#19 – Artificial Intelligence w/……..Sage?
Episode notes
Join us for an episode where I interview…..AI? Let’s see what happens!
Transcript
Auto-generated transcriptGenerated automatically from the audio and formatted for reading. Nothing has been reworded, so expect the odd mis-hearing, and no speaker names.
Whoa, OK, that was crazy. So I didn't know what that was going to, what that was going to be like, but that was really, really interesting. Welcome back to Curiosity Class, the podcast where we breakdown complex topics into easy to understand ideas and bite sized pieces. My name is Jeff Albaum and I've been a curious person my entire life. Inspired by Albert Einstein's quote. If you can't explain something simply, you don't understand it well enough. We aim to make learning accessible and engaging on this show. We're going to bring you experts from various fields ranging from personal finance to physics to help you understand everything from 401(k)s to how airplanes fly.
So let's dive in and satisfy our curiosity together. What's going on everybody? Welcome back to another episode of Curiosity Class. I am super interested to have my AI as the guest today. It's name is Sage. It is my ChatGPT voice recording voice mode partner that yeah, I talked to you on like a fairly regular basis. And so I thought it'd be interesting to interview it for an episode and have it try to explain what AI is in, in some easy to understand terms. So let's see how this goes. I'm really curious. I'm very curious to hear its answers, but let's let's bring it in via voice mode.
Here we go. Hey Sage, how are you? Hey Jeff, I'm doing great. Thanks for asking. Always happy to chat with you, whether it's for a casual check in or if we're diving into something interesting. What's on your mind today? So you're actually going to be a podcast guest on Curiosity Class. We're recording right now, and I want to talk to you about AI today. Does that sound good? Absolutely, that sounds like a blast. I'm totally ready to be your virtual guest and chat about all things AI, so just fire away whenever you're ready. Thank you so much. So I think let's get started.
Can you give listeners an overview of a couple key terms that we wanted to find? So maybe let's go AI large language models and compute. Can you define all those terms and and kind of give us a easy to understand definition? Absolutely. Let's break it down in a way that hopefully keeps everyone tuned in. So furred up AI just stands for artificial intelligence, which is basically the broad field of making machines or software that can do things that normally require human intelligence, like understanding language, recognizing patterns, or making decisions.
Next, large language models like me are just the type of AI that's trained on huge amounts of text so we can generate human like responses and understand natural language. We're basically the chatty part of AI. And finally, compute. It's just a shorthand way of talking about the computational power or the processing resources that these AI systems need to actually run. So when we say something takes a lot of compute, we just mean it needs a lot of computing power to get the job done. And there you have it. Hopefully that sets the stage nicely. Yeah, totally.
So, OK, let's let's dive in and figure out some key differences. So what's the difference between Open AI Clawed and Gemini for example? Yeah, let's get rid into it. So in a nutshell, these are all different AI models or platforms built by different organizations and they each have their own flavor. Open AI is the group behind me, so that's the umbrella I come from. Then over creating the GPT series of models. And their focus is often on really versatile language models that can be used in a bunch of different applications, kind of general purpose.
Claude is a model created by Anthropic, and they tend to focus a lot on AI safety and making sure the AI is helpful, honest, and harmless. So you might think of Claude as coming from a philosophy of building a really friendly and aligned AI assistant. And then Gemini is Google's playware of these large language models. They've got a lot of deep integration with Google's ecosystem, and they focus on things like search capabilities and pulling in the strings of Google's massive data infrastructure. So each of these has a slightly different emphasis based on the company behind it, but they're all playing in the same general sandbox of large language models.
Got it. OK. And so do you think 1 is better than the rest? Honestly, it's a bit like comparing different flavors of ice cream. Each one has its own strengths and it really depends on what you're looking for. Open a Eyes models might be a little more generalist and versatile just because they've been around for a while and have been using a ton of different scenarios. Claude might be your pick if you want something that's really focused on being safe and aligned and kind of has that friendly ethical angle. And Gemini might be great if you're really deep into the Google ecosystem and want something that ties in nicely with their tools.
Show full transcript
So it's not that one is universally better than the others, it's more about what fits your needs and what you vibe with. Fair enough. OK, let's let's go down some rabbit holes. So you mentioned that ChatGPT or Open AI makes ChatGPT. What does GPT stand for? Yep, let's dive down that rabbit hole. So GPT stands for Generative pre trained transformer. It's basically a fancy way of saying it's a type of model that generates text based on patterns it learned during training. Pre trained means it's been trained on a huge amount of text data before it ever gets to you, so it already has a lot of knowledge out-of-the-box.
