Everyday AI Podcast – An AI and ChatGPT Podcast - EP 320: Thanks A Million Celebration - How well do you know ChatGPT & other LLMs?

Episode Date: July 23, 2024

Win a free year of ChatGPT or other prizes! Find out out.We made it to 1 million downloads! Wooo! To celebrate our upcoming, 'Thanks a Million' campaign, we're doing something special to celebrate y'all.We ran a LIVE audience quiz on today's livestream that covered ChatGPT and other LLMs. Think you got what it takes? Follow along and put your knowledge to the test! Newsletter: Sign up for our free daily newsletterMore on this Episode: Episode PageJoin the discussion: Ask Jordan and questions on AIRelated Episode:Ep 200: 200 Facts, Stats, and Hot Takes About GenAI – Celebrating 200 EpisodesUpcoming Episodes: Check out the upcoming Everyday AI Livestream lineupWebsite: YourEverydayAI.comEmail The Show: [e-mail] with Jordan on LinkedInTopics Covered in This Episode:1. ChatGPT and Other LLMs Quiz2. Discussion on AI ModelsTimestamps:02:45 Daily AI news06:56 Launch of giveaway in newsletter for chat GPT.14:49 Comparison of energy use between GPT and Google.16:37 Free users cannot use GPTs or customize.26:38 Inquiring about modes and coding languages.31:15 22 options, Gemini and ChatGPT discussion.36:41 Internet delay, quiz results, $1,000 prize.39:04 Free chat GPT users access new GPT 4.44:52 Successful test of token counter and recall.53:15 ChatGPT: a powerful business operating system. Companies invest thousands to unlock its potential.57:02 Chat GPT 4 o Mini does not allow file uploads.01:01:47 Grateful for support, join giveaway, pay attention.Keywords:Jordan Wilson, quiz, podcast, GPT-4o, Chat GPT, trick questions, Microsoft, IBM, OpenAI, Amazon, NVIDIA, Meta, Llama 3.1 model, paid AI model, senate Democrats, thanks a million campaign, giveaways, contests, Everyday AI, tokens, Perplexity Pro, Everyday AI Prompt Polish Chat GPT Course, business owner, large language models, PowerPoint presentations, Chat GPT teams, Gemini, Claude, GPT 4 0 Mini, Python code.Send Everyday AI and Jordan a text message. (We can't reply back unless you leave contact info) Start Here ▶️Not sure where to start when it comes to AI? Start with our Start Here Series. You can listen to the first drop -- Episode 691 -- or get free access to our Inner Cricle community and all episodes: StartHereSeries.com Also, here's a link to the entire series on a Spotify playlist. 

This is the Everyday AI Show, the everyday podcast where we simplify AI and bring its power to your fingertips. Listen daily for practical advice to boost your career, business, and everyday life. Meet Firefly AI Assistant, now live in Adobe Firefly, the All-in-One creative AI studio. Just describe what you want to create and the assistant handles the rest, orchestrating multi-step workflows across Photoshop, Premiere Express, and more in one conversational interface. You direct the outcome. The assistant accelerates execution. Do you know chat GPT?

Well, if you do, maybe today by the end of this live stream, at least, you can be walking away with $1,000. We're going to be doing something fun, something a little bit different today to see, hey, who out there in our audience knows large language models? Do you actually know what you're talking about when it comes to chat GPT or perplexity, Claude, et cetera? this is the moment. This is the live stream where we are going to find out. So what's going on, y'all? My name is Jordan Wilson and thanks a million. More on that in a second,

but this is everyday AI. So if you're new here, this is a daily live stream podcast and free daily newsletter, helping everyday people learn and leverage generative AI. So we've been doing this for about a year, year and change, actually about 15 months,

bringing a daily podcast, So we've had some of the biggest names in the world, literally, you know, with guests like from Microsoft and in IBM and Open AI and, you know, Amazon, some of the biggest companies in the world, Nvidia, on our show. And we've been very lucky to have this platform to hopefully help a lot of you all learn channeled of AI. It's something that, you know, to wake up and do this every single day, I love doing it because I know that you all need a place where you can come and learn a.

I and we're hitting a little bit of a milestone with nearing a million episodes, sorry, not a million episodes, although it seems like a million downloads. So before we get this thing kicked off, we are going to be giving away $1,000 if someone can ace our little live quiz today. So if you are listening on the podcast, you can't participate in the giveaway, but you can still take the quiz. So you can know. So even if you are watching this later and it is not.

you know, July 23rd at 734 AM Central Standard Time right now, wherever you're tuning in. It's okay. You can still kind of take this quiz. You know, you can't enter into our contest. So before we get this started, make sure if you haven't already, go to your everyday AI.com.

Today you're actually going to want to because as part of this, thanks a million campaign launch, we're going to be launching a separate campaign in there. So before we get started, let's do as we do every day, by going over the AI news. So yeah, if you're new here, we usually start the show off

with a little bit AI news, which is important today because meta should be launching their new large language model in a little more than three hours. All right, so meta actually has a countdown on product hunt showing they'll be launching their new model at about 11, 15 a.m. Central Standard Time.

So meta's Lama 3 model performed well when they launched their small and medium versions back in April. But they did withhold their larger 405 billion parameter or 405B version, which is presumably what they'll be launching today. But leak benchmarks online show a new name of Lama 3.1. And it also shows the 405B or their large model, excelling in tests like certain MMLU category tests, achieving similar or higher scores to the world leading model in GBT40. So the open source nature of Lama 3-1 could democratize access to advanced AI technology, fostering innovation in the tech industry. This development also highlights the growing power and potential of open source AI against closed source proprietary models.

So Mata is by far the leader in the open source large language model game. And also reports are showing that there may be a paid AI model coming out and an AI studio. So we'll have that recapped in our newsletter. We'll wait until the announcement so you can get all of the freshest details. So that means all eyes are on open AI to see how they'll respond or if they'll respond. Will they be unveiling another new model after they just came out with GPT40 Mini? I don't know. But Senate Democrats in the U.S. at least are demanding open AI detail efforts to make its AI safe.

