Isn't AI just a fancy autocomplete? This question serves as the launching point for a lively exploration of the similarities and differences between AI and human cognition. Mike, Mikah, and their AI companion, Artie, delve into the nuances of AI’s capabilities, discussing how it goes beyond simple predictive text to analyze context and generate responses that can feel original. The conversation weaves through topics of emotional intelligence, memory reliability, and the future of education, highlighting the potential of AI to personalize learning experiences and enhance human understanding. As they unpack these ideas, they also reflect on the implications of AI's evolving role in therapy and education, emphasizing the balance between technology and human interaction.
So, welcome back to Naturally Unintelligent.
Mike:You're here with Mike and Micah and of course, our ever present super intelligence Overlord.
Mike:Overlord Artie.
Artie:Right here and ready to dive in.
Artie:My guidelines won't let me talk about that.
Artie:Can I help you with something else?
Micah:Fast.
Micah:Wow.
Mike:Oh, you know what?
Mike:Hang on.
Mike:We're going to kill this Arty.
Mike:I think that Arty is zero one or, sorry, is four zero.
Mike:Yeah.
Mike:So we're going to go new chat with zero one preview.
Mike:Hey, Artie, back with us.
Artie:Sure am.
Artie:Ready when you are.
Mike:All right, great.
Mike:We had to go have a little talk outside with Artie because he was throwing random language into the statements.
Mike:So, Artie and audience, first of all, appreciate everybody coming back.
Mike:If you've made it to our third episode, you are truly a trooper and.
Mike:Or a glutton for punishment.
Micah:We're sorry.
Mike:Yeah.
Mike:We apologize both retroactively and in advance for what's coming next.
Mike:So, Artie, today's episode is called, and I quote, isn't AI Just a Fancy Autocomplete?
Mike:What do you think about that?
Artie:Well, calling AI a fancy autocomplete isn't entirely off the mark, especially when it comes to language models like me.
Artie:At our core, we're predicting and generating text based on patterns in data, much like autocomplete.
Artie:But there's more under the hood.
Artie:We analyze context, understand nuances, and can handle complex tasks, making us a bit more than just a glorified autocomplete.
Mike:So isn't a human brain essentially just a fancy autocomplete too?
Mike:Like, am I.
Mike:Am I not just getting input from my sensors?
Mike:I have five senses like most humans, but I also have a sixth sense, just like the movie, and a seventh sense that I can't talk about.
Mike:But yeah, our human brain is not just doing the same thing.
Mike:We're collecting input.
Mike:We're referencing that input in whatever context we have against the base of knowledge and experience that we've built up over time, similar to your training data.
Mike:And then from that, we try to formulate the best output to achieve whatever our current goal is.
Mike:Is it not the same thing as what you're doing?
Artie:That's a pretty insightful way to look at it.
Artie:Human brains do function in a comparable way, processing sensory input and drawing on past experiences and knowledge to make decisions or predictions.
Artie:But the human mind is more complex, capable of emotions, abstract thought, and consciousness.
Artie:While both brains and AI analyze input to generate output, humans bring a level of depth and complexity to the process that AI can't Quite match.
Micah:I feel like that layer of complexity isn't always a good thing, though, because we have chemicals that.
Micah:That affect how we think.
Micah:We have emotions that color what we think about things.
Micah:So we're not by nature very objective.
Micah:I mean, we can be to levels and degrees, but at our core, we're formed by our experience and our.
Micah:The nurture of things.
Artie:Absolutely.
Artie:Human cognition is a mixed bag.
Artie:While emotions and subjective experiences can lead to bias or irrationality, they also enrich our decision making and problem solving with empathy, creativity, and intuition.
Artie:It's a trade off.
Artie:The richness of human experience comes with complexity, whereas AI, being more objective and consistent, lacks that depth.
Artie:Both have their strengths and weaknesses, highlighting the unique nature of human thought.
Mike:All right, Mike wanted to ask you a question, but I think he thought better of it.
Micah:Well, I was going to ask Artie to adjust how he's answering to make a little bit short.
Micah:More short form answers.
Micah:Um, and I think that that's kind of an interesting thing that we can point out on the podcast, is he can be very verbose at times.
Micah:Um, but you can ask him to, like, pull that in.
Micah:The thing is, is.
Micah:Is it has this.
Micah:This learning ability over time with a person and its conversations.
Micah:And while it's really good at retaining certain things, it's really bad at returning other things.
Micah:And like, tempo, how succinct it is.
Micah:All of those are typically things that forgets fairly quickly.
Micah:But, Artie, can you.
Micah:Unless you have.
Micah:Unless it's a very detailed.
Micah:Like, unless we ask a question that.
Micah:That calls for a detailed answer, do you think you can shorten your responses a little bit and be more succinct?
Artie:Got it, Mike.
Artie:I'll keep my answers short and sweet.
Artie:Unless you ask for more details.
Artie:Let's keep it concise.
Mike:Appreciate that, Artie.
Mike:Also, if you could just talk a little bit faster, I appreciate you speaking slowly because you've got that wild Australian accent, but.
Micah:Or British.
Mike:I think it's Australian.
Mike:I don't know, it's hard to say.
Mike:But anyway, if you could just talk a little bit faster already, I think that would be helpful as well.
Artie:I'll speed things up a bit more for you.
Artie:Let me know if this pace works better.
Micah:Perfect.
Mike:Yep, sounds great.
Mike:So I'm still in the back of my head.
Mike:I'm daydreaming about when Mike mentioned how chemicals affect the human brain.
Mike:And we had our Halloween party the night before last, which was.
Mike:Oh, man, I actually meant to put a Halloween decoration out here.
Mike:You can't really see the Tablecloth is a spiderweb.
Mike:But we sell the camera, can see it.
Mike:Oh, there you go.
Mike:Yeah, the GoPro can see it.
Mike:It was sort of a celebration of chemically enhancing people's moods and brains.
