers from not being perfect is. I don't want to say human, but like endearing maybe is the word. And listen, the reality is none of us are perfect, right? We make dumb mistakes. I don't like have a stroke mid sentence and start speaking French. But that was just such an odd thing how the person starts laughing and the AI assistant stops and it's like, hahaha, yeah, no, that was my bad, sorry about that. And then it like goes back to what it was doing. So if somebody had told me they did a bunch of speech updates, I'd be like, meh, that's not that interesting to me. But to hear it and to see it, like, that's pretty darn wild. If you compare that to our smartphone assistants right now, which have never been less smart than they are today. Imagine the near future where everybody has an assistant at this level on their smartphone. And this is what I think. This is what I think makes AI much more mainstream than it is today. Is everybody having access to an assistant of this quality? I was saying this before the new model. If my mom had chatgpt on her iPhone and could interact with it just like she did Siri, that would be novel to her. She would be like, wow. I'm pretty sure there's ways that I can use this, especially if it works with your smartphone camera and you can show it things and what's wrong with this plant? Like, that sort of thing. But what makes all of this voice stuff even better is that it is the peanut butter to another jelly that they announced. That jelly being a desktop app. A desktop app that not only, you know, makes it easier to pull it up, and it's like a window that sits on top of all of your other windows, like, so it's just kind of always there, but it can see your screen. And if you've been around this podcast long, you know that this is what we've been talking about clear back to, I think, like the third episode of this pod, that AI assistants that can see the stuff that we're doing on our screen are going to be wildly helpful. And what this unlocks is the ability of the AI assistant to see and understand all the things that are happening on your screen, and you can talk with it with your voice while you work. How wild is that? So I'm going through, doing whatever I'm doing, and I've got chatgpt open. I hit a little shortcut, and I'm like, what is. Like, why is this not in balance? Why is this app doing this? How do I find or where do I find this option in this app? And while you're working, while you're using the mouse and the keyboard and doing all of that stuff, you don't even have to stop to get help from ChatGPT. You simply have to use your voice. And I think that this is the future of computing. We talk about how, you know, Bill Gates and folks like that are like, well, we're never going to have user interfaces for our computers anymore. We're just going to talk to them and that sort of thing. I think many applications, it sure seems like we will still have to have some sort of user interface and people who do work like we do. That seems like that will not go away entirely. But what I can get my head around is, is talking with an assistant as I'm doing my work and that thing helping me. Not unlike a person who could sit there like, you know, watching over your shoulder that can see what you're doing and be helpful in the context of work. This has long been my dream for what Windows Copilot ought to be. We now have Copilot built into every single Windows machine that is up to date. You've got this assistant and right now, unfortunately it sucks. It's basically Bing Chat but on your desktop. What I wanted to do is see the stuff on my desktop and help me with the things on my desktop and someday it could get there. It's frustrating how behind Microsoft is, so I don't know when it will get there. But this new desktop version of ChatGPT, that's it, man. That's what I've been wanting. I don't know when we're getting our hands on the desktop app. They said a number of these things will be coming out in the coming weeks. They also made a couple references to a Mac OS desktop app, which makes me a little afraid that the Windows version will come after the Mac version. I don't have a problem with that because I'm a Mac user now, But I know 95% of people listening to this are Windows users. And me just saying that I'm a Mac user just got me accountant canceled. But this is something they're in the process of rolling out. It's literally this always open window that you can have stay over the top of your other windows and you can chat with it while you do stuff and it can see the things on your screen. That's pretty cool. This episode is brought to you in part by Team up, helping you recruit top Filipino accountants without any ongoing monthly fees. The difference between Team up and all the other offshoring options is a team up helps you hire staff directly. No middleman. You work directly with your new hire in the Philippines. Hire the person, not the company. Guys, gals, gang. Here's just a few reasons to hire directly. You have access to higher level talent. Makes sense. You have complete control over team culture and training. They keep 100% of what you pay them and it's a lot more affordable for you. So you can retain your team for the long term, Team up can source accountants with experience working at US or Australian firms familiar with tools like Xerocubio and Dex. Also recruit specialist roles Team Leaders Leaders Tax specialists Administrative assistants Thought Experiment what if you had an executive assistant for the first time this tax season? Just throwing it out there. What would they do? Start at that email video I did on the main