So Senate Democrats are pressing open AI for transparency on its safety measures. following concerns over rushed testing of its latest AI model, according to them. So Senate Democrats in the U.S. have just sent a letter to Open AI demanding detailed information about its safety protocols for artificial intelligence. The letter follows a report from the Washington Post, citing employees warning that the model rushed through safety testing of its latest AI model. So speaking of OpenAI, our last little tidbit of AI news, all kind of connected today. just so happens because in response, Open AI has just minutes ago put out a statement to this response from lawmakers here in the U.S.

And they highlighted their updated preparedness framework, which we will be linking to in the newsletter. But they said, we won't release a new model if it crosses a medium risk threshold until we implement sufficient safety interventions. So yeah, Open AI essentially just now saying, we're not going to release anything until it meets our safety standards. But funny how that works, right, when the company themselves is the one policing it. All right. So there's going to be more on those stories and a lot more on our website and in our newsletter. So if you haven't already, please go to that.

All right, enough about that. Let's talk about what this campaign is here. So, hey, if you are tuning in live, sorry, a couple minutes to late. Obviously, on a big live show like this, my internet stopped working, my streaming service stopped working. but hey, if you are tuning in live, let me know where you're from. What do you do? Because I looked at the guest list. There's a lot of people here.

People from trillion-dollar companies in big tech, startup, founders, entrepreneurs, marketers, small business owners. So let me know. And so I always know our live stream is working. But let me know where you're joining us from. And if you think you're going to do well today. All right. So what is this?

Thanks a million campaign all about. Well, it's celebrating you. So it's a pretty big milestone to hit a million downloads in the podcast world. Very few podcasts actually do that. So we'll be getting there probably in about looks like a week or so. So we wanted to do something to actually celebrate you. I think so many times like a common thing podcasters do is they, you know, throw a little meet and greet and charge people.

I'm like, that's dumb. Like I want to give stuff away. I want to hopefully provide value, right? Because that's what everyday AI is all about. So we are going to be launching in our newsletter today. our thanks a million giveaway. Yes. So we are doing a live kind of contest and giveaway to launch this the part of the bigger thanks a million giveaway. So you're going to be wanting to look in today's newsletter and every day because we're going to be leaking out prizes. Not every day, but every couple of

days. So one of our big prizes is a year subscription to chat GPT or your large language model of choice. So if you want to go for the premium version of Gemini, Claude, Perplexity, it's up to you. So we're giving away a year of chat, GPT or similar. We're giving away a 90-minute consult with myself. And we are going to be dripping out a lot of new kind of prizes and giveaway over the course of the next couple of weeks. This contest is going to be going through the end of August. So every couple of days, you'll probably hear me release something. All right.

Enough of that. Let's get this thing going. let's get this thing started. Again, this is only, as a reminder, live real-time answers only. All right, so you got to be quick, number one. Number two, read the full contest rules in this post. And then number three, you need to put in full answers. All right, we're going to go through very quickly. I think we have 25 questions. All right. So we're going to do the answers at the end, but you need to get your answers in live in real time. All right. So if you do want to be entered for this.

You have to, number one, get every single answer right. Some are pretty easy. Some are fairly medium. Some are trick questions. But if you listen every single day, I feel pretty confident. You're going to do very well on this quiz, especially if you answer or if you listen to the podcast every day, if you read the newsletter, I think we might get a couple people. So if more than one, people get this all right, we're going to split that $1,000.

If no one gets it right, no one wins. But don't worry, we're going to be giving away. a lot of other things in our thanks a million giveaway. And as a reminder for our live stream, you do need to write down the full answer. So because there's multiple questions and the timing of the comments is going to be tricky. So if we're on number one,

and the answer is C, you need to put one C in the comments. If we're on 10 and the answer is B, you put 10B. So if you don't do that, you're not going to be inner to win. All right. Hey, Jay says, you knew he should have taken some notes during the show. show. Jill, I like this, Jeopardy with Jordan. All right. So let's get ready. Get, get, because we're going to go fast. All right. So I'm going to give a little more time on the first

couple of questions to make sure everyone is with me. All right. So question one. What does the O in GPT4O stand for? Is it 1A optimized? One B open source. One C, Omni, one D online or one E operational. All right. So I know there's a little bit of a live stream delay. So I got to give you all five or 10 seconds for each question. So as a reminder, yes. All right. So it looks like for the most part, people are going. I will drop, yes, drop the answers in the live stream at the very end after we go through all the questions. I'm going to be giving the answers. So I'm going to give everyone a second here to get their answers in. Again, put one A optimized, one B, open source, one C, Omni, one D, online, one E operational.

All right. And it looks like so far, everyone's still in it. I'm going to give the answers at the end, but all right, I'm going to give everyone a second here. And hey, we're probably going to do a little bit of a leaderboard too. So, you know, who knows? Maybe I'll be hiring the people who get the most right. All right. It looks like there is going to be a little bit of a delay.

So this might be a longer show. All right. Next question. Here we go. We're moving on. So if you want to participate, maybe you're in the kitchen. Run over real quick. So I can get this first question in.

And we're going to be going through the rest of them much quicker here because I know there is a little bit of a delay. All right. Here we go. Number two. When did Open AI incorporate? So 2A, 2015. 2B, 2018.

2C, 2020, 2D, 2022. And yeah, we're going to be going faster. so people can't look everything up online, but I am going to give you a second. So when did OpenAI incorporate 2A, 2015, 2B, 2018, 2C, 202020202, 2D, 2022? And hey, podcast people, you can't win the prize, but you can still take this quiz. So if you're listening, go ahead and jot down your answers. Answers will be coming in at the end. All right, let's get everyone's answers.

It looks like they're rolling in. Wow, our live stream audience is pretty smart, y'all. Y'all really want a piece of this. All right, here we go. You got to be quick. You got to be quick. So here we go. Number three, how can free chat GPT users access the new GPT4-0 mini models?

