Watch This Before Using GPT-4o For Your Business

I want to see if the new chat GPD 40 is actually as good as open AI says it is so I'm putting it to the test today with five Super common use cases for me as an online business owner and I'm going to compare the results of those use cases from GPT 40 gp4 turbo and Claude 3 Opus I'm first going to test having the AI write an email then I'm going to test its data analytic skills then I'm going to test comprehension of a long PDF then its summarization skills of a call transcript and then finally I'm going to test its brainstorming capabilities and have it write me ad copy specifically meta ad copy so Facebook and Instagram ad copy I'll do each of these use cases in each of these models and then see which results I like best so the first use case of writing an email I've already written the prompt all the prompts I'm going to use today are the exact same ones that I will use in GPT 4 and also in Claud so whatever I use in gp4 I'm also using in CLA the only difference is going to be the slight difference in formatting of the section header so in gbt 4 I'm going to use markdown and in claw I'll use XML uh tags when appropriate for some of the prts I don't all right so on the the screen here this is the prompt I've already written to write an email this is actually an email series we're only going to do one email though and I've already written the entire prompt I have added in my target audience already I've added in my tone of voice my writing style so it writes in that format so I've already done all those things I'm going to copy the prompt and now let's go over to gbd4 okay so I have pasted that prompt right here into chat GPT I'm using the chat GPT 4 or excuse me the GPT 4 o model o stands for Omni by the way and let's hit go and put this prompt into action so the first thing that I notice about GPT 40 as it writes is super fast so all of the hype around the speed of GPT 40 I would say yes absolutely it's very fast so it is written me a fullon email and then given me five as instructed in the prompt give me f five subject line ideas so everything looks good I like the sections in the email and I did ask it to ask me at the end of the email that it's written for feedback and approval it will not move on to the next email to write the next email in this case here until I give it approval so it's saying how does this first email sound let me know if there any tweak or additional details you like to add and so I could just simply say approved and it would write the next email but for this example here I'm not going to I think this is pretty good I like the formatting skimmable which is how I how I write so now let's test this in GPT 4 Turbo and see what happens now open AI has put in this really handy model changer at the bottom and rather than starting a new chat or anything all I have to do is click this GPT 4 which is Turbo and it will automatically start executing that prompt and I'll have two different tabs here with the results so let's do that so as you can see here right off the bat it's slower okay so check out what GPT 4 Turbo started with it says great let's start with the first email which is all about delivering a lead magnet before I write the email could you tell me what the lead magnet is this will help you tailor the content and subject lines more effectively isn't that interesting because gbt 40 didn't do that it just went right into fill-in the blank of your lead magnet name I absolutely like the fact that it's asking me what the lead magnet is that it is delivering this email about right here in the prompt so let's give it the lead magnet so I've just now put a this is a fictional lead magnet I'm calling at the AI prompt generator and I just gave a simple oneline benefit driven statement about it so let's see what it does with this information so I think it's actually still pretty fast uh it's not quite as fast as GPT 40 so it's just finished generating this output here and again I like the fact that it asked me the question of what lead magnet so this email sequence prompt is for the delivery of a lead magnet so this first email is delivering a lead magnet and letting people know what's to what to expect so I like the fact that because it now knows that name of the and benefit of the lead magnet it's contextually relevant now to in in the email here so I like that aspect of it what I think GPT 40 did a better job of is simply the uh format of the email I think this is cleaner uh especially with the sections like what's next and then it gives a your action plan so I like that better again but I don't like the fact that it didn't ask me up front for the name of the lead magnet so that it could make this email more contextually relevant so both sides to it it's not a clear winner for me in terms of GPT 40 I would lean more towards GPT 4 Turbo for this specific use case now let's go over and test this in Claude 3 opus okay here we are in Claude you can see I've chosen the Opus model and it is the exact same prompt I just used in chat gbt the only thing I've done is changed the formatting a little little bit changing the markdown that I used for the section headers and instead put XML tags for the section headers and to end each section only change I've done so let's see what Claude gives us here all right so I love claude's response right off the bat it also asked for the name of the lead magnet because that's what the email is delivering that we're writing here and then also a brief description of the lead magnet saying once you provide these details I'll then generate this first email after receiving your feedback and approval I'll proceed to create the subsequent emails in the sequence following the outline structuring guidelines okay cool I love that response let's put the name and the same exact thing I put into GPT so now it's writing subject lines and the email itself I would say the speed is Ain to gp4 Turbo which I mean it's writing it pretty fast not quite as fast as GPT 4 but in this case here speed doesn't really matter a whole lot to me I want to see a really good response here okay I've just read through