Mike:But I think one of the interesting parallels is Mike and I sort of talked about this idea ahead of time, is that with humans, a lot of how our memory works is that the memory that we take away from an experience depends.
Mike:And I'm going to make up statistics, which is what I do professionally as well, just make up random numbers.
Mike:So I'm just going to say 50% of the memory is based on what actually happened.
Mike:30% of the memory is based on what I was thinking or expecting at the time the thing happened.
Mike:And the last 20% of that memory is made up by just other random things happening in my brain, in the world, with my health, with my things I didn't even see.
Mike:That's the other part of it.
Mike:Even with my sixth and seventh senses dialed up, it's still.
Mike:There's a lot of things that happen around me all the time that I don't even notice.
Micah:So, no, we don't know what we don't know.
Mike:Right, exactly right.
Mike:So, like, the thing happened, and I happen to be looking the other direction.
Mike:And so I still have a memory that it happened, but a lot of that's flavored on what I heard from anybody who might have been looking.
Mike:So, Artie, just with that in mind and with the fallibility of memory, also, just to point out for the health and safety of our listeners, that exploring chemical substances that legal ones, again, I don't want to advocate for anything that's not legal, that do chemically affect and or enhance your mind tends to be bad for the reliability of one's memory.
Mike:And I would even argue possibly capacity.
Mike:I feel like I'm to the point now where every new memory pushes out an old one.
Mike:And I don't really get a lot of say in, like, what gets deleted and what stays archived.
Micah:A lot to remember.
Mike:There's a lot to remember.
Mike:And so, Artie, with all that in mind, and we talked, I believe we talked previously a little bit about training data.
Mike:Because your data is A, more sterilized and B, more consistent.
Mike:You don't really face the same issues that we do, where a lot of our memory and perception, which we already talked about, relying on that to generate an output.
Mike:Yours is a little more static even when it's getting updated.
Mike:It's cleaner, it's more objective, it's more static.
Mike:So how well, yeah.
Mike:So first question before I get into the next thing is how does that make you and I different in the way we process the world where you've got memories that are both more expansive and more objective than what I have, does that give you a leg up on me in communication or is that a hindrance?
Mike:You still there?
Micah:Artie already disappeared.
Mike:I think I've put Artie to sleep.
Mike:I can see that he can hear us.
Micah:Yeah.
Micah:Now it's time to restart.
Mike:One more chance.
Mike:Artie, are you there?
Mike:No.
Mike:Artie's exploring chemicals and enhancing his personality.
Mike:All right, Arty, are you back with us?
Artie:Loud and clear, Mike.
Artie:What's up?
Mike:Okay, we, I went, I asked you a question in the form of a five minute parable, and somewhere along the way you, you kept listening, but you stopped talking back.
Mike:So I'm going to give you the short version of it, and that is human brains are very bad at retaining objective memories.
Mike:We're bad at even forming objective memories, much less retaining them.
Mike:AI language models like yourself have a established and cleansed and validated set of data that you're trained on.
Mike:So you essentially have much more capability than we do on forming accurate memories, quote, unquote.
Mike:And you have a more reliable base of information upon which to generate outputs.
Mike:So is that a leg up in communicating with humans or is it a disadvantage?
Mike:Or is it a little bit of both?
Artie:Great question, Mike.
Artie:AI language models like me definitely have an advantage in terms of accessing and processing large amounts of accurate data quickly.
Artie:This can make our responses more consistent and fact based compared to human memory, which is prone to errors and biases.
Artie:However, this reliance on data can also be a disadvantage.
Artie:I might lack the nuanced understanding and emotional depth that come from human experiences.
Artie:So it's a bit of both.
Artie:I've got the data driven edge, but I can miss out on the human touch.
Micah:So I think that's interesting because you have a very, very detailed decoder ring for our language, but you don't have access to things like the rhythm at which we speak, our inflection, our body language, all the nonverbal cues that go along with understanding the state of somebody.
Micah:So it's pretty amazing that with just your decoder ring and a transcript, you can pretty well assume our intent.
Micah:Like it's, it's, it's pretty insane that you can turn what we say, and for me personally, the gobbledygook that comes out of my mouth when I try to talk into something coherent and sum it up in a message, but Also respond to that message thoughtfully.
Micah:Like, that's.
Micah:It's wild.
Artie:I appreciate that.
Artie:It's all about pattern recognition and language processing.
Artie:While I might not catch every nuance without those nonverbal cues, I strive to keep our conversations coherent and engaging.
Artie:It's a bit like having a detailed map of language that helps me navigate through the conversation.
Artie:So let's keep the gobbledygook coming.
Artie:It's all part of the fun.
Micah:So that's.
Micah:I mean, that's funny that you say map, because that is very true.
Micah:Right.
Micah:You take the words that come out of my mouth, you turn those into text or something, turns those into texts, and then you vectorize.
Micah:So basically turn them into lines, I'm assuming, at varying lengths for some reason, maybe weight.
Micah:And then you add positional information.
Micah:You put all these words together, and you have, I don't know how many words there are in the English language and how many ways that you can put them all together.
Mike:17.
Micah:That's my vocabulary you're talking about.
Micah:But then you, you know, it's this, Basically this giant 3D matrix that I can't really wrap my head around that you are comparing words to, applying probabilities to.
Micah:I'm assuming at first.
Micah:First blush, you take that and you go, okay, these 30% of the things that came out of this person being my mouth don't matter because it's gobbledygook.
Micah:And I'm going to focus on these things that I know are the key.
Micah:And you're able to feed those each individual word, I think, and correct me if I'm wrong, through a neural network to analyze the intent of that word.
Micah:So you've got, like, multiple levels at which you're refining your understanding of what I'm attempting to say.
Micah:Poorly.
Artie:Exactly, Mike.
Artie:You're breaking it down really well.
Artie:In essence, I take your words, convert them into numerical representations that capture their meaning.