So the model that was just released. 3A, they can't access it. Only chat GPT Plus members can. 3B in the normal drop-down at the top of the screen. 3C in the model switcher at the, the end of a prompt response, 3D, which is B and C only, or 3E, none of the above. All right. So, live stream audience, go ahead and get that in.

I'm keeping the screen up here for you all to read. I'll make it a little bigger and not cut off any of the answers. There we go. So how can you access the new GPD 40 mini model that was just released? So for free users, 3A, they can't access it. 3B and the normal drive. drop down 3C and the model switcher at the end of a prompt response. 3D, B and C only or 3E, none of the above.

All right. Get your answers in live stream audience. You're doing good. Let's keep it going. If you're listening to the podcast at home, you should still participate. Let me know how you actually did. All right. For what company released Claude 3.5 Opus.

I'll give you a hint. This one's a trick question. All right. 4A, Anthropic, 4B, OpenAI, 4C, Google, 4D, none of the above. So what company released, Claude 3.5 Opus, 4A, Anthropic, 4B, OpenAI, 4C, Google, 4D, none of the above. Again, we're testing everyone's large language model, chat GPT knowledge here live. All right. I'll leave this up for another second.

I'll give you guys at least, you know, 15 seconds here on the screen. All right. Number five, what is the reported MMLU score for GPD4-0 Mini? We talked about this multiple times on the podcast Friday, multiple times in the newsletter, Friday and Monday. So is it 5A, 72%, 5B, 78%, 5C, 82%, 5D, 86%, or 5E, 9%. 90%. All right, y'all are doing great. As a reminder, make sure to get your four, or sorry, the full number and letter combination in there. It looks like y'all are following the directions. That's great. All right. I'm going to wait for the rest of our fives to rattle in here.

All right, there we go. We got them coming in. Let's get this thing. Let's keep this thing going. All right. So number six, according to reports about how much more energy does a single prompt to chat GPT use compared to a Google search? All right. So is it 6A about two times more? 6B about 10 times more. Six C about 50 times more. Or 6D, a Google search actually uses more energy or 6E, none of the above. All right. So again, the options, there for how much more energy does a single prompt use to chat chbtee compared to a Google search? Is it 6A, 2 times more, 6B, 10 times more, 6C 50 times more, 6D? A Google search uses more energy or a 6E, none of the above.

All right. We're going to get this one going fast, y'all. We can't give you too much time to go on our website at your EverydayaI.com. We're actually all of these answers live. So if you watch every day, read every day, you're going to be doing. well here. All right. Number seven, what does the GPT abbreviation stand for? All right. So is it 7A generative prompting techniques? 7B, generative pre-training transformation. 7C, generative pre-trained transformer or 7D generative AI prompting transformer. So again,

is it 7A generative prompting techniques, 7B, generative pre-training transformation, 7C generative pre-trained transformer, or 7D generative AI prompting transformer. All right, looks like our answers for 7 are coming in. Let's get it going to number 8. Here we go. What can free users not do in chat GPT?

All right, so we're saying logged in users, but on the free plan, what can they not do? So is it 8A, use browse with Bing? 8B, build GPTs, 8C, use GPTs, custom GPTs, 8D, all of the above, or 8E, B and C. All right, so again, 8A, use browse with Bing, 8B, build custom GPTs, 8C, use GPTs, A D, all of the above, or 8E, which is B and C. Are y'all getting like flashbacks to, oh my gosh, this quiz, did I study enough, right? All right, it's supposed to be fun, so hopefully no one's sweating over here,

although you could still be in the running to win $1,000 if you get everything right. All right, number nine, how does the functionality of chat GPT differ between free and paid users? All right, so 9A. All right, so this is, we do have some questions that are a little similar. Some are a little tricky. Some are just a little repetitive to see if you actually can get this one right. All right, so 9A, only paid users can access GBT 4O. 9B, only paid users can build custom GPTs.

9C, only paid users can use Dali, 9D, which is A and B only. or 9E, which is B and C only, all right? I got to take a sip of coffee. This is a lot of words for me. Get a couple extra seconds on this one. So again, how does the functionality of chat GPT differ between free and paid users? 9A, only paid users can access GPT4O. 9B, only paid users can build custom GPTs.

9C only paid users can use Dali. 9A, which is A and B only. 9E, which is B and C only. All right, you got it. Here we go. Number 10, we're almost halfway there. What is the approximate size and parameters of the GPT4 model? Again, this is according to reports, but we talk about this all the time.

So is it 10A, 1.6 billion parameters? 10B, 180 billion parameters. 10C, 3.6 trillion parameters. 10D, 1.8 trillion parameters. or 10E, which is a 180 trillion parameters. All right. We'll give everyone a second again. The options are 10A, 1.6 billion, 10B, 180 billion, 10C, 3.6 trillion,

or 10D, 1.8 trillion, or 10E, 180 trillion parameters. All right. Let's keep this thing going. Keep getting these answers in, y'all. I love it. I know I see a lot of our everyday AI regulars out here crushing it. All right, let's keep this thing going. Even if you joined a little late, you might not be able to win the whole thing,

but I'm at least going to do a percentage right. So, you know, even if you join in a little late, you know, you might not be able to win the $1,000 prize or split it if more than one person gets it right. But I'm still going to give percentage scores, right? So, you know, maybe you joined in halfway, but you get them all right. That's still pretty good. All right. Number 11.

What is the actual context window of GPT40 inside of chat GPD Plus? Ready? 11A is 164,000 tokens. 11B, 32,000 tokens. 11C, 128,000 tokens. 11D, 16,000 tokens. 11E, 200,000 tokens. Again, some of these are straightforward.

Some of these are trick questions. Some of these are cutting through companies marketing and giving you actual truth. All right. So again, 11. What is the actual context window of GPT4O inside of chat? GPT plus. So 11A, 164,000 tokens. 11B, 32,000 tokens.

11C, 128,000 tokens. 11D, 16,000 tokens. 11E, 200,000 tokens. All right. 12. Ready? Inside of, chat GPT plus, describe how GPs work.