this by far it it sounds more human than GPT 4 o or Turbo it's kind of what I expected I find Claude writes a lot more humanlike and I like the subject lines I mean they're nothing ground breaking here I would probably put into the prompt the length of the subject line that I want it to generate and I like the email itself I would probably do a slightly different formatting but for the actual email writing itself and how it sounds I like Claude better so I'd say maybe for this use case of writing an email I would say Claude is the winner of this one because number one it asks me the questions of the lead magnet and a little information about it before writing the email GPT 40 did not do that and then secondly I think the writing of the actual email how it sounds is more like me and more humanlike so I'm going to give this one to Claude 3 Opus for use case number two I want to test the data analysis skills of these three models here my expectation is that GPT 40 should be superior between gbt 40 turbo and Cloud 3 Opus Cloud 3 Opus is not necessarily known for its data analysis but let's see what happens here so on the screen here you can see this is just hypothetical sales performance data and I've just taken a screenshot of this information so not only am I going to test the vision aspect of the models but I'm also going to test its data analysis so I've taken a screenshot of that data in the spreadsheet I've attached atted it here in chat gbt and I've written a prompt here it's not very long and I also haven't uh formatted The Prompt like I normally do but it is a very detailed and explicit prompt so I just want to see what it does with this so basically what I'm do what I'm doing is I'm asking it to analyze the data on that image for monthly sales growth customer retention website conversion rate average order value sales forecast customer segmentation and then once it gives me that data I then want it to provide some actionable suggestions for how to continue to scale the business based on this data I think this is a really good test let's do it so it's giving me the output pretty quickly but I will say the analysis of the data it took longer than I expected especially from this newest model in reviewing the results here it's pretty good I like how it broke it out gave me the monthly sales growth by percentage month over month customer attention rate in a percentage format website conversion rate trends average order value I wonder if I could do this in a table format so it's easier to do something with but that's okay uh it gave me the sales forecast for the predicted sales uh for the next quarter customer segmentation ratio of new to returning customers this is really good and then it gave me a bunch of recommendations for for sustainable growth for this hypothetical business based on all of that data so I like it I like this uh output quite a bit so with that now let's test uh gbt 4 Turbo okay and the results are in it took I think it actually was a slightly faster in the analysis of that data and I got to say I like this output a lot better because it broke it you know it just simply broke it out by the month like January to February gave the percentage increase and then it gave a little additional information that analyzes this data so the sales growth rate shows a robust upward Trend so it just sort of summarizes that data and it did that throughout all of the data analysis then down at the recommendations for sustainable growth it looks to be about the same as GPT 40 but I got to say I like this better I like the output better than GPT 40 it's it's I think it's more helpful so let's go test this in Claude 3 opus okay we're in Claude now and I have put in the exact same prompt I haven't made any changes to it even the formatting and I have again included a screenshot of the sales performance data so let's see what it gives us okay and here is the output I will say that Claude was faster than both GPT 40 or GPT T4 turbo and the results are more similar to what gbt 4 Turbo gave me in terms of the percentages I would say I probably like the gbd4 turbo results a little bit better there it gave the customer retention rates Broken Out by month and G again gave a little analysis here website conversion rates average order value sales forting in each one and then gave six different recommendations I would say the recommend ations are about the same across GPT and also clae this is what I find really interesting between these three results each of these models gave different forecasts for Q3 so Claud obviously broke this out by month and then gave a total for Q3 forecast in GPT 40 gave me a Q3 forecast of 35,000 it didn't really Break It Out by month or anything like that and then in Turbo it gave me a completely different uh this is closer to I would say at least closer to Claude but again it gave a completely different answer here to forecasting I like how it broke it out I would have preferred it give me a total so I would want to definitely fact check all of these results from any of these models with that said I would say I like claud's response better again it's telling me an average 177% monthly growth rate versus GPT telling me uh a conservative average growth rate of 10% for month now it's it's act it's being more conservative so you know all right cool I would say overall though I like a little bit of a combination of gbt 4 Turbo results and the Claude results I just like they're they're fairly similar but what I like that Claude did is it gave more of an analysis on potential improvements to the data in addition to the recommendations at the bottom I like the recommendations at the bottom from gbg4 Turbo better but either way I think gbt 40 did not perform as well on this data analysis now with that said we we need to fact check these numbers because each of these models gave different results in terms of forecasted numbers and so forth so we'd want to check those but yeah I liked the gbt 4 Turbo and Claud responses better for this data analysis use case for this third use case I want to test the comprehension skills of the AI models what I'm going to do here is