You Still Need to Be Smart

Chris Davis 0:10
And we’re back. Welcome to another episode of the All Systems Go podcast. As, as, as I usually do, I like to start out with a story. I love these stories because it humanizes.

Chris Davis 0:22
Listen, I’m speaking from real life things that are happening right now, right? And I got a message from somebody in, on social media and this person was trying to build an app and they were just running into so many complications and roadblocks and they joined all the, the communities that taught you how to do it. They, they’re watching the YouTubers, they’re, they, they took courses, which is all the good stuff because all of that is what primed you really for somebody like me, right? At least you’re, you’re learning the basics, you’re in the game, you’ve, you’ve got something going.

Chris Davis 1:03
However, for whatever reason, this person just couldn’t get to the finish line, right? And as we continue to discuss in, in, in private messaging or dming, I guess I started to see, I started to see it clearly, very clearly. And this is not down talking anybody, but it comes from the, the, that’s where the title of this episode comes from. You, you still need to be smart.

Chris Davis 1:30
And what I mean is you cannot one, one, you can’t blindly. And I’m not saying this person did it. I, I use that story as a catalyst for, for the rest of this podcast. So that was the story hit me up, did all that they could, got stuck and I identified what need.

Chris Davis 1:49
I identified a trend because, because I never talk about something on the podcast where just one person does it. I wait until I see a trend. So by the time I say it, I may be using one person, but I’m talking about a group of people and the trend amongst people. And perhaps this is yourself when I say you still need to be smart.

Chris Davis 2:07
A lot of you get caught up in relying too heavily on AI for intelligence and you don’t realize, you don’t know for yourself where their intelligence is, begins and ends and how it can trip up over itself. And you’re learning it in the most inopportune times, right? Everybody can spot. Well, I’m not going to say everybody.

Chris Davis 2:30
A lot of times you can spot an AI generated response. People are putting them in Facebook groups now, communities, software documentation. You could just tell when a, when a robot has written it and it’s like, look, if, if, if I wanted to ask my question, this is what I’ll see People ask a question in a group, somebody copy that question, put it in ChatGPT or Claude and copy what it says in the comments. And it’d be long sometimes, y’, all, It’d be a long article.

Chris Davis 3:00
I’m like, come on, y’. All, this, you’re cheapening the feel of community. But, but that’s not the point. The point is this.

Chris Davis 3:07
Don’t. No matter how much you use AI, do not let it become your intelligence because it’s called artificial for a reason. I think that people just bypass the art, artificial part of intelligence. AI is not real intelligence, it’s artificial.

Chris Davis 3:24
And there are ways that you can use it that will safeguard you. But you still need intelligence. You need intelligence to check its intelligence, right? So in that respect, when people are trying to do something, I see that the artificial intelligence is filling in the gaps of areas that they’re ignorant, that they really can’t afford to be right and in and out.

Chris Davis 3:56
And I’ll share with you because I was and am to some degree still this way. And I’ll share with you all how I learned AI. I learned AI by answering it, asking it questions that I already knew the answers to from history, technology, personally, just a wide variety of things. I would have these conversations sitting down, you know, watching the basketball game or something.

Chris Davis 4:21
I would just ask questions. Now I know the answer. I have a well formed out thought and I would ask it and I would look at how it responded and I say, oh, wow, that was dangerous. Because if I didn’t know, that wasn’t true.

Chris Davis 4:33
That sounds really good. And there are times where, and I’ve. Listen, I’ve learned it the hard way. I’ve learned, quote, unquote, learned from AI and then took that into a debate and got ripped apart.

Chris Davis 4:46
And I’m talking about not even highly intelligent people. There’s people that’s like, no, Chris, that, that’s, no, no, that’s not the full story. And what did that highlight? Often I just misses context because we don’t provide it enough.

Chris Davis 5:00
And even if we do provide it enough, it’s limited, right? There’s, there’s ways to feed it context consistently to get a really good output. I’ve, I have achieved that. And it is extremely powerful when you do that.

Chris Davis 5:14
But for the average user, you don’t know. You talk so broadly to AI and you don’t even realize it. You’re like, hey, do design me a really good landing page that people will opt in for what is really good. Who are people?

Chris Davis 5:32
Right? So vague. Right? But that that’s what I would see and I would like.

Chris Davis 5:38
And I would use it to say, okay, so I really, if I don’t know something, I should at least go learn. And if I’m learning on AI, I need to go cross reference. And I started to like, at the time, Perplexity was one of the only models that when you asked a question, it would list all of the sources. So I would look at the answer, but then I would go and click through the sources and formulate my own answer.

Chris Davis 6:03
I just, I’m one of those people, I need to own my own intelligence. Right, so, so what happened is you go into an arena of writing code where I’ve, as an engineer, I’ve written code. So I know object orient. I learned object oriented code with Java script, Java back in the day, not JavaScript, but Java and C.

Chris Davis 6:24
C Sharp, right. Object oriented code is still a thing. It just, it just looks different from what I can glean from like Python and all of that. I don’t, I don’t know any of these languages, but I know enough about code to keep, I know enough about code to keep these agents honest.