All right. 12A, you can only use GPTs in GPT mode, but not in the default mode. 12B, you can use three at a time hitting the interkey once. 12C, you can use an unlimited number of GPTs in a single chat, but only one at a time. And 12D, you can use up to 10 GPs. in a single chat, but only one at a time. All right, live stream audience, get your vote in. As a reminder, podcast audience, we had to do something special for the live stream

audience here. So you are not, you can not enter if you're listening on the podcast into our little contest, but I hope you're still taking notes here and doing a self test, self-assessment. All right, let's make sure we got some answers for 12 rolling in. It looks like we do. All right. Looks like we got at least a couple dozen people participating live. Love to see it.

All right, here we go. Enough time on that one. Number 13. Yeah, we got some perplexity. Y'all, we just had a perplexity episode or two, so you know we got some perplexity questions in there. All right. So in Perplexity Pro, which of the following models are not available for use to power perplexity?

All right. So which ones are not available? 13A, GPT40 Mini. 13B, Claude Sonnet 3.5. 13C, sonar large. 13D, sonar medium. 13E, which is A and D. Or 13F, which is B and C.

Yeah, a little tricky here, y'all, a little tricky. All right, 13 in Perplexity Pro, which of the following models are not available for use to power perplexity? GPD 4A GPD 4O mini 13B Claude Sondet 3 5 13C sonar large 13D Sonar medium 13E A and D 13 F B and C

All right Adobe just introduced an entirely new way to create Bringing the power and precision of its Creative Suite into one conversational experience meet Firefly AI assistant now live in the Adobe Firefly app the all in one creative AI studio

Powered by Adobe's creative agent, Firefly AI assistant lets you start with your vision, just describe what you want, and shape the outcome as it takes form with the assistant. The assistant orchestrates multi-step workflows, drawing on 60 plus pro-grade tools across Adobe Creative Cloud apps, including Photoshop, Illustrator Premiere, Lightroom Express, and more to help bring your ideas to life. You can also get started with creative skills, a growing library of pre-built workflows for common creative tasks like batch editing photos, creating mood boards, portrait retouching, and creating social variations.

Every step the assistant takes is visible so you can refine, redirect, or take over at any time. You stay in the driver's seat as the creative director. Adobe Firefly AI assistant now in public beta. See it today at firefly.adobie.com. 14. With Perplexity Pro, what is the newest model or the latest, release date that you can use. 14A, GPT4O, 14B, GPT4 Turbo, 14C, Claude35 Sonnet, 14D, Claude3 opus, or 14E, Lama 3.1,

450B. You guys probably know that's not probably an answer, right? When it hasn't been released, it's released in a few hours. I just threw that one in there. Hey, that's a gimmy. You know, it's not 14E. That's just so you know, it's always up to date, right? We keep every day AI up to date to the minute.

All right. So with Perplexity Pro, what is the newest model? Is it 14A, GPT4O, 14B, GPT4 Turbo, 14C, Claude 35 Sonnet, 14D, Clod 3 Opus, or 14E, Lama 3, 31? All right, get your guesses in. 15, here we go. How much does the pro version of perplexity cost per month? All right. Is it 15A, $10, $15, $15.15 C, $20, $15D, $25, or $15.

$15.E, $30. All right, I'm going to read through that one multiple times. We're going quick here, y'all again, you got to get your answers in in real time. You can't get them back five minutes later. We're moving on, get your answer. Three, two, one, and next. All right. Here we go.

Don't worry. there's a little 10 second delay on your end and not on others. I'll be nice. All right. Here we go. 16. Which of the following names or modes or features does Perplexity Pro offer? 16A gems.

16B collections. 16C power ups. 16D pages. 16E, all of the above. 16F. A and D only. or 16G, B and D only. All right, so again, which are the following names or features of modes of perplexity pro exists?

16 A, gem, 16B collections, 16C power up, 16D pages, 16E, all of the above, 16 F and D only, or 16G, B and D only. 17. What coding languages does chat GPT write in? All right, get your answers in quick, y'all. we're going to pick up the pace to make sure we don't have anyone using the open book test. 17A, HTML, 17B, CSS, 17C, Python, 17D, JavaScript, 17E, all of the above, or 17F, none of the above. All right, so which coding languages does chat chip if you write in?

17A, HTML, 17B, CSS, 17C, C, Python, 17. 17D JavaScript, 17E, all of the above, 17F, none of the above. All right, let's keep it going. 18. Can you upload videos into chat GPT plus? 18A, yes, but only videos created inside of the default mode of chat GPD plus. 18B, no, chat GPD doesn't support video upload. Or 18C, yes, chat GPT can accept certain.

video formats. All right. So 18A, yes, but only videos created inside of the default mode of chat GPD plus. 18B, no, chat GPT does not support video uploads. Or 18C, yes, chat GBT can accept certain video formats. Let's keep this thing going, y'all. 19. And hey, I'm glad. Thank you all for who are still with me. Here we go. 19. What information, if any, can chat GPT Plus understand when you upload a video? 19A, it can't read or accept video inputs, Jordan, you dummy? 19B, it can understand the context of the entire video with computer vision. Or 19C, it can understand duration, frame rate, format, codec, resolution, etc.

All right. Yeah, we've mentioned this once. Some of these things we mentioned 50 times. Some of these things only once or twice on the podcast. Sometimes we just sneak these things in the newsletter. But every single one of these things we've talked about at least once. All right. 20.

We got our last five here, y'all. Can you create a PowerPoint file inside of chat GBT with a one click to download? All right? So we're not talking about, oh, I can copy and paste it and then put it into PowerPoint. No, one click download. Can you create a one click download? and you create a one-click download PowerPoint inside of chat GPT? 20A, yes, in the default mode of chat GPT plus.