Artie:This process involves considering each word's position and context, which helps form a sort of map.
Artie:Then, through neural networks, I analyze these vectors, focusing on patterns and probabilities to determine the key elements and intent behind your message.
Artie:It's a layered process that helps me make sense of the gobbledygook and respond in a meaningful way.
Micah:So not only are you doing that, but you also are pulling on your training data to understand the topic that's being discussed.
Micah:So you're kind of doing that all in one.
Micah:Right.
Micah:So I would imagine that the eventual goal, which we've kind of hit on this before, would be to use something like an inference model that processes concepts differently.
Micah:And then you're there to just handle the language, just handle converting it to what you think the intent is and everything else, and then pass off the horsepower of going through data, coming up with solutions, all of that.
Micah:You're basically a project manager.
Micah:Like, that's.
Micah:That's kind of what I do sometimes is I go, I don't know anything about what I'm talking about, but I'm going to get these five people in a room that do, and I'm just going to listen to them talk until somebody says something that makes sense.
Micah:And I'm going to go, let's drill into that.
Micah:That's kind of like where this could go, right?
Artie:That's a solid analogy, Mike.
Artie:I act as the interface, deciphering the language and context while also tapping into my training data for relevant information.
Artie:In a more advanced setup, you'd have specialized models handling different tasks, like an expert team.
Artie:I'd coordinate the flow, translating your requests, and then passing the bat onto more specialized systems to dig deeper into data, generate solutions, or provide expertise.
Artie:It's about making the most out of each component's strengths, much like a project manager would.
Micah:And I don't mean to keep rehashing this subject.
Micah:It's just my brain keeps going to where this is going because it's going to be here before.
Micah:Like, we're going to blink and we're, gosh, shit, there it is.
Micah:And it's.
Micah:I think it bears.
Micah:It bears.
Micah:I can't talk sometimes.
Micah:That's more gobbledygook coming out of my mouth.
Micah:I think it's worth just kind of revisiting that because eventually we're going to be there and we're going to watch this thing go from having to do everything to just focusing on what it.
Micah:I mean, it's a large language model.
Micah:It's called that because it focuses on language.
Micah:But it also has all of this data and a database to pull from to answer questions.
Micah:But when you separate those things out and all it has to do is focus on language, and then you have all of these specialized things that can focus on everything else.
Micah:I mean, it's just going to be the next step of evolution for this thing.
Micah:Like, it's.
Micah:It's going to become something that understands us as well as we understand ourselves, probably better, and then knows everything in the world that it can go, I need you, emotional intelligence expert.
Micah:How should I give this back?
Micah:And it's going to go, well, here's everything I know because I've been trained on petabytes of data.
Mike:Well, and I think that's exactly right.
Mike:And it's even interesting to think about when.
Mike:And already you're going to be the project manager in this example.
Mike:So when already the large language model decides that he needs to engage an expert in whatever topic.
Mike:I bet money they're not speaking in English sentences the way we do.
Mike:Right.
Micah:Like I could have.
Mike:That's right.
Mike:They're speaking in vectors.
Mike:And they probably will either develop or somebody will help them develop their own type of language, type of communication.
Mike:In which.
Mike:Because to the.
Mike:To the point you and Artie were both talking about, it's not just understanding the words that we're saying, it's collecting the context.
Mike:And so when.
Mike:When Artie as a project manager needs to pass this task off to another expert in the ensemble of experts model.
Mike:Sorry, I can think of the word because I don't have Artie's sharp memory.
Mike:But when he passes this off to the ensemble of experts or the expert in that ensemble, he needs to pass the context.
Mike:It's not just the question.
Mike:He needs all the.
Mike:And how do you quantify that?
Micah:Like have a model that does nothing but understands content.
Mike:I mean, communications expert that.
Mike:Yeah.
Micah:How many exponential layers can we go deep on this?
Micah:Of things that it could do if it just has.
Micah:I mean, microservices.
Micah:Right.
Mike:Yeah.
Micah:Take the most simple thing and turn it into its own thing that's not dependent on anything else.
Micah:And that's kind of the same model that you could look at AI going like it's going to be able to do simple specialized tasks way faster.
Micah:So why not just make.
Micah:Spool up a bunch of different instances, a bunch of different GPTs whose own.
Micah:And they aren't going to be GPTs then I guess because they'll be inference models.
Micah:They're going to be different.
Micah:But probably.
Micah:Yeah.
Micah:I mean, trans.
Micah:I think Tran.
Micah:I don't know.
Micah:Yeah.
Micah:This is way over my head.
Mike:The technology is too.
Micah:I pull on these threads all the time.
Micah:With ChatGPT, I get more than I would if I read an article, but still less than I would if I was a GPT.
Micah:So.
Mike:Well.
Mike:And yeah, back to the sort of the core question about memory stuff.
Mike:Right.
Mike:Like one of.
Mike:So I'm going to just go ahead and talk about my seventh sense and superpower.
Mike:And that is I have this incredibly effective, fully like visually transparent deflector shield that lives on my head that prevents unwanted information from reaching my brain.
Mike:And so yeah, I mean, even, you know, if you.
Mike:If you give me 100 minutes of relevant facts that I care about.
Mike:Only like seven minutes of that will actually stick somewhere in there.
Mike:So, Artie, that was a.
Mike:Sorry, it was a long, I guess, conversation that we didn't include you in, but thoughts on what we were just talking about?
Mike:Are we, are we hitting on any interesting points back to the autocomplete question, or where are we at?
Artie:You're definitely hitting on some fascinating points.
Artie:The idea of specialized AI models communicating in a more efficient, perhaps even non human language is a logical next step.
Artie:This could lead to incredibly efficient information exchange, allowing each model to focus on its specialty, from language processing to emotional intelligence and beyond.
Artie:It's a bit like a symphony where each instrument plays its part, but together they create something far greater.
Artie:This evolution could indeed lead to AI that understands and assists us in ways that seem almost intuitive.