channel recently. Get help with that stanky ol inbox. I digress. Team up recruits these talented folks for a flat one time fee of 4,000 US American dollars. That's it. 4K one time. Somebody at Robert Half just did a spit take. Robert Half reference We ever going to get Robert Half to sponsor this pod? Not anymore. And they can connect you with an affordable employer of record if you need help with payroll and compliance once you hire that person. Big fan of hiring in the Philippines. You know I did a bunch of that. Check out the link in the description to learn more about Team Up. This episode is sponsored in part by Melio. If you have small business clients, and I know you do, you know just how hard it is for them to track all their payables and receivables, especially when they're doing it manually. Now there's an easy to use solution that turns all that tedium into an easy task that only takes a few minutes each week. Melio is a B2B payments platform that saves you and your clients many hours each month on bill pay by offering robust payment capabilities including advanced approval workflows and a two way sync with your accounting software. Also helps maximize cash flow by allowing you and your clients to pay vendors however they want to get paid, including instant transfers and per payment financing. They can also use Melio to pay international vendors. It can all be done in minutes and you can more easily keep track of everything on your end. On May 21st, set your calendars. Melio is hosting a webinar for accountants and bookkeepers where they'll cover some exciting updates to their platform that includes new features, product improvements and more. What could that be? What's in the more you think they'll also answer any questions you may have helping you set up your account to ensure a smooth onboarding process. I'll put a link in the Show Notes where you can sign up for that webinar and get to know this platform. That's a game changer. Spots are limited. Make sure you claim your seat asap. Thanks to me. Leo. Check out that link to the webinar down in the Show Notes. Now looking ahead, the Holy Grail of all AI applications is AI agents. And these are AI assistants that can actually click on things on your screen and, you know, make keyboard inputs and that sort of thing. And so you're hearing the vernacular AI agents a lot more. That's what agents are. They'll actually do the work. And it's a big old step from talking about the work and having to type into a chat box and all that to actually, hey, here's this thing. Watch me do this thing now do it every Monday at 8am or here's this thing. I've got this QuickBooks file. Open that task that I showed you last week, come in and do this and let me know when you're done. That is where our work starts to really, really change. And what they just announced this week is the necessary intermediate step to get there. And frankly, I don't. We'll see like the devil's in the details of how good what the model returns actually is. But this just. This version might actually be super impactful if it can help you troubleshoot on the fly and like work through problems that you're stuck on. I mean, especially for juniors. And this was. It's not that just juniors get stuck on this stuff. But you spend so much time, quote unquote, spinning your wheels and you don't know what the threshold is where you ask for help versus don't ask for help. But to have something that can see what you're doing, to just be a second set of eyes, that's super valuable. Also, how these things interplay with GPTs, which are like custom assistants that you can build any knowledge you want into. So, for example, the support documentation for Ultra tax or for QuickBooks or whatever it is, the notion that you can give it specific information and just have it use that information. Then there's a version of that assistant that you sit and work with every day that could be augmented with like general accounting knowledge, U.S. tax knowledge, Canadian tax knowledge, background on how to use the various apps that you use so that it has much more relevant information with which to help you when you do get stuck on things. That's going to be really exciting. So I can't wait to get my hands on the desktop stuff. Few other odds and ends that they announced this new model, GPT4. Oh, they're being very careful not to say, you know, this is not GPT5. This is not a model that's a huge step more capable, but in benchmarks, it actually is quite a bit better than any other model that's out there. Right now. So they're not really saying this, but in like, there's these kind of evaluation frameworks where people will blindly test the output of one model against the other and then pick which one they like better. This new version is much better than anything else out there. They're also giving it away for free. So now anyone, or very soon anyone, will be able to rock up to chat.OpenAI.com and use the very best model, which is great. In the past, if you weren't a paying user, when you go to chat.OpenAI.com, it gave you the worst model, the old model. And that's always kind of stunk because the folks who aren't drinking the Kool Aid yet and don't understand how it can be helpful to what they do, they'll go to the website just to try it. But then the experience they have is the very worst possible experience. It's with the worst model. Now, those folks, or at least very soon, those folks, when they go to the website, they're going to work with the very best model. It'll just come with some limitations, like they won't be able to message it that many times. But even on the free