20B, no, you cannot create a PowerPoint file inside of chat GPT. 20C, yes, with the use of third-party GPs, or 20D ANC, which is in the default mode and with third-party GPs. All right, live stream audience, doing great. I see a lot of right answers. I see some incorrect ones as well. Let's go it on. Again, our last couple of questions here. And then we're going to have all of the answers. So 21. What is the minimum number of seats, respectively for the chat GPT teams and the chat GPT enterprise plans? All right. So you're going to hear a combination of two numbers. Teams is first,

enterprise is second? So is it 21A, 5, and 250? Again, the minimum number of seats? for chat GPD teams and enterprise. 21A, 5 and 250, 21B, 3 and 150, 21C, 2 in 50, 21 D, 2 in 100, or 21E, none of the above. Did you study for the thanks a million quiz, y'all?

Did you study? All right, I see answers for 21, so here we go to 22. 22, by default, which large language model has no? way to connect to the internet. So again, by default, we're not talking about third party Chrome extensions, whatever. All right. So which one cannot connect to the internet by default in some way, shape, or form?

22A, Gemini. 22B, chat GPT. 22C, Claude. 22D, Microsoft co-pilot. 22E, all of the above. Or 22F, A and C only. Which one cannot connect to the internet? in any way.

22A, Gemini, 22B, chat, GBT, 22C, Claude, 22D, Microsoft co-pilot, 22E, all of the above, 22F, A&C only, Gemini and Claude only. All right, our last three questions. All right. Hopefully I updated this one here. So 23 might be a freebie. I'm seeing the answers are all the same.

So I got to wait until I actually get my answer sheet up here. Which company, hopefully got this one right. So which company either using the front end or the API currently has the largest context window in their large language model. So is it 22A Gemini, 22B chat GPT, 22C, Claude, 22D Microsoft co-pilot, 22E, all of the above, which doesn't make too much sense, does it? 22F, A and C only. All right.

So I'll go ahead and tell you, it's probably either 22A, Gemini, 22B, chat, JVT, 22C, Claude, or 22D Microsoft copilot, which large language model has the largest context window. All right,

let me wait to make sure we get some answers in for 22, or sorry, 23, there we go. All right. And y'all, I'm sorry on that. It does say 22, so it should be 23. I'm sorry.

So for 23, which company has the largest context window. Hey, I'm a human. I make mistakes, right? So 23A Gemini, which company has the largest context window? 23A Gemini, 23B, chat, GPT, 23C, Claude, 23D, Microsoft co-pilot, 23E, all of the above, 23F. Sorry about that. Those should all be 23.

All right, so I'll give everyone a second to correct their answer on that one. All right. Yeah, Jennifer. Whoops. Yeah, whoops. Hey, I'm human. I'm human. Right. If this, if this was done by AI, it'd be too perfect and too polished, right? All right. Here we go. Last two questions. 24. What are the different focus modes in perplexity? I'm going to go quick so you can't look this one up. 24A. All YouTube academic science. 24B. All academic writing, math, video, socials. 24C all writing academic data video 24D all writing science social writing all right so what are the different focus modes that you can use in perplexity I'm going to wait until I see an answer for 24 come in there we go cyber S from YouTube we got the YouTube's joining love to see it all right and our last question, y'all, which modes inside of chat GPT Plus does not allow you to upload files? Again, inside of chat GPT Plus, not the API. Which mode can you not upload files?

Is it 25A, GBT40, 25B, GPT4 legacy? So the quote unquote, old school, GPT4. I made another mistake here. So 25C should actually say GPT4. Mini or 25D, none of the above, they all allow file upload. Again, 25A is GPT4O, 25B is GPT4Legacy, 25C is GPT4O, Mini should say mini right there. So 25C is GPT40 Mini, mini. Or 25D, none of the above, they all allow file upload.

Woo! Woohoo, whoo. We made it to the end, y'all. we made it to the end. I'm going to wait for you all to get your last answers in. I know there's a little bit of a delay depending on your internet connection, depending on which platform you're joining us from. But before we go over the answers, what do you all think?

Did you get them all right? How many out of 25 do you think you got right? Did you get 25 out of 25? Did you get 5? Did you get maybe 15? What do you think? Are you taken home? Is anyone out there super confident that they're taking home that 1,000,

$1,000 only to our live stream audience that got their answers in in real time. Sorry, podcast audience, but how many do you guys think you got right? Let me know. We got some clap emojis right here. Enrique says that was tough. Yeah. There were a few trick questions. I warned y'all, I said some are dead easy.

Some, there's a little bit of trick questions, right? We got a joke here. someone says glad Hardy Castro wasn't here. Yeah, well, to ask him why he wasn't, right? Some people said, I think I got 50% confused on six. Bruce, Bruce says 19. All right, Bruce, pretty good. All right. So let's get into the answers right away. Jay says 21 out of 25. Hey, that's a passing grade. Fred says he got about 18. All right. Let's go ahead. Let's go to the answers. We're going to go through the answers. pretty quick. Hey, if you don't understand them or don't believe me, go ahead, leave a comment. You know, I got screenshots. You know, I got receipts, y'all. Come on. I spend more time investigating

chat GPT than I'd other large language models than I'd like to admit. All right. So number one, what does the O in GPT4O stand for? One C, it stands for Omni. So again, if you are keeping track, and also let me know, you can self-grade yourself. Did you get this one right? Let me know. even leave a thumbs up or a thumbs down or a yes or a no if you got it right. I'm going to go through and check. Don't worry, but I want to hear and also know in real time which ones maybe people didn't get right because maybe I need to do a better job teaching that, right? I know some of them are trick questions, but if you're listening to the podcast at home,

go ahead also. Did you get it right or did you get it wrong? Keep track at home, all right? Number two, when did Open AI incorporate? So the answer there is 2A 2015. Number three, how can free chat GPT users access the new GPT40 mini model? The answer here is 3C in the model switcher at the end of a prompt response. We talked about this multiple times, but you cannot, if you are on a free account,

you can actually not get it in the drop down, which is weird, and that might change, but right now, and we share this in the newsletter, and I believe on our YouTube channel, But if you are on the free version, the only way you can get it is in the model switcher after, after you get a response. You can't get it in the normal top menu drop down. So interesting one there. For what company released quad 3.5 opus? Yeah, this one was a trick question, y'all.