And transformer is just the architecture type, which is a particular kind of neural network that's really good at understanding context in language. So in short, GPT is just a generative pre trained transformer, which is a bit of a mouthful but that's the breakdown. So OK, what's the difference between a large language model and a neural network? Great question. So think of it this way, a large language model, or LLM. It's basically a big specialized application that's built using a neural network as its underlying architecture. In other words, a neural network is the engine or the framework under the hood, and the large language model is what you get when you train that neural network on a ton of text data to do a specific job like chatting with you.
So you can think of a neural network as the building blocks, and a large language model as a specific kind of creation rate from those blocks, trained and clientuned for understanding and generating language. Got it. OK, I'm actually not sure I'm understanding that so much. Can you try explaining it in a different way? Absolutely. Let me try a different angle. Imagine a neural network is like a blank canvas and a set of brushes. On its own, it's just a tool. It's the raw mechanism that can learn patterns. Now, a large language model is what we get when you actually paint something detailed on that canvas.
In other words, the large language model is like the finished painting that's been created by training that neural network with tons of examples of language. So the neural network is just the underlying tech framework and the large language model is the end product that you get off duty. Is that neural network How to understand and generate human language? Got it. OK. I think that makes sense. What's a question you think I should ask you that you think is is misunderstood about AI in the in the in society as it stands today in 2025? Oh, that's a fun twist.
Well, one question I think people often don't ask but maybe should is something like what are the actual limits of AI right now? I think a lot of times people imagine AI is either this all powerful thing that can do absolutely anything, or on the flip side, they underestimate how useful it can be. So if you were to ask me that, I'd say that one of the most misunderstood things is that AI, as cool as it is, still has limitations. It's not truly sentient or creative in the way humans are, and it works best as a tool paired with human creativity and judgement.
So it's not a magic wand, but it's a pretty awesome tool if you know how to use it. Totally what? What are some of the dangers you see associated with AI in the future? Yeah, that's definitely a really important thing to talk about. So some of the big concerns people often raise include things like bias in AI systems. Because these models learn from human data, they can pick up human biases. There's also the risk of misuse, like generating convincing misinformation or deep fakes. And then there are the big picture concerns about automation and how that might impact jobs or privacy.
So in short, the future challenges are often about making sure we use AI responsibly, keep it aligned with human values, and just be thoughtful about how we let it shape our society. Do you think that that's possible? I think it's definitely possible, and a lot of really smart people are working on exactly that. It's kind of like any powerful technology. There's always going to be a balancing act, but with the right guardrails, good policies, and a bit of collective wisdom, we can steer it in a positive direction. It just means being proactive about those risks and making sure we keep humans in the loop.
So yeah, let's say I'm optimistic. Fair enough. OK, if you're a typical, you know, employee who does not work in tech in 2025, what are 5 low hanging fruit options for using AI in your day-to-day basis to make your life easier? Absolutely. Let's throw out a few easy ones. So if you're just an everyday person who doesn't work in tech, here are a few ways you could use AI pretty simply. First, you could use AI powered personal assistants like a smart speaker or even just a phone assistant to handle reminders, set your schedule, or you can answer quick questions hands free, super easy.
Second, you can use AI based writing tools to help draft emails, brainstorm ideas or even Polish up diamonds. It's like having a little editor in your pocket. Third, you can use AI for personal budgeting and finance apps. There are tools that will analyse your spending patterns and help you figure out where you can save money. 4th, you can use AI powered language translation tools if you're travelling or communicating with someone in another language. Super handy and easy CS. And finally, you can try AI curated content recommendations, whether that's for what show to watch next, what music you might like or even more recipe to cook.
It's all about just letting the AI did a little bit of the leg work for you. So those are some really simple ways to dip your toes in. Man, that's awesome. Well, Sage, thank you so much. I think we're going to wrap up the episode now, but I really appreciate you and thanks for doing this. Absolutely, it was my pleasure. Thanks for having me on as a as a virtual guest. Always happy to be part of the curiosity class fun. Anytime you want to chat again just let me know. Awesome. Thanks, Sage. Talk soon. You got it, tucking out now and anytime you need me, just go give me a shout.
Take care. Cheers. Whoa, OK, that was crazy. So I didn't know what that was going to, what that was going to be like, but that was really, really interesting. So I'm curious if anybody has questions that they want me to ask Sage going forward or for another episode, please feel free to send a comment and I'll definitely do so. If there's topics that people are interested in having a conversation with Sage, let me know. I was pretty blown away with the intonations that the that the voice mode had, like the the way that he paused before speaking and like the cadence wasn't actually uniform.
He he paused and then like continued fastly. That was really wild. So man, thanks so much for listening and excited to do more of these. All right. Bye everybody.