I've written a prompt that analyzes an attached PDF it's a long PDF and I want it to give me a solid comprehension of the text and make recommendations to me based on its comprehension of that text and I've chosen for the text How to Win Friends and Influence People by Dale Carnegie it's a long PDF it's like 265 pages so it first has to analyze that long PDF and then I've asked as it to give me daily practice suggestions on how I can implement the principles from this book so here we are in GPT 40 I've already put the prompt in here again it's going to be the exact same prompt and let's see what this says the first thing I noticed with gbd 40 is it is super fast it is analyzing a 265 page document very quickly and let's review this complete output it broke down the different sections of the book fundamental techniques six ways to make people like you etc etc so basically all the principles from the book which is great but I asked it to create a daily practice schedule based on the principles I didn't necessarily ask it to give me the principles but hey that's cool that it did so it actually gave me a full day's routine and it broke it down by time blocks throughout the day so morning reflection reflect on things that you appreciate write down three things you're grateful for review one principle from the book each day and think of how you can apply it I don't love that because I asked it to give me a daily practice putting into play or putting into use the principles from the book so it's almost like I have to go back and reference the principles and I wanted to incorporate that into the daily plan start with positive feedback encourage team members to share their thoughts and ideas respond to customer emails or comments with genuine interest and appreciation that's kind of weird so I'm not super impressed with this and then it just said suggestions for tracking progress which I did ask it to give me which I like that I like those responses then it gave me some additional tips and resources thank you GPT 40 now let's see what gp4 turbo does with this use case so the first thing I noticed was it was definitely slower in its analysis but it still astounds me how fast it just analyzed a 265 page PDF now what it's doing here is it's following the same sort of thing that it did in with GPT 40 and it's giving me the different sections and Main principles and then giving me bullet points and and so forth then for the daily practice schedule it broke it down into morning midday and evening time blocks essentially it didn't give me specific times which is fine and let's see here I like what it did I think it's I feel it feels more specific to the actual principles of the book meaning like outline your key business interactions focusing on how you can make each interaction Pleasant think about specific interactions you anticipate for the day and how you can apply principles during business interactions focus on becoming genuinely interested in other people smiling and using their names those are direct things from the book so ra I like this better than GPT 4 oh I think this is more it's I think it's a better more actionable daily plan which is what I'm asking for in the prompt so winner for this one goes to GPT 4 Turbo let's see what cloud 3 Opus can do okay I've just pasted the exact same prompt into Cloud 3 you can see I'm using the Opus model again I have attached the PDF of the book and let's see what it gives us so the first thing I notice is by far Claude is the slowest in processing that book and analyzing the PDF however as after it's processed it you can see how fast it is giving the results here okay so in looking at the output from Claud here I very much like both GPT models in their principles and how they formatted the principles I like that a lot better than just you know 1 through 20 that Claude has just given me however the daily practice schedule which is what I'm really looking for from this prompt I like this better than either of the GPT models because it's more specific it breaks down client communication content creation afternoon Outreach marketing meeting evening reflection I like how specific it is rather than saying oh do this during this time block or something like that so I will say that for Claude I like the actual output from the practice schedule which is what I was really after however I like gbt 4 Turbo and gbt 40 better in terms of the principles a lot so again I would say this is a combined winner gbt 4 Turbo and Claude three Opus if I could combine the two I would like that the best in terms of an output but again I'm not impressed with GPT 4 in terms of how it handled this use case for this next use case I want to test the summarization skills of a transcript this is a transcript of a call I had with somebody on my team last week and I've written a prompt it's very simple so basically what I'm doing is I'm asking it to provide a summary of the transcript of the call and then also give me the main action items and deliverables from that call in addition to the summary the call is about an hour long so the transcript is fairly robust again let's start with gbt 40 and see what it gives us okay so again the speed that GPT 40 is doing this specific task and a couple of the other ones there is very fast I like how it is breaking down the main points the summary of the call and it's even doing it in little sections giving me a section header I likee all that and I really like the action will takeaways and deliverables who's responsible the deadline and the details I like everything that it did with this so this one is going to be hard to beat so I think this result here from gbd 40 is going to be pretty hard to beat but let's see what gp4 turbo does with this and here is the result from gp4 Turbo it GES gave me three bullet points at the top in terms of key points discussed some decisions made next steps and then actionable takeaways and deliverables which is pretty good let's again take a quick look at gbt 40 and I like this result a lot better so in terms of gbt I very much like gbt 40 better in terms of summarization for of this transcript and giving me