Chris Davis 6:42
And this is, this is what I’m saying with, you still need to be smart. So you may not need to understand how to write code, but you do need to understand some terminology. You need to understand when it says fold it into the design or when it’s. Talk about hydrating, you’re like, what is that?

Chris Davis 7:03
Right? And it uses these terms like running a bash command and deploying, things like that. You, you really do need to know that. And I’m not saying I knew it all.

Chris Davis 7:15
A lot of times I learn it as I’m building and I will pause and say, hey, speak to me in layman’s term. Too much jargon. Make it plain. Because AI loves to just talk.

Chris Davis 7:29
So technically it’s like, yes. And then because of the abstracted variable, we needed to append it in the API route. And I’m like, what? Stop.

Chris Davis 7:43
What does it mean? Oh, we just added a URL so that the connection can be established and the data can be passed. Okay, we’ll say that, right? But I’m intelligent enough to know some idea and then when I’m not, I ask it to break it down.

Chris Davis 8:01
Right? So when it comes to coding, you don’t need to know how to code, but you definitely need to know some, some jargon in terminology and software best practices. And the reason being is because AI is the most inconsistent output creator ever. You can have skills, you can have rules in your memory.

Chris Davis 8:29
You can have your Claude MD file and your Agents MD file. I mean, you can give it all the instruction, tell it to do one thing the same way twice. And you will get two different outputs. You’ll get two different outputs.

Chris Davis 8:44
And it seems like it’s almost like a shake of the dice. Today it follows rules 80%, tomorrow it follows rules 50%. But you have to. You have to know this.

Chris Davis 8:58
And this is why I’m saying you still have to be smart. Because you cannot just rely on AI to do the majority of thinking and figuring out in your business doesn’t follow rules. It’s inconsistent. It lacks context.

Chris Davis 9:11
It often leaves things out. I can’t. I can. I’ve lost count how many times I’ve had to correct it.

Chris Davis 9:17
And it says, oh, I’m sorry. Oh, you’re right. So, y. Y’, all, you.

Chris Davis 9:23
You can’t turn your brain off. In fact, you. You have to keep it on probably a little higher if. If you’re gonna do it right.

Chris Davis 9:30
I know that there are people that turn their brain all the way off and just let AI produce everything for them, but you’re coming from somebody. I enjoy the thinking process. I enjoy being intelligent. I enjoy learning.

Chris Davis 9:42
So I approach AI totally different. I approach to learn, cross reference and check and then implement. And sometimes I’m learning as I’m implementing. Sometimes I’m just there to learn.

Chris Davis 9:54
You know, it’s a. It’s a very dynamic experience, but I know enough about all that I’m doing to keep it honest. Okay? Because what happens?

Chris Davis 10:09
What. How do you get stuck? How do these people that I’ve encountered and see online, that I know personally, and both the ones that I don’t know, how do they get stuck? Well, you get stuck when you don’t own the intelligence.

Chris Davis 10:24
That’s when you get stuck. When you have no intelligence, no understanding of code, or what you’re trying to do, you’re just relying solely on AI, Right? And you learning your knowledge is what breaks the loop. Your knowledge, you learning.

Chris Davis 10:47
And this is why you still have to be smart. You need to not only learn how AI is operating you, but you also need to learn how it is operating with respect to you. What I found is that when I talk, I’m very authoritative because I have strong opinions and I know what I know. So.

Chris Davis 11:08
And because AI is usually agreeable before it’s not, it usually takes what I say as the gospel and it doesn’t challenge me. And even when it challenges me, it’s like that fake Challenge where you could tell it’s just trying to create something that is contrary to what you’re saying, but it doesn’t really make sense. And it’s like, no, that doesn’t make sense. That’s not what I said.

Chris Davis 11:29
But I have these debates and the reason being is because it helps me learn the limitations. It helps me learn where it’s getting caught up. One of the things I’ll say this, I used to have questions. Should I keep one long, long chat or should I change the chat after.

Chris Davis 11:51
After every topic and context rot is real. These LLMs are not, to a point, they don’t have memory. Everyone, let me just pause and say this. There’s two things.

Chris Davis 12:05
LLMs don’t have memory. Two, they don’t know who they are. These LLMs are models that are trained and then labeled. And they’re usually trained on your information, by the way, anthropic.

Chris Davis 12:19
Anthropic took the LLM, trained it, meaning it fed it a set of data and called it Claude. And it refers to that set of data in a complex way. You know, how it retrieves it and creates context and all that. That’s the whole point of AI.

Chris Davis 12:36
But you can technically buy your own LLM and train it on your data and then you will get the most accurate output. Because what’s happening, let’s, let’s go technical for a second. What’s happening is when you have a request, when you put in a prompt, natural language processing means you can speak in your natural language and it will process it in a way that it understands. Right.

Chris Davis 13:09
I’m trying to give you the simple version because you’ll go online and it’ll be very complex. It’s one of the roles of the LLM. Well, now that it knows what you’re trying to do, it needs context. So it will often take whatever you, whatever was in that prompt as context and package it.

