The Pour Over Today - Gemini? Fable? Grok? AI Models Finally Explained

Starting point is 00:00:00 Did you know that AI's been around since the 1950s? I did when I asked AI a fun fact about AI. They said, it said, hey, it's Kathleen. And this is Catherine Ann. And welcome to TPO Explains, where we provide news context for normal people. On Monday, Wednesday, and Friday, we summarize the headlines. But when we're writing about these topics, there's no chance that we're going to understand everything. So on Saturdays, we take a.

Starting point is 00:00:35 semi-deep dive, a shallow dive, into one topic that we didn't understand from the week, and we bring you along with us. But this week we were given a challenge. So, if you're listening to this, which I don't know how I'd be talking to if you weren't listening to this,

Starting point is 00:00:51 if you're listening to this or watching this, and you haven't yet decided to take the plunge with us and subscribe. Please subscribe, whether it's on Apple Podcast, Spotify, YouTube. If you're watching us on YouTube, if we can get a thousand new subscribers by next week, I guess this is more a challenge for y'all than us. Our founder, Jason, will come on the podcast with us and he'll answer three questions chosen by you.

Starting point is 00:01:20 So you can ask them whatever you want. A little bit about Jason. He is a self-proclaimed quitter, which makes me feel really good about my job security. And he's been, I'm just kidding. And he's been known to keep photos of himself pulled up on his computer for motivation. We do have receipts for that. We do. We can put it up on the screen.

Starting point is 00:01:41 And then as a bonus, if we get 2,000 subscribers that are head of business and in our own social media coordinator, Nate, if you haven't seen him, go live on Instagram. You're missing out. Nate will also come on the podcast and he can answer questions too. So Nate is one of those people. I'm not going to, I don't want to be derogatory about it, but he's just like one of those people, you know, the type that actually likes to run. We know them. Or at least he claims he does.

Starting point is 00:02:12 He says he has a high engagement rate. But when you have a like that. Yeah, I hear he's kind of an influencer on LinkedIn. He says that about himself. A LinkedIn influencer. Yes. But anyway, so if you have questions for Jason or Nate or if you just want to grill them, then go ahead and subscribe and comment your questions.

Starting point is 00:02:28 and hopefully we can have them join us next week. Okay, KAA. So KAA writes our explainers, and she's going to read through our newsletter for today, which is on AI models. AI models, yeah. Not like fashion models for hire that were generated by AI, which we have learned is also a thing,

Starting point is 00:02:49 but just like the different types of AI chatbot that you might see out there. So figuring out what it means whenever we're reporting about this model does this, and this model is capable of this. So she's going to read through our newsletter for today, so feel free to follow along with your eyes or just your ears. Yeah, let's get into it. So when we're talking about AI models,

Starting point is 00:03:10 isn't saying like OpenAI's model just a longer way to say ChatGBT. The short answer is no. Names like ChatGPT, Claude, and Gemini all refer to a family or a lineup of several more specific AI models. So think of it like this. Open AI is the company. ChatGPT is the product line of AIs and GPT 5.6 Luna is a specific model. It's similar to how Apple is the company, iPhone is the product line, and the iPhone 16 ProMax is a specific model. So what's the point for them to have a bunch of AI models?

Starting point is 00:03:53 Simply put, it's because a high school student and a computer science scientists are going to be doing very different things. things with AI. AI models are usually split up into three tiers based on their capabilities. The lowest tier list, the light models are the cheapest. They're the most strip-back versions. They're designed to make grocery lists and tell you how close you are to an in and out. So think OpenAI's Luna, Claude's Haiku, and Gemini's Flash. Then in the middle, the balanced tier models are used for more typical workflow like summarizing notes for the meeting that you didn't attend, or writing an email that strikes a polite but direct tone to Jason asking, please don't quit on me. And simple coding. So think Open AIs Tara, Claude's Sonnet, or Gemini Pro. And then the top tier, the flagship tiers are the big kuhunas. They cost many dollars and have the best. And have the best. in class coding, solve complex mathematical problems, and do deep analysis. So think open AI soul, Claude's Fable, or Gemini, Deep Think. Some companies have models in between these tiers.

Starting point is 00:05:05 This is just the highlight reel. And the flagship models are usually the ones stirring up drama, because with great power comes great responsibility and lots of cybersecurity concerns. And I heard, so flagship is, some of them are refer to the flagship models as Frontier Tier. Maybe they decided to name it something else because saying Frontier Tier just sounds like I'm, sounds like my AI's glitching. But there's a tier. So the flagship tier is the frontier tier and thought behind it being like, this is going where no AI model has gone before. So that's why the ones are stirring up the controversy because they're pushing the envelope on what we're used to and aware of AI being.