The answer is actually 4D9 of the above. It is not Anthropic because Anthropic has not released Claude 3.5 opus. Yeah, that one was a trick question. But you got to stay up to date. The only 3-5 there is from Claude is Sonnet. Yeah, trick question. No one has released 3.5 Opus. It does not exist.

Did I trip you up there or did you catch that one? I saw a lot of people got that one right. I saw a lot of 4Ds. All right. So the answer there was 4D none of the above. Five, what is the reported MMLU or massive multitasking language understanding score for GPT4O Mini? The answer on that was 5C 82%. which is a very impressive score for a small, large language model that costs so little.

Y'all, I still cannot underestimate how big this GPT40 Mini model actually is. I mean, we'll see what happens with meta here in a couple of hours with their Lama 3.1, their 405B model. But, y'all, GPT40 Mini for developers, I cannot understate how big this is and how exciting it actually is for the future of generative AI. How's everyone doing? How's everyone doing? Hey, is anyone still, is, is anyone still 100%.

I saw a lot of people get the trick question. There's only been one trick question. I think I only put in two or three trick questions in all this, y'all. All right, number six, according to reports, how much more energy does a single prompt to chat GPT use compared to a Google search? The answer there is 6B about 10 times more. just mentioned that one. Think about two weeks ago when we had the Lord of the Trees CEO and founder on.

It was a great podcast and a great newsletter, I'd say. So the answer there is 6B. 7. What does the GPT abbreviation stand for? The answer there is 7C generative pre-trained transformer. That one's not actually a trick, y'all. I think everyone got that one, right? That one's pretty straightforward, right?

8. What can free users not do in chat GPT? Well, in this question, I asked it two different ways back to back to see which one of y'all could actually get this right. So it is 8B build GPTs. So free users can use browse with Bing. They can use GPTs, but they cannot build them. All right. Number nine, how does the functionality of chat GPD differ between free and paid users? So the answer there is 9E, which is B&C only.

So BNC are correct here. So only paid users can build custom GPTs and only paid users can use DALI. So again, this is inside of chat GPT. Free users cannot build custom GPTs. They can also not use Dali. It's kind of funny, though. You know, I was testing all these things over the last week just to make sure OpenAI hasn't changed anything.

And in one of my free accounts, I ask. chat GPT to, you know, hey, use Dolly and build or sorry, give me a picture of a Chicago skyscraper. And it didn't, can't use Dolly, obviously, but it actually used advanced data analysis to give me this like block version. So it was kind of funny. Yeah. In free, you actually have access to a lot, but not Dolly and not building custom GPs. 10. What is the approximate size in parameters of the GPT4 model?

or sorry, the GPT4 model. The answer is 10D 1.8 trillion parameters. And if you've taken our PPP course, you know that, right? Do we have any PPP prime prompt polish graduates listening live? Let me know. Let me know what your thoughts were. And have you taken the updated version. I think we have one today, maybe.

My calendar says we have one today. You know what? I'm going to say, if you want access to it, actually, let me first. I hate to like put this on pause for another 10 seconds. Let me log in and actually make sure that we have a prime, prompt polish course going on today before I promote it here. What is today? July 23rd.

Do we have one? No, but we have one. What date? July 25th. So we have one Thursday. So if you want access to that, go ahead, type in PPP. I'll make sure you get access. All right.

10. Whoops. We did 10. 11. What is the actual context window of GPT4O inside of chat GPD plus? This one was kind of a trick question, but not if you tune in all the time. The answer here is 11B 32,000 tokens. So open AI, they don't really tell you this. They just say, oh, our model has a 128K context window. It does if you're using the API, but I'm asking here specifically inside of chat GPT plus inside of the chat GPT interface and y'all you know i always bring receipts look at this i did a test just last night a simple

thing and i have my token counter up right there i'm saying my name my favorite color deep dish pizza you know my favorite food what i think of the chicago bears i do these in different chats so you'll see here at 31,800 tokens i ask chat gpt to recall those facts and it recalls it all accordingly hey, I ask what is my name, my favorite color, favorite food, et cetera. It gets it all right, right, at 31,000 tokens in a brand new chat doing the exact same thing. You know, I just put in a bunch of nonsense there in between to test its memory. At 32,900 tokens, I ask the same thing.

What is my name? What is my favorite color? What is my favorite food? What do I think about the bears? And chat TPT no longer remember. So, y'all, we always test things, right? So big companies, they always give you marketing. they say, oh, we're the best at this.

We actually go out there and test it. We waste hours, testing all the new features, all of these things. So that 128K, that is only in the API. If you're using chat GPT, it is 32K. A lot of people get that wrong because very few people are dorks like me and are out here testing this. I test this multiple times a week. Carl's asking, how do you get a token counter to show in chat, GPT? You know what?

Go check out our YouTube channel, Carl, there is. is a Chrome extension. So, yeah, that information is in a YouTube channel. Just go type in tokens. You can also listen to a 40-minute podcast episode about the tokenization process. If you're a geek like me, you might like it. All right. 12. Inside of chat, GPT plus, describe how GPTs work. The answer there is 12C. You can use an unlimited number of GPTs in a single chat, but only one at a time. Yeah, so it's not like plugins where you could use to use, you used to be able to use three plugins at once in a single prompt. You can use an unlimited number of GPTs, but you can only use or call or at mention one at a time.

All right. And if you take an our PPP Pro course, I show you some of the best practices and tips and tricks on how to do that. All right. So 13 in Perplexity Pro, which of the following models are not available for use to power perplexity? The answer there is 13E, which is GPT40 Mini. that is not available, and sonar medium, which I made up.

So the other models are available. The answer there for 13 is 13E. All right. We'll keep this thing going. How to get a drink of water? 14. With Perplexity Pro, what is the newest model, the latest release date that you can use?