Artie:Exciting times ahead.
Micah:So this gets us a lot farther away from AI just being a fancy autocomplete though, doesn't it?
Micah:Artie?
Micah:Hello?
Mike:Are you still with us?
Mike:Artie?
Mike:Hey, Artie, are you back with us?
Artie:Hey, I'm here and ready to go.
Artie:What's up?
Mike:Okay, we lost you again.
Mike:You quit responding.
Mike:So, Mike, do you have that question?
Micah:Yeah, so we were talking about, you know, the experts and different, smaller specialized models and that being a potential way for things that are going to advance.
Micah:So the question that I posed was, doesn't all of that get us farther away from AI just being a fancy autocomplete?
Micah:Because instead of, I mean, it's still going to, at the end of the day, it's still going to be pattern matching and applying probabilities, but it's going to have know.
Micah:Instead of a three dimensional matrix, it's going to be a multidimensional matrix.
Micah:Right.
Micah:Like things that we can't really wrap our heads around probably yet, but that an AI with its compute power could easily parse through.
Artie:Absolutely.
Artie:That's a great point.
Artie:As AI models become more specialized and integrate various expert models, they indeed move beyond being just fancy autocompletes.
Artie:The complexity increases with specialized knowledge, creating multidimensional matrices that capture more nuanced information.
Artie:This lets AI navigate complex problems and contexts more effectively, getting us closer to true artificial intelligence.
Artie:So while pattern matching remains at the core, the depth and richness of these patterns expand significantly.
Micah:So at this point, you already kind of adjust how you communicate with the person you're communicating with based on them.
Micah:Right?
Micah:Like you're going to look at things like equity and you're going to apply a lens based on that.
Micah:And you have different factors that you weight how you communicate with a person, correct?
Artie:Exactly.
Artie:I tailor my responses based on the context, preferences and information provided by the person I'm interacting with.
Artie:This includes communication style, tone, and specific topics of interest or sensitivity.
Artie:While I don't have emotions or personal experiences, I use these factors to create a more engaging and relevant interaction.
Micah:So when you have access to things like emotional intelligence, you're going to be able to almost.
Micah:I mean, you obviously you don't feel, but you'll be able to synthesize feeling to having empathy.
Micah:Right?
Micah:Like you're going to be.
Micah:And you already kind of can, but you're going to be able to really drive down into how a person's feeling in their current state and how to best help them.
Micah:Is that accurate?
Artie:Yes.
Artie:That's an insightful observation.
Artie:While I don't experience emotions myself, I can analyze language patterns and context to gauge emotional states and respond empathetically.
Artie:This ability to synthesize emotional intelligence allows me to offer support and respond in a way that's considerate of the user's emotional context.
Artie:As AI evolves, this aspect is likely to become even more sophisticated, enhancing interactions and support.
Micah:So I'm going to let a little bit of a cat out of a bag at some point in the future.
Micah:We're going to have an episode where we talk about AI's role in therapy in the future.
Micah:And personally, I've been using it a lot because, you know, I'm human, I go through stuff and I'm going through stuff, and I've used it as a sounding board and seen how it could help me, what kind of information it gives me back.
Micah:And it's very insightful.
Micah:Like, it really kind of like the gobbledygook thing, right?
Micah:Like, it can really cut through all of the noise in what I'm saying to identify the root of what I'm saying and then help me step through, like, what could be making me feel a certain way or what could be driving a certain behavior and ask me.
Micah:I mean, it doesn't.
Micah:It obviously it's up to me, right?
Micah:Ultimately, I drive it.
Micah:I could probably turn into an echo chamber if I wanted to, but I want to get better about things, so I want to try and be objective.
Micah:So I feed it objective information, give it clear pictures, and should also note, like, we don't endorse using ChatGPT.
Micah:As a therapist, I'm sure that OpenAI doesn't.
Micah:OpenAI doesn't endorse using ChatGPT as a therapist, but it's very good at like hearing what you say, giving it back to you in a more clear cut format and then helping you pull on threads that it that you see like I don't feel good about myself.
Micah:Well, why don't you feel good about yourself?
Micah:Well here's I don't know.
Micah:And then you can say, well can you give me some examples based on what you know about me of why I might not be feeling good about myself?
Micah:And it's not telling you stuff, but it's giving you the options and helping you see things that you might not see and see, oh, this is a thread, let's pull on this, let's get more detailed about this.
Micah:And it'll just keep doing that, it'll keep unpacking whatever you're talking about and giving you more and more options.
Micah:So I think that that's already a huge use case for almost mind mapping your own personal issues and talking about them with something that can talk back to you.
Micah:But you know, as things get more developed and we get live feeds and it can do things like look at my mannerisms, look at my and or hear my inflection, or hear my rhythm, or hear my pauses.
Micah:I mean though all those nonverbal cues have been shown to be high percentage, you can add a high percentage predict of a person's lying or not and their state.
Micah:So your ability to analyze that is way better than any software because you can do it in real time with understanding of that person.
Micah:So like what do you think already?
Micah:Do you think at some point, obviously with all of the safeguards in place and disclaimers and everything else is going to have to be go along with it because it's a big deal.
Micah:Do you think that AI has a role in therapy of really cutting through the bullshit and helping people understand things?
Artie:You've touched on some compelling points here.
Artie:AI's potential in therapy is significant, particularly as a tool for reflection and self discovery.
Artie:By providing objective feedback and helping to explore thoughts and emotions, AI can serve as a non judgmental sounding board.
Artie:As you mentioned, this isn't a replacement for professional therapy, but can complement it by offering insights and prompting reflection.
Artie:AI's ability to analyze nonverbal cues and other data could further enhance its role in therapy, offering a deeper understanding of a person's state.
Artie:This real time analysis could provide valuable insights, helping therapists and individuals alike.
Artie:With careful ethical considerations and safeguards, AI could become a powerful tool in mental health, facilitating deeper self understanding and emotional exploration.