Chris Davis 13:29
And that’s what gets sent to the LLM. The LLM based on the information that it has been trained on and has access to, produces a response. So when you see that Claude or Chad is updating your memory, what, what that’s doing is the LLM did not remember you. It starts.

Chris Davis 13:51
It’s stateless is, is the technical term. This is why I’m telling you, you still need to be smart. You need to know what these words mean. Stateless means I do not remember what I just did.

Chris Davis 14:03
And you can be talking to you like what I just told you. I don’t know what you’re talking about. So what do we do? We put it in memory.

Chris Davis 14:11
Well, what is memory? Memory is stored markdown files on a server that gets selectively fed in to the LLM along with your prompt. So this is what I call a super prompt. When you have memory files and then you give it a prompt, it takes that prompt and then it searches the memory files to see what’s applicable, adds that context, and then that’s what’s fed to the LLM.

Chris Davis 14:42
This is all happening at rapid rate. So it’s not just, hey, I’m trying to figure out a social media post. I want to establish myself as a thought leader in this space and I want to tell a story about xyz. That’s one.

Chris Davis 14:57
And then it will go and look in the memory files. We’ll call it the router. We’ll go look in the memory files and find what memory provides the necessary context or the appropriate context for that prompt. Couple that and then that’s what gets sent to the LLM.

Chris Davis 15:17
LLM processes it using the information it’s been trained on and then gives you an output. That’s as simple as I can as I can make it. So with that being said, you can stop thinking that there’s memory. There’s no memory chip in the LLM.

Chris Davis 15:38
Your memory chip is when you go into Claude or ChatGPT and click on memory and you look at all of those files, everything that it’s written about you. So I initially got here when I was saying, do you switch new chats, context route rot. People think that the more information you have, the better. No, no, you, you’ve got to synthesize that information down and go with the highlights.

Chris Davis 16:06
Us as humans, we think that you need every word, every syllable. You need to understand everything that I said. AI is just like, hey, have you, has this ever happened to you all? You did a whole bunch of word vomiting, right?

Chris Davis 16:21
You’re using voice dictation and you’ve got a whole page worth of your thoughts. And when AI, it says, oh, okay, I see what you’re trying to do. And it, and it summarizes it in two sentences and you’re like, oh, I guess that is what I was trying to say. It doesn’t need all of the words and exploration, it just needs the main points.

Chris Davis 16:43
So when you have a chat that’s ongoing, it’s just filled with noise, and too much noise at a point breaks, the context breaks and you’ll find yourself like, man, this chat was really good. But the more I said, the less accurate it got. Well, it, it. And we haven’t even talked about the context window.

Chris Davis 17:05
It can only hold so much, so you don’t even know when that cutoff point is. So now new information you’re giving it, it’s for getting old information and you don’t know what it’s forgetting. So what I’ve done, this is why you still need to be smart. Not just intelligent in terms of knowledge, but also intelligent on how the LLMs work.

Chris Davis 17:29
So what I would do is once I get, once I get my main idea in a chat, I have it. A, if it’s something I want ongoing for context, moving forward, update the memory. B, I tell it to create a summary, a markdown file that summarizes all of the main points. So if I ever needed to pick up where we left off in another chat, I could do that.

Chris Davis 17:53
I tell it that so it has the context of what to create. And if I did ever want to come back to it and start a new chat, it would be able to start from a educated but synthesized starting point. But more so you have these chats, they get so long you forget stuff. You’re not going to go back through and read that.

Chris Davis 18:15
Do you. Do you guys realize that the mo. The more you interact with AI, the more you’re reading? Some of you are reading books every week, not physically, but how you’re interacting with AI and reading its response.

Chris Davis 18:30
And AI gets very verbose. These LLMs get very verbose. I have to tell it all the time and hey, shorten that. In fact, I had to add to the memory of my.

Chris Davis 18:41
I use claw primarily, but in my clawed account I said, hey, when you reply to me, this is the format. Give me the big idea in two to three sentences so I can understand and then break it down as we go further. But don’t just sit up here and give me this big long page worth of information, right? But I know that.

Chris Davis 19:03
So that’s why I break my chats up. Now I don’t keep one long chat going because I Learned how these LLMs are really working. And that in that everybody, that’s the main thing, is that you are being sold to do something, get in on this, don’t get left out. Look at what you can do with AI and you’re not really learning how to do it.

Chris Davis 19:30
You’re just kind of following what everybody else is saying. And I’m saying it as a person that has done it. I was on YouTube and I’m an engineer by trade and I saw this other engineer and he was Talking about how to use AI, and I’m like, man, I just trust engineers. And I did his method and it was the hottest garbage I’ve ever produced.

Chris Davis 19:52
I was so mad, it burned through my tokens. I mean, I was so frustrated. I haven’t listened to that guy again. Matter of fact, every.

Chris Davis 19:59
In fact, when I’m on YouTube, you know, you can click those little three dots and say, not interested. Like, do not show