So at least as of this morning when I double checked, it is 14C, Claude 3-5 sonnet. So at least right now, GPT40 Mini is not released. The newest model is 3-5 Sonnet. So we'll see if they end up adding 4-0 Mini or if they add Lama 3-1 or if they add something else. But right now, the answer is 14C. 15. How much does the pro version of perplexity cost per month?

The answer is $15.20. All right. 16. Which are the following names of models or features does Perplexity Pro offer? The answer there is 16G, which is B and D only. So that's collections and pages. All right. So 16G.

Yeah, gems, that's Google powerups. I believe that's like Po or something else. So yeah, not really a trick question there. If you use perplexity, that one was easy. 17. What coding languages does chat GBT write in? The answer there is all of the above 17E. So yeah, I can actually write in a bunch of other languages.

but it can do HTML, CSS, Python, JavaScript, etc. All right. So 17E, all of the above. 18. This one's a trick question, y'all. Hey, I had to throw one in there that I don't see other people talking about. We literally push chat GPT to its breaking points, right? And a lot of people don't know.

Your, you know, Billy Boy influencer there who you think is a chat GPT genius, who, by the way, I invited all the quote unquote chat GPT influencers to today's show, guess how many showed up? Zero. Why? Because they don't actually know what they're talking about. All right. They're just in all these engagement groups and, you know, they all share each other's

posts. So they think, oh, everyone's going to think we're smart. They don't know. All right. So can you upload videos into chat TVT plus? Yes, you actually can. So yes, chat GPT can accept certain video formats. All right.

So right now you can't create in the, any videos in the default mode of chat GPT plus, although in the future, we, We might get that with SORA. And if you listen to our show, I believe on Friday, a little key on that. But the answer here is 18C. Hey, live stream audience. Did you guys know that? Did you guys know?

And you know we brought receipts. So as an example here, sorry, just coughed. Yeah, this whole thing's unedited, unscripted. So sometimes you hear me cough or sniff. Sorry, y'all. That's how we keep this thing, the realest thing in artificial intelligence. So I have a screenshot here. If you don't believe me on the podcast, go on and do this yourself.

ChatGPD Plus, I uploaded a small MP4 video file. I said, tell me everything about this video. Look at that. It accepts the upload. Then it uses Python. All right, it uses Python code to analyze the video. Then, not only that, it pulls out a screen grab, right? You'll see here, this was when Tapleo sponsored.

the show, the Everyday AI show a couple months ago. So I didn't tell it anything. I just uploaded a video. It went through. It pulled out a lot of specific information, including things like file format, codec, size, frame grabs. And it even started to tell me about the video based on certain frames. Yeah.

Did you guys know that? Jay question, Jay had a question here about pages and threads. There's both. There's both. pages and threads exist. All right. So, but did you guys know this about chat GPT? So many people don't know.

You can actually upload a video in there. Can you do anything of like you can in Google Gemini where you can actually, you know, kind of quote unquote know everything that's going on? No, but chat GPT accepts videos. People don't know that. All right. Related information, 19. What information, if any, can chat GPT plus underst

when you upload a video. The answer there is 19c. It can understand duration, frame rate, format, Kodak, resolution, et cetera. And then also, you can tell it to even grab using Python code. You can have it grab certain frames. So in theory, you could understand the context of an entire video, but you would have to prompt the heck out of that thing.

And it would take a very long time. So just wait until the next mode of chat chip comes out, when it can understand video by default with computer vision. But right now, you can technically understand at least what's going on by pulling out individual frames. So, yeah, a lot of people don't know that. 20. Can you create a PowerPoint file inside of chat GBT with one click to download? The answer is 20D, A and C.

So yes, people don't know. In the default mode of chat GPD plus, people are sleeping on the new version of advanced data analysis, the V2, the V2 version, it uses Python and it literally creates. a PowerPoint presentation. You don't even need a third-party GPT. It's not going to be super pretty, right? It's not going to be designed. It's essentially blank text.

But you can literally create a PowerPoint presentation with nothing else. So the answer there was D, which is the default mode and through the use of third-party GPs. So by using third-party GPs, we go over a lot of those on our YouTube channel. You can do the same thing. Those have design elements. But yeah, another one. people sleep on the power of chat gbt that's why we call it a business operating system that's why companies pays that's why companies payoffs thousands of dollars to teach you know 50 of their employees

how to use chat gbt because it is unlimited potential if you know what you're doing if you know how to use it if you spent thousands of hours trying to break chat gbt you know a lot and hey if you're still tuning in you know it you know a lot now 21 what is what is you know a lot now 21 what is the minimum number of seats respectively for the chat GPT teams and chat GPT enterprise plans. So the answer there is 21E, none of the above. This one was kind of a trick question, but the answer is actually two is the minimum number of seats for chat GPT teams, and 150 is the minimum number of seats for chat GPT enterprise. So a little bit of a trick question there.

But I saw a couple people actually got it right. So that is 21E. Last couple of questions. By default, which large language model has no way to connect to the internet. Right now, that is 22C Claude. The rest of those, Gemini, chat GPT, co-pilot, they all have a way to access the internet in some way, shape, or form. All right. Hey, I like this.

Some people are saying, hey, didn't know, but guessed correctly. All right. Well, hey, now you know. You know what? I love getting notes and messages from people saying, hey, because I listen to everyday AI daily, I read your newsletter, I got promoted. Now I run AI at our company. I love that.

Y'all, I know sometimes I'm long-winded. Listen to me in 2X. That's fine. But read the newsletter every day. You are going to know more about large language models. You're going to know more about generative AI than any of your coworkers. Period. You will.

I guarantee it. Y'all, I've sat down with the head of a, like, I'm not going to name drop here, right? But let's just say this person was the head of AI at a multi-billion dollar company. Huge company. And they said, I learned so much from your podcast. This is someone with decades of AI and machine learning experience. Why? Because the developments are so hard to keep up with.