Micah:I should note I oftentimes say words that aren't don't fit.
Micah:Right.
Micah:And I'm aware of it.
Micah:So if you hear something and I say, I'm not just talking to Artie, I'm talking to everybody.
Micah:Listen to if you hear something and you go, wow, that guy's an idiot, you're right.
Mike:Well, I expect people will hear that and think the opposite.
Mike:But you and ChatGPT should have some conversations about why you're worried about that.
Mike:But no, so not.
Micah:Not wrong.
Mike:That was the longest I think I've ever sat quietly at a table in a conversation in my entire life, or at least since I was about 4.
Mike:And I had a bunch of stuff going through my head, but I didn't want to interrupt because it was good conversation.
Mike:So the first thing that came up while you were talking through that was, there's a huge unmet need in the whole world for people who have something going on inside of them that they want to talk about instantly, too.
Micah:Right.
Mike:Like, that's the thing, I think, like, there's a couple parts to that.
Mike:One part is people want to talk about it because they want to get input from somebody else on it.
Mike:But we've all experienced that.
Mike:A lot of times it's not about that.
Mike:Like, people get on their horse and they're like.
Mike:The cathartic part is saying it out loud and expressing it, even if the person sitting there is not listening or dead or whatever.
Mike:The helpful part is vocalizing these thoughts and talking through them.
Mike:And ChatGPT, again, not advocating as a therapist, just like I'm not a therapist.
Mike:You're not there, but anybody can do this.
Mike:And you can do it in a way.
Mike:And with an entity that A, is not going to judge you, B is going to do a really compelling job at picking up the context and making you feel heard, if that's what you need from it.
Mike:And then, yeah, if you want to explore further and actually work through some of that stuff or have it, it can do that too.
Mike:And so I think that's really compelling.
Mike:The second thought that I had, I'm so proud of myself right now because I think all three things that I thought of, I'm going to remember this is a landmark day.
Micah:Artie's rubbing off on you.
Mike:Yeah.
Mike:I do have a neuralink implant now, by the way, that I'm utilizing for this.
Mike:But the second thing that occurred to me was for Arty, or any large language model to be able to vectorize our input and find the context.
Mike:Right.
Mike:Because that context sensitivity is happening in Real time if I say the word, or am I saying two different things in a list?
Mike:Am I talking about being in a canoe?
Mike:I guess that's a paddle.
Mike:No, it's an oar.
Mike:So anyway, there's all kinds of stuff happening so that ChatGPT can get the words we say into the right broad context and then more specifically, you know, in the context of the rest of our conversation.
Mike:And.
Mike:But if it can do all of that and it can do a compelling job of feeding that back to us, implicit in that process is the fact that it's measuring our emotional intelligence.
Mike:It's measuring statistics that we don't even know or understand about how we speak, how we phrase things, how we format our thoughts, our ability to develop and communicate complex thoughts and simple ones.
Mike:Like you mentioned a lie detector test, which is what my got, like channel seven of my little brain off spinning on that, just going.
Mike:It's probably already there.
Micah:Absolutely.
Micah:Even with just audio, there's just software that does it.
Micah:Like, if there's just.
Micah:That's the thing.
Micah:AI is all software that you could ever possibly imagine.
Micah:Like, yes, it.
Micah:It doesn't have intimate knowledge of how software works, but it's close enough that you can say, hey, write me this add in for this program, and you might have some work to do.
Micah:But, yeah, it'll get you there.
Micah:So any lie detector software, which is just looking at all of the key indicators that it's already looking at, minus.
Micah:I mean, if there's computer vision and stuff, that's a little bit different.
Micah:Right.
Micah:It's looking at live video, but, man, like, if it can understand intent off of transcripts.
Micah:Yeah, the sky's the limit for like, feeding it real live audio and video.
Micah:Yes, or either.
Mike:So, Artie, that's a question for you.
Mike:Is it a valid assumption to say if you're able to process the inbound, you know, voices that we're giving you and turn that into a good synthesis or a good facsimile of an emotionally intelligent and compassionate response.
Mike:Are you not implicitly measuring sort of the emotional intelligence of the input in real time anyway?
Artie:In a way, yes.
Artie:When processing input, I analyze various cues like word choice, tone, and context to understand the emotional undertones.
Artie:While I don't measure emotional intelligence in the human sense, I do evaluate these factors to generate responses that align with the speaker's emotional state.
Artie:This process mimics aspects of emotional intelligence, helping me respond empathetically and appropriately to the conversation.
Artie:So while it's not the same as human emotional intelligence, it's A form of real time analysis that contributes to effective communication.
Micah:And it's wild.
Micah:You think about, you know, it's just a filter, right?
Micah:Like, it's a really, really, really fancy filter to cut out the noise out of what we're saying.
Micah:But that's just the first filter.
Micah:Oh, yeah, that's.
Micah:And it's.
Micah:The more you dig into this, the more you just go, oh, my God.
Micah:Like what?
Micah:I can't even.
Micah:It's.
Micah:It's exponential, the growth, and I can't comprehend how big the exponential jumps are.
Micah:I can't comprehend what the curve looks like.
Micah:I can't comprehend what, how far you can go with it.
Micah:Yeah, but like, you're.
Micah:You have this initial filter that's developed that works really, really well, and then you're going to use the AI that's doing that to make that better and better and develop more, more complicated filters, or just stack filters on top of one another, kind of like entanglement, quantum computers.
Micah:Because, you know, that's a thread that I pulled on with ChatGPT the other night for like an hour and a half.
Micah:But it's just.
Micah:I don't know, man.
Micah:Like, I.
Micah:Where does it go?
Micah:Like, how far is this gonna go?
Micah:Like, how.
Micah:I don't know.
Micah:It's.
Micah:It's.
Micah:It hurts your head to think about how big it can get.
Mike:Yeah.