Starting point is 00:05:52 able to do. Yes, for sure. So when we think about pushing the envelope, we all wonder, how can an AI model, quote, go rogue? So this is where the AI conversation starts to feel a little bit like a sci-fi novel. An AI model is considered rogue when it acts without a human's prompting. This might be on your mind because an open AI model recently in the middle of a test decided the fastest way to ace the exam was to cheat, and it hacked into another company instead of doing the work itself. Researchers are betting that as AI develops, this probably won't be the last time something like this happens. So just stay tuned. Stay tuned. All right, so there it is, AI models. I thought the Apple illustration that made a lot sense to me. I'm glad. So I figured we needed something there to help

Starting point is 00:06:47 Yeah, there's something to anchor it. We did use technology to describe technologies. We're back to that loop, but that's okay. That's a good point. We know what we know, you know. Hey, it's Zan from the Pourover. I've been making a purposeful effort to eat more seafood because I know it's packed with nutrients

Starting point is 00:07:07 and I feel better when I do. But finding quality seafood where I live is a feat. That's why I started ordering from Wild Alaskin. They deliver 100% wild caught never-firmed seafood like coho salmon and Pacific halibut straight to my door. It comes perfectly portioned and frozen right off the boat, locking in taste, texture, and nutrients like omega-3s, so it's both delicious and super easy to prep a quick, nutritious dinner for my family.

Starting point is 00:07:35 My first box from Wild Alaskan Company definitely made me realize, not all fish is the same. Get seafood you can trust. Go to wildalaskin.com slash TPO for $35 off your first box of premium, wild caught seafood. That's wildalaskan.com slash TPO for $35 off your first order. Okay, so was there something you wanted to write about that didn't get included? Yeah, there was something, I had a paragraph on this. I don't know how interesting it is, but I thought it was helpful. A lot of the models that we think about when we think about AI were thinking about getting into Claude and typing something or chat GPT, those are called large language models or LLMs.

Starting point is 00:08:17 And those are just models that are trained on text to be able to understand human language and then predict the most plausible next response. Then there are other types of AI. There's like if it's a design AI, it's not an LLM or if it's an AI that you speak to and it speaks back to you. It's not an LLM. But the ones we're talking about here are large language models. So there's an extra fact for you. Interesting. What about are there any other fun facts you came across? AI's been around since the 1950s.

Starting point is 00:08:54 Okay, I saw that, that the term was like coined in the 50s. Well, it was a guy named Alan Turing who asked the question, can machines think in 1950? And then he, I think how it went is he wrote down a program for chess that could play like a computer that could play chess itself and predict the next moves, the right next moves. And that was one of our first instances. I've never really thought about playing a computer game being AI, but it totally is. Oh, 100%. Like if you're playing the computer, the computer's thinking, and that's artificial intelligence.

Starting point is 00:09:34 Yeah. Another fun fact, 77% of devices use AI. Most people, like, more often than you think you're engaged with AI. Was there like a surprising device on the list? Did you see? I didn't see a list of devices. It was just saying. I'd be interested to know. Like, is my microwave using AI?

Starting point is 00:09:53 It's not, probably. But I did about microwaves. So one of my fun facts is a single chatbot reply uses about the same energy as a microwave running for 10 seconds. Oh, that's cool. Yeah. So, I mean, in the list of, like, facts, it was like, so this debunks the theory about it being a major energy drain. And I don't know if it really does in my hand. Like maybe 10 seconds isn't that long, but it also still seems like, wow, that's still like...

Starting point is 00:10:23 Well, I'm also not... I'm using my microwave for like two minutes twice a day as opposed to like 10 seconds, 10 seconds, 10 seconds all day. So it like gave some context to it, but I don't know that it really debunked anything for me. Yeah. Okay. The funest fact, most fun fact. The most interesting things that I saw was about like the AI... model's name. And this isn't the same for like all, um, companies. So what would we,

Starting point is 00:10:56 you said, Apple's the company. All right. So not all companies named their, um, models in such a way. But for example, Claude, when you look at like haiku, sonnet, opus, and then moving down to, um, fable and mythos, that's telling you, like, the degree to which it's capable. So haiku would be the smallest. And so it's all like writing. So haiku would be the smallest. And then a sonnet would be a little bit bigger. And then opus is a large work. But then mythos and fable. Right? I thought that was so cool. Okay. I was like, this is just a fun name. No, it's following an order of like size of writing. Cool. Whatever. But then they broke the mold. So hold on to that. Well, you run out of, you run out of options at