actionable takeaways and deliverables let's go see what Claude can do same exact prompt just attach the transcription to the prompt and let's see what Claude comes up with and here are the results and I will say that it's good but by far I like GPT 40's uh output better it's just I it it just looks just crammed in here it's not organized super well it's not easy to read and flow like GPT 40 was so for me for the summarization GPT 40 by far is the winner for this use case okay for this last use case I wanted to test the brainstorming skills of each of these models and of course you can do this in any number of ways I wanted to because this is a common use case in my business I wanted to see if it could brainstorm some great meta ad copy so Facebook and Instagram ad copy I've written a prompt for that it's a really good prompt and again we're going to start in GPT 40 I've also already put in all the necessary information in terms of the offer target audience my brand voice all that stuff I've already put that into the prompt but we're going to use the exact same prompt again across all of the models so let's see what gbt 40 gives us so the first thing I do notice is that like the other use cases foro is really fast in how it's executing this task so what I've asked it to do in the prompt is to write four different variations of AD copy both long and short form ad copy and then for each ad copy variation I've asked it to give me five different hook ideas that I potentially could use in the first sentence of the add copy so check this out and I'm looking at it and it looks okay I mean I would I wouldn't just copy and paste this copy by any stretch um so yeah I mean it's it's okay it's decent I will say the one thing that it did not do is give me a any kind of longer form copy variations now I fully understand understand that AI still has a challenge with shorter and longer but I still would have expected it to at least especially GPT 40 since it's a new model to follow that a little bit more closely I will say also the one thing I didn't do in the prompt is I didn't Define what longer form ad copy looks like I just wanted to see what it would give me and see how well it followed the prompt all right let's test this in GPT 4 Turbo so far I am not impressed with the results that uh 40 gave me and here are the GPT 4 Turbo results I mean look at the difference here this is not very good number one right off the bat in the prompt I am very specific saying to make the copy skimmable short sentences more paragraphs GPT 40 did that turbo here did not do that absolutely did not do that also there's no longer form copy variations here as well notice too that it put the hook ideas for each ad copy in its own section I liked how gbt 40 actually included those with each with the respective ad copy variation so not impressed with four turbo at all on this the winner for this one in terms of GPT definitely goes to 40 but let's see what Claude can do with this all right in Claud I've just put the exact same prompt in here again and by the way in this prompt here I'm having it write ad copy for my aifi lab membership Community uh if you're interested in checking that out I'll put a link down below of the video here let's see what Claude gives us in terms of add copy variations and here is the output from Claude it went a little overboard with the Emojis but that can be cleaned up I don't specify emojis in the uh brand voice paragraph in the prompt but this sounds number one the formatting is much better with the bullet points I like it again it sounds more human in the copy so it's it's better it's also seems a little bit longer than anything that gbt gave again same thing I like the formatting of it I like the hook better and then also gave the hook ideas along with with the actual variation the only thing I don't love is that it kind of took the same format all throughout each of these variations here and I mean obviously it changed up the copy and the hooks but it didn't get creative at all in terms of the format of the actual ad copies so I would ask it to change that I would also ask it to use fewer emojis but in terms of how it's written Claude is by far are better in terms of how the copy sounds so this use case I'm going to give the wi to Claud simply because of the fact that the copy sounds more human and I can more easily adapt this prompt to get more to what I would be actually looking for and either or either model I would also give it a an example of longer form ad copy versus short order form ad copy so it can be more specific in the output there you have it comparing the three models GPT 40 gp4 turbo and Claude 3 Opus across five Super common online business use cases from my business as an online business owner I got to say I'm not super impressed with GPT 40's output I love its speed but I think between these use cases between GPT 4 Turbo and Claude I liked the responses a lot better except for the summarization and uh analyzing the transcript so I will give it to GPT 40 on that one but otherwise I'm not super impressed with it in terms of its outputs from its prompts with that said I am very excited about the voice aspects of GPT 40 having it be more and act more as like a personal assistant that you can talk to and have conversations with there is a motion now baked into GPT 40 and that's exciting so that'll be super fun to be able to play with and test in terms of prompt output not super impressed other than the speed so I do hope that that improves but yeah I think this is a perfect example here today in this video that you still need to test the other models to get the best results possible for whatever use case that you you are doing in your business all right my friend thank you so much if you like this video please give it a thumbs up also subscribe to the channel and if you have any questions comments feedback as always put them in the comments below and I'll see you in the next video

Channel: Rick Mulready

Views: 4,179

Keywords: ChatGPT-4o, ChatGPT, AI, online business, productivity, ChatGPT4

Id: QU8eRjZflWw

Length: 30min 16sec (1816 seconds)

Published: Sat May 18 2024