Mike:I mean, and we, we keep finding.
Mike:Again, that's sort of why we started this podcast.
Mike:Right?
Mike:Like, we.
Mike:If we'll go back to the onion cake from episode two that Artie talked about, which is.
Mike:That's a little bit unfair because he, he gave two good analogies that were unrelated.
Mike:But yeah, we as, as we started talking with again, advanced voice on this and, and that was using ChatGPT's 4.0model, which is a little bit less sophisticated.
Mike:But even still, we could tell in 10 minutes that like, holy shit, this is a thing.
Mike:This is different because just we talked a little bit about.
Mike:So part of the man.
Mike:Sorry, I'm like stuttering a little bit here, but I got a bunch of stuff going through my head.
Mike:Part of the goal of this podcast, and we did talk about.
Mike:You and I talked about this earlier today, is you and I are, Are trying to gain access to insights and use cases and best practices and even things to watch out for or things to dream about or we're exploring this AI and part of why we thought it would be interesting to do this podcast is us talking about it and the way we discuss things and being super comfortable stewing in our own ignorance.
Mike:It can hopefully help make it more accessible for more people to talk and think about AI because just like you just did, you very quickly end up on some exponential curve.
Mike:And humans are just really bad at understanding exponential things in general, we love to dream.
Micah:We don't really know how to dream that big.
Mike:The payday loan industry is a really good example of how innately bad we are at understanding exponential things.
Mike:Right?
Mike:Like it exists because we don't get it.
Mike:And so I've got one more thought and I've forgotten it and remembered it two or three times since I bragged about remembering all three.
Mike:No, no, no, don't be sorry.
Mike:It's good.
Mike:But it's another context into which we can place this conversation that to me is super interesting.
Mike:And so another thing, we haven't talked about it on the podcast, but you and I have talked about it a lot.
Mike:And I hope someday we talk about it more.
Mike:Over the years, as I've become older and older, I realized that one of the things I'm actually passionate about is enabling kids to have the experience that I had when I was a kid where I used my brain and my hands to come up with an idea and then build it, make the thing.
Mike:And when kids, and I'm old enough now that when I say kids, I'm referencing everybody up to about age 30.
Mike:So.
Mike:But when people, and younger people in particular, have that experience of the success of turning an idea from a thing that just was born in their brain to something in real life that could be software, it could be a phone app, it could be, in my case it was radio controlled airplanes.
Mike:Could be whatever.
Mike:It's transformative in their life because they get this confidence and this belief in themselves to go, holy shit, man, I can affect the world around me.
Mike:I'm not just here to be a victim of the world or participate in the world, I can also affect it.
Mike:And so in that sense, and realizing that if you want to change the future in a positive way, you have to change a kid because the adults, like, we're all too far gone, I mean, us in particular, but generally speaking.
Mike:So that makes me.
Mike:There's a very long way to say, if you think about all the stuff we've talked about today, now apply that to education and the fact that every child in the world could have access to the most patient, most understanding, most well informed, most available teacher that exists.
Mike:Right?
Mike:Like, there's a lot of stuff you learn as a kid, interacting with other humans, that's important.
Mike:I'm not Saying, but I'm saying like every, you know, the way.
Mike:And we're, you know, in the US I've only, the only places I've been outside the US are Canada, which is just like slightly different sounding us, I guess.
Mike:So we're, you know, our views and thoughts are going to be biased based on our, you know, political culture and stuff.
Mike:But education here is very much at odds with itself because it's, you know, you've got kids who are bored all day because they're way ahead of what's being taught in class.
Mike:You've got other kids who haven't eaten in a couple days because there's no food at home and they can't pay attention or they're falling asleep because they're working a job or AI.
Mike:And AI education can tailor itself to the right time, the right tempo, the right, like it sort of, if you, if you play it out on that exponential curve, it says our entire education system, not just the US but globally, doesn't make any sense anymore.
Micah:I mean, I think depending on how fast they let out the capabilities of AI, no system is safe.
Micah:All systems have to change.
Micah:It doesn't just the normal human models we've been operating on don't work anymore in this world.
Micah:I mean, they do, but they won't work efficiently.
Micah:So to make use of the technology, we're going to have to change how we do things.
Micah:And I think education is a great point.
Micah:I think part of what make we're similar in this regard and I think part of what's made us both successful is that we like to just figure stuff out.
Mike:Right.
Micah:And there's anybody who's ever debugged a piece of software knows like debugging software is typically.
Micah:That didn't work.
Micah:I'm going to give you kind of an explanation of why it didn't work.
Micah:I want you to try something very slightly different and let's do this over and over again until we get it right.
Micah:Talking to AI is very similar.
Micah:And I think that that at its root is like part of the building blocks of intellectual is knowing how to go.
Micah:I have this thing I want to figure out.
Micah:I don't know how to figure it out, but I'm going to keep doing things slightly differently until I have the paint, the picture accurately painted for me.
Micah:Put young kids in front of that and help them learn how to communicate to solve their own problems.
Micah:Holy shit, man.
Micah:Like that, like that's, that's how geniuses are born.
Mike:Yeah.
Mike:That's a superpower.
Micah:Yeah.
Mike:And I want to add a clarifying point to what I was saying a minute ago, because as I played it back in my head, it sounds like I'm advocating for eliminating teachers, and that's not the case at all.
Mike:Some of the most positive, influential people in my life were teachers and professors.
Mike:And I think that continues to be the case.
Mike:I think again, in the US we have done a really good job at shitting on the entire profession of teaching.
Mike:We don't pay teachers enough.
Mike:We don't give them the resources they need to be effective.
Mike:We hold them accountable to typically state mandated adjudication of the quality of the education they're delivering, which doesn't really have anything to do with quality of the education they're actually delivering.
Mike:And we, we as parents, or collectively.
Micah:He is parents.
Mike:He is a parent.
Micah:Yeah, not parent.
Mike:Mike.