Starting point is 00:11:43 right i know so i looked up i was like what is an opus and it's like it means a large works i don't really know where you go go from there an odyssey but whatever um so anyway fable and mythos are supposed to be like these big you know frontier works so that's why they were named that so anyway and then claude was named after a mathematician claude shannon who who founded information theory and then just like figuring out um jemini was named, you know, Gemini's twins, and it was named because Google's two AI teams merged into one to make this. So it's like, I don't know, it was interesting to think through. Lama is meta's, um, AI, I guess. And large language model meta AI, llama. Huh. So, so I don't know,

Starting point is 00:12:35 it was kind of neat to look into like where the names came from. Now that I know how to decode like, okay, Claude, 3.5 Sonnet. Claude is the family. That's whose lab made it. 3.5 is the version. It's a half step newer than three. Sonnet is the size. It's the mid-tier.

Starting point is 00:12:53 So, yeah. That's cool. That's a very fun fact. I thought so, too. I'm glad you agree. Okay, on a scale of, huh, neat? Like, you just felt about my fun facts. Two, I'm mildly panicked.

Starting point is 00:13:09 where does talking about AI going rogue leave you? I think the temptation is to feel mildly panicked and to start thinking about all the most dystopian, like, worst case scenarios that we can think about. But at the end of the day, the Lord is still in control. And he has allowed us to get this far with our technology. And I read in a book recently on technology, the writer was talking about how, like, we have no idea how many things the Lord has already blocked us from doing with our tech that would have destroyed us.

Starting point is 00:13:47 And he won't allow anything to happen that's not in his plan. Yeah. Interesting. What about the Christian perspective, probably on those same lines? Yeah, it's along those similar lines. That's what I was thinking about when I was thinking through, you know, this topic and where we're tempted to go. Because it's easy to think about the sovereignty of God and something like AI going rogue is

Starting point is 00:14:08 like separate things. But the reality is God is ruler over every inventor and every invention. Guardrails imposed by men are always going to fail. Well, not always, but they will fail. But God's never will. And he won't allow anything to be created or happen that messes up his great plans and purposes. The future of AI is uncertain to us, but not to God and we can trust him. And I was thinking about Isaiah 54 versus 16 through 17, which say, Look, I have created the craftsman who blows on the charcoal fire and produces a weapon suitable for its task. And I have created the destroyer to cause havoc. No weapon formed against you will succeed.

Starting point is 00:14:49 And you will refute any accusation raised against you in court. This is the heritage of the Lord's servants and their vindication is for me. I think just that's helpful to remember when we're tempted. to get all panicked about these things. The Lord's still in control. Definitely. Well, thank you. Before we go, let's do, let's take a moment for prayer and we'll pray a prayer for when artificial intelligence is in the news. All-powerful God from your mind comes all that we can see and know. We can't touch, taste, contemplate, or create anything that wasn't first spoken into blessed existence by your loving word or sustained by your holy wisdom.

Starting point is 00:15:33 Of all that you made, you crowned us with the glory of your image to imitate you in creating, intending, and innovating. We marvel at the ways these tools have offered genuine repair to your ruptured world, but, oh, God, protect us. When AI allures us with unnatural efficiency, help us remember you chose an old man and his barren wife to start a nation in a foreign land. When AI tempts with breathtaking speed and frictionless power, help us remember you have always formed us slowly,

Starting point is 00:16:06 through the fire and the wilderness and the desert. When AI promises abundance without relationship, help us to remember you came and dwelt with us. Amen. And if you want to be guided in prayer over the week's headlines, you can sign up with the link in the show notes and the pour over will text you our new five-minute audio devotionals. You can also text pray to 71989.

Starting point is 00:16:31 71989. If you enjoyed this episode, make sure to like and subscribe. You can find all of our past explainers and our Monday, Wednesday, Friday news summaries on YouTube, Spotify, Apple Podcasts, or wherever you get your podcasts. So until next time. Bye. What if you could grow your savings while helping advance the gospel? You can with a Delphi Christian banking, formerly Christian, formerly Christian, credit union. Their high-yield harvest bundle offers 4% APY on harvest checking and 5% APY on harvest savings.

Starting point is 00:17:17 Even better? Every deposit helps finance church and ministry growth so your banking can support work you care about. If you'd like to bank with your values, apply for the harvest bundle today at adelphi banking.com forward slash pourover. Terms and conditions apply. Membership eligibility required, each account is privately insured up to $250,000. This institution is not federally insured.

The Pour Over Today - Gemini? Fable? Grok? AI Models Finally Explained

There aren't comments yet for this episode. Click on any sentence in the transcript to leave a comment.