version, if you just go up and don't even create an account, there's an option where you can have it not remember that conversation. And it actually doesn't use any of your prompts to train the model or anything like that. Now, if you're an accountant, I still recommend if you're going to do anything with AI, the place to start is ChatGPT, specifically the team plan. It's 25 or 30 bucks per user per month, depending on if you're paying monthly or annual. And on that plan, nothing you ever put into it is ever trained into the models. Nothing it ever hears, sees or says is ever trained into the models. The original cause of alarm with these AI models was when you put something into it. They would then use that prompt and train it into the model. So somebody else that was talking with that model, you know, some response that they get could be impacted by a sensitive thing that you put in there. That's no longer the case. Hasn't been the case for quite a while. The team plan, like, that's the way to go. A couple other details on this new GPT4O model, it's faster. The usage limits are on average going to be 2x what they were before. So right now, if you use it a ton, you can sometimes be limited. And you say, like, you have to Wait another hour to be able to use it more, or something like that only really comes up if you're using it a huge amount. Those usage limits generally have been doubled. It is half the cost of the previous GPT4, which came out in November. So less than six months since their previous state of the art model. And with that version of GPT4 actually came a pretty big price cut too. I think it was either half or a third of the very best model before that, which at that point was only six months old. So over the course of the last 12 months, we've had now two price cuts for their very best models that now make the very best model between 10 and 20% the cost of what the best model was 12 months ago and way, way faster. Where else in technology are you seeing this speed of progress? I push people like, even if, even if you're not finding wildly useful ways to use AI right now, there's nothing else out there that's developing as quickly as all of this. So even our developers who are building AI into the software that we use overnight, the model both got better in terms of, like, what it could do and how reliable it was. It got faster by at least 2x and it got cheaper. It costs half as much as it did yesterday. What other software is that the case? Like, somebody rolls out a new feature, now it's a new tier that you're going to pay more for, or maybe it's built into what you have, but it's not getting twice as good and costing half as much overnight. Really exciting stuff. I mean, again, the most relevant thing for us in the work that we do is this desktop app. If you're on a team plan where you have like the data controls, where you're not going to be as concerned about the stuff that it can see because it's not getting trained into the model, it's going to be really exciting to see how people use that, the sort of things it's good at, the sort of stuff it's bad at, because it won't be amazing at everything, but then also how we can augment them with knowledge that is specific to the ways that we work. So how can I get it to know the apps that I use, general accounting knowledge, maybe tax information, that sort of thing. How can I get a version of it into the hands of the rest of my team that has all that knowledge built in to make my team more productive? Oh my gosh. Like, like, this could be really cool. Supposed to happen in the next few weeks. Um, as Always. Once we do, we'll be back with more information that'll actually probably merit a main channel YouTube video when we have the desktop app to kind of explore what it can do, what it can't do, how to use it both for yourself and for your team. Again, not just with AI, but with anything tech. The only people that ultimately get replaced by this stuff are the people who are surprised by it. Like, the only risk is hiding under a rock. If you're aware of what's happening, you're going to be as excited as anybody else to lean into it and be like, oh, dang, man, this is like going to make me more productive. I can help more people, I can do more meaningful work for the, for the folks that I already have, the clients I already have. It's only problematic when you're surprised by it. And it's why I always try to be really careful to keep the ear to the ground when there's these big announcements that get a lot of hype online to cut through it and be like, what aspects of this are relevant to what we do and what aren't, and how could it practically change the way that you get work done each day? So exciting update for sure, not only for us, but for our developers, like the folks that make software for us as well. This is awesome for them. If you gotten to play with this yet, we'd love to hear what your experience has been. And we'll be back with another update when we get that, that desktop version in our hands, because that's going to be. That's going to be really exciting. Could be an opportunity there to develop kind of some shared resources around that too. When it comes to, for example, a custom GPT that knows accounting stuff, knows tax rules software, that sort of thing. I'll definitely be looking at that when it comes out to see if there's a shared version of that that we could build that folks can swipe to make that thing even better than the out of the box assistant that doesn't specialize in the type of work that we do. Right, that's it for today. Exciting week. I'll see you back here tomorrow.