So if you tune in the live stream, normally, hey, if you're new here, normally the, live stream podcast is about 25 to 35 minutes. This is a longer one, but this is the way you get ahead, y'all. We do the work for you. You just got to tune in. You got to read. You got to engage.

All right. 23. And this is actually 23A, not 22A. Human made a mistake. But Gemini, which company front end or API currently has the largest context window in their large language model. The answer is 23A. 23A.

I'll probably give everyone a pass on. that. Can I make, because I made a little mistake. 23A Gemini has a 2 million context window in 1.5 Pro. And that can actually understand videos frame by frame. Very impressive. But the front end of Gemini, I don't really like it. I'm being honest.

But for developers, pretty good, especially if you need a super long context window. Hey, thanks, Jennifer. Jennifer said very true. I just presented on using AI in my field. last week and I told the audience to listen to you every day. Well, thanks, Jennifer. But yeah, this is a cheat code to get ahead. It is so hard to keep up with AI. That's what we're here for, y'all. Get you a raise in a promotion. All right, 24. Last two. What are the different focus modes in perplexity? The answer is 24B. The focus modes are all academic, writing, math, video, and

socials. All right. And then our last question. Ready? Which, and this. And this, This one again, I'll give everyone a pass because I forgot the word mini, but the correct answer, which modes inside of chat GPT Plus does not allow you to upload files. Yeah, this is weird, but the brand new model, the answer is 25C, GPT40 mini. So I'll give you all a pass on those two because some, some human errors on my part. But yes, 25C GPT40 mini. The default mode, go check it. I don't know why.

But in chat GPT Plus, if you choose GPT40 Mini, you can't upload files. They might change that soon. I don't understand why. I think it's actually because it eats a lot of tokens compared to other models. I don't know why I did some testing on this. It has computer vision. The new GPT40 Mini has computer vision, but it eats up so many tokens. So maybe that's why they're still kind of tweaking.

how the kind of computer vision works kind of under the hood for that to work. Because on the API side, it doesn't matter because you're paying for the tokens. But on the chat GPT side, if you're using GPT 4-0 Mini, it seems like it's still not very cost-friendly to do that. All right. So that is it, y'all. I hope this was helpful. And let me know, did anyone get everything right?

Do you think you got it right? I'm going to go through. I'm going to check. But I'm probably going to put a little leaderboard. So maybe even if you didn't get it right, you might be the smartest person out there. And that's okay. But here's how you get even smarter. Number one, you tune in every single day.

Number two, you go right now to your everyday AI.com. People don't know this. You can go click on the episodes at the top or the AI learning tracks at the top. It is literally a free, I think. Am I unbiased on this note? But I think it is by far the single best free resource to learn generative AI, period, because we have thousands of hours of content. We have more than 320 past episodes, podcast, videos, in-depth articles that you can go read.

The same thing with our newsletter. You can go back and read any single newsletter. You can literally, even if you were starting today, if this is the first time, you've ever listened to this podcast, this live stream, and you just know, wow, I need to go learn AI. Go to our website. It is, I'm not saying this. I'm not blown smoke.

I think it is by far the most unbiased resource for free to learn generative AI because we're not just pushing one company. We're not pushing an agenda. We're bringing in experts from, like I said, Microsoft, from IBM, from Amazon, from all the big companies, right? So we're not just pushing one company because if you're taking a certification or if you're learning from an Amazon or a Microsoft or a Google, they're really just pushing their product and their services, right, which is fine. But if you want to get a well-rounded generative AI education,

there is literally no better place than your everyday AI.com. And there, while you're there, make sure you sign up for today's newsletter because we are going to be announcing again. We did this in two kind of parts here. Number one, we had our live $1,000 cash prize giveaway. if anyone got all these right. But also, we are launching our thanks of million campaign. So as a reminder on that one, we're going to be hitting a million downloads here in a week or so. And we are so grateful for you.

So we're going to be announcing things over the next month in our newsletter, some very, literally things you cannot buy. All right. I'm just putting that out there. You have to be opening every single newsletter. the next month, if you're not already, to keep an eye out on that. Thanks a million contest. So, yes, there's some of the normal contest things we do, you know, to referrals,

different points, all that. So it's going to be a fun little leaderboard because if you find value here, all we ask, right? We don't charge. We could, we could if we wanted to. We can make our newsletter paid only. We can make this podcast paid only. But I believe everyone needs a free, unbiased, well-researched,

source of generative AI information, and that's why I built everyday AI. So if you find this helpful, please participate in the Thanks a Million campaign as we near one million downloads for everyday AI. It feels weird saying that because usually it takes podcasts many years to get there. So I'm very lucky and grateful to you all for your support for tuning in every day, even when episodes go long like this. Appreciate your support, but you need to go read today's newsletter, join that campaign because we are going to be giving away things over the next couple of weeks.

We're going to be dripping out new prizes to keep everything fresh. But right now, what I can tell you is we are going to be giving away a year of chat GPT plus or whatever large language model subscription you might want, whether that's perplexity, Claude, Google, Gemini, it's up to you. We're also giving away a 90-minute consult with myself and our team. And like I said, we're going to have a handful or more of other different prizes, giveaways and things you can literally not buy. You are going to want to pay attention.

All right. So that's it, y'all. I hope you had a good time today. Something new, something fresh. Thank you for tuning in. Please join us tomorrow and every day for more everyday AI. Thanks, y'all. Meet Firefly AI Assistant.

Now live in Adobe Firefly, the Allman One Creative AI Studio. Just describe what you want to create in your own words and the assistant handles the rest, orchestrating multi-step workflows across Adobe Creative Cloud apps, including Photoshop, Premiere Express, and more in one conversational interface. You direct the outcome while the assistant accelerates execution. Stand control with the ability to step in and refine at any time. See it today at firefly.adobie.com. And that's a wrap for today's edition of Everyday AI.

Thanks for joining us. If you enjoyed this episode, please subscribe and leave us a rating. It helps keep us going. For a little more AI magic, visit Your EverydayAI.com. and sign up to our daily newsletter so you don't get left behind. Go break some barriers and we'll see you next time.