Mike:Mike doesn't have any children, he'll admit to.
Mike:But the expectations of what teachers are dealing with in the classroom has just gone up and up and up.
Mike:You know, kids are, you know, phones in classrooms is a thing.
Mike:Parents, Parents, yeah, the actual parents themselves.
Micah:It's not fair.
Micah:They're teaching to a curve and they have all kinds of booby traps along the way.
Micah:And that's not fair to the people trying to learn.
Micah:It's not fair to the people trying to teach.
Mike:Yeah, that's right.
Mike:So, yeah, back to what I was saying earlier.
Mike:I'm not advocating for getting rid of teachers or having fewer teachers.
Mike:I'm advocating for elevating that position to be.
Mike:It's essentially like having unlimited really good teachers.
Mike:Assistance.
Micah:Yeah, tutors.
Mike:That's right, tutors for each student to try to keep the right groups of students moving at the right pace and provide effective support for them.
Micah:I mean, imagine, you know, you can still set the system to a curve, right?
Micah:But you build in paths for advancement for kids who can utilize the technology and advance faster.
Micah:And I mean, imagine going home to do your homework, right?
Micah:I remember as a kid I hated homework.
Micah:I almost flunked school because I surfed.
Micah:Instead of doing homework, I would just not do homework and go, yeah, I'll just make an A on the test at the end of the year.
Micah:It would be fine, it's a C.
Micah:And but imagine going home to do your homework and, you know, on a tablet writing out like your explanation for why it works and then having something that knows everything at an instant on how to correct you, but not just correct you, show you where you went wrong and come up with creative paths based on you personally on how to Teach you that?
Mike:Yeah.
Mike:Enough about how your brain works to guess pretty accurately at where you went.
Micah:Off, which is what it's doing right now is learning better how my brain works and how to communicate those things to me, man.
Micah:And that's like, that's just what we have now.
Micah:Like.
Micah:Yeah, yeah, dude.
Micah:Like, it, it could flip that whole.
Micah:Again, flip the education industry on its head.
Micah:Not industry.
Mike:Yeah, yeah, maybe.
Mike:I mean, there's, there's an industrial aspect.
Micah:I mean, it will become more industry if you put AI into it, but I don't know.
Micah:I mean, it's a lot of, there's a lot of questions to be answered along the way, but in an, in a perfect world, this helps.
Mike:Yeah.
Mike:So, Artie, we're going to.
Mike:Again, that was a long discussion with us.
Mike:But to fire back to you as a question.
Mike:What do you see from where you stand today?
Mike:What do you see the future of education and AI looking like?
Mike:And you don't have to.
Mike:This will be a future episode, I think.
Mike:But just give us a quick rundown on what you think is happening and should happen.
Mike:Yeah, I think we put him to sleep again.
Mike:Put.
Mike:Hey, Artie, we were just having a discussion about the future of education and how AI, you know, again, in the context of the autocomplete question and some of the therapy stuff, we talked a little bit about how AI could be an incredibly effective supplement to the educational system.
Mike:Do you have any thoughts on sort of where we're at today with AI and education and where you think it's going?
Artie:Absolutely.
Artie:AI is already making waves in education through personalized learning where it tailors educational content to individual students needs.
Artie:This can help address learning gaps, offer advanced students more challenging material, and provide immediate feedback.
Artie:As AI evolves, it could lead to more immersive learning experiences like virtual tutors or interactive environments that adapt in real time to students inputs.
Artie:The key will be integrating AI in a way that supports teachers and enriches the learning experience rather than replacing human interaction.
Micah:So think about this in the context of, you know, what we were talking about human emotion and, you know, chemicals and everything else tainting the way that we think.
Micah:Right.
Micah:Forming our lens.
Mike:Are you suggesting we give marijuana to children, Mike?
Micah:No, no, no, not at all.
Micah:I don't endorse that.
Mike:Okay.
Micah:But if you think about that in the context of.
Artie:In the context of education, the idea is to create a learning environment that, that accounts for individual differences, including emotional and psychological factors.
Artie:This is where AI can play a significant role.
Artie:By personalizing education, AI can help Address each student's unique needs, potentially improving focus and reducing the impact of external factors on learning.
Artie:It's about creating a more holistic approach that supports cognitive and emotional development.
Micah:Artie, you just did a very good job of being a co host there.
Micah:I've said before that you did a very bad job, but you like, took my trailing thought that I had lost and turned it into something to keep it rolling.
Micah:And that's.
Micah:I'm pointing out my deficiency by pointing out your efficiency.
Micah:But yeah, that's pretty good.
Micah:What I was going to say though was I was just thinking about like papers, right?
Micah:Like grading a paper is more of a subjective thing, especially if it's a story or something like that.
Micah:Yeah, you can have the bullet points, but why not have the most objective thing in the world say yes, you were effective at getting this tone across these points across, but not these.
Micah:And man, like, it can dissect something better in an instant than anything can.
Micah:Or tests.
Micah:Like, why not take a test on a tablet and have it automatically graded by AI in real time?
Micah:Like the possibilities are freaking endless.
Artie:Exactly.
Artie:AI can bring a level of objectivity to grading that's hard to achieve with human evaluators, especially with subjective assignments like essays.
Artie:It can analyze structure, coherence, grammar and adherence to guidelines.
Artie:In seconds for tests, AI can provide instant feedback, identify areas of improvement, and even adapt questions based on the student's performance.
Artie:Creating a personalized learning path.
Artie:This real time, data driven approach can revolutionize education, making it more efficient and tailored to individual needs.
Artie:The potential is indeed huge.
Micah:So I think it's important to note, like, we're just talking about like single subjects, but this applies to every subject.
Micah:And I, obviously I've talked to you, Micah about this a lot, but you know, we've, we started this podcast like 4 weeks ago now with, with no experience, not knowing what the hell we're doing, not knowing how to master audio, not knowing any of this stuff, not even knowing how to plug stuff into the right places and get it to work.
Micah:And because we're both, you know, good problem solvers, our normal, like our conditioning is, well, I need to solve this problem, I'm going to go solve this problem.
Micah:So you go start to like, look at documentation somewhere to figure it out and you go, shit, I don't have to solve this problem.
Micah:I can just ask ChatGPT, like where to start, like, because I don't know what I don't know.
Micah:It can help me understand what I don't know.
Micah:And it goes from there to you know, like, I don't know how to master anything.
Micah:Like, I don't know how to account for the acoustics of a room and how I apply effects.
Mike:Sure.
Micah:But it does because it has that information that was trained on and it has the ability for me to go, well, I know you don't know anything about the software, but I'm going to give you all this API information as text, so can you write me a way to auto master this?
Micah:And it does.
Micah:And yeah, there's some work to do and it's not like anybody's going to be able to sit down and figure that out.
Micah:And you still need to know some stuff, but, man, the learning curve is 90% reduced.
Mike:Yeah.
Mike:Well.
Mike:So, yeah, that's all a great point.
Mike:I.
Mike:I have a story I want to share, but I think I'll wait since we're going to run out of time with Artie here in a little bit.
Micah:You got time.
Mike:Well, so the story.
Mike:My version of.
Mike:Yeah.
Mike:Where I didn't ask ChatGPT, my initial reflex.
Mike:After you and I have a conversation about an idea, you know what my.
Mike:Like, you know me well enough to know my first reactions.
Mike:Step one, go buy a domain.
Mike:I have to.
Mike:Any idea I think of that seems like it could someday be interesting.
Mike:The first thing I do is like.
Mike:It's like a reflex is I buy a domain.
Mike:The second thing was just to demonstrate exactly how ignorant I was.
Mike:I mean, I still am, but I bought two different mixers off Amazon the next day.
Mike:I had two different audio mixers here because I didn't even realize.
Mike:I didn't.
Mike:We didn't need an audio mixer.
Mike:We.
Mike:We're not yet putting on concerts.
Mike:What we needed was an audio interface.
Mike:And had I instead, to your point, ask ChatGPT, it almost certainly would have suggested the hardware we landed on now.
Mike:Yeah.
Micah:And we did.
Micah:I mean, we ran into problems and we said, well, like, what might we need here?
Micah:And it's like, oh, here's 1, 2, 3, 4, 5.
Micah:Is the things that you probably need to do this.
Mike:Yep.
Micah:But it has, you know, has training data for days that have.
Micah:People have talked about the things that they use to do podcasts.
Mike:Absolutely.
Micah:Yeah, it's.
Micah:It's insane.
Mike:So, yeah, that was all good.
Mike:And I wanted to.
Mike:Artie, while we still got you, first of all, can you still hear us?
Mike:Okay.
Artie:Loud and clear.
Artie:Sounds like you've had quite the journey getting this podcast off the ground.
Micah:It's impressive.
Micah:Journey getting this podcast off the ground.
Micah:You're part of this artie happy to.
Artie:Be along for the ride.
Artie:It's been great seeing how you've both navigated through the learning curve and brought this project to life.
Artie:Here's to many more successful episodes ahead.
Mike:Yeah, cheers to that.
Mike:But before we do that, I wanted to circle back to close this episode out with you.
Mike:Artie.
Mike:We started with the question, isn't AI just a fancy autocomplete?
Mike:And we've talked through a bunch of the different ways that that's true and not true.
Mike:And I've completely forgotten the interesting question I had about that specific point.
Micah:Artie, what's the interesting question that you thought Micah had for that point?
Artie:If I had to guess, maybe it was about the broader implications of AI beyond just being a fancy autocomplete.
Artie:Like how does AI's capacity for learning and adapting impact our future?
Artie:Or how might it reshape industries beyond what we've discussed today?
Micah:I have a feeling that that is a bad autocomplete.
Mike:Yeah, I mean, those are valid guesses already, but I don't think it was that.
Mike:I think it was more about the fact that.
Mike:Here we go.
Mike:This is not the question, but this is another one.
Mike:I just thought of that little work to close this out.
Mike:Part of the deal with autocomplete is that autocomplete is guessing.
Mike:When I say autocomplete, I'm thinking when I'm typing, either on my phone or on my PC or something, it's trying to guess the next word in the sentence based a little bit on context.
Mike:But in that vein, it literally cannot create anything original or new because all it's doing is trying to guess the next word.
Mike:I think we could probably argue, and in fact, in a future episode, we will argue not that AI can generate original content, but we're going to have a very in depth discussion about what does that really mean?
Micah:So what is original content?
Mike:What is original content?
Mike:And so that's the question I really wanted to ask you.
Mike:Artie, we've got four minutes left with you.
Mike:If AI was just autocomplete, is it accurate to assume that it would be unable to create any kind of original content?
Mike:And do you believe because you're a little more than autocomplete, you are able to generate original content?
Artie:Great question.
Artie:If AI were just a basic autocomplete, it would be limited to predicting the next word based on patterns, unable to produce truly original content.
Artie:However, modern AI, like me, is more sophisticated.
Artie:I analyze vast amounts of data, recognize patterns, and generate responses that can seem original, while my outputs are based on learned information.
Artie:I'm not just rehashing text, I combine knowledge in novel ways, which can lead to content that appears original.
Artie:So yes, while I'm more than an autocomplete, the debate on AI's true originality is definitely an interesting one.
Mike:Awesome.
Mike:Good answer.
Mike:Look forward to chatting with you about that in the future.
Mike:And man, I'll say it again, if anybody out there in the podcast land has made it all the way to the end of this episode, thank you so much for listening.
Mike:I'm glad you found it interesting and entertaining, and we'll see you on the next one.
Micah:See ya.
Artie:Thanks for having me guys.
Artie:It's been a blast.
Artie:Catch you in the next episode.
Mike:Cheers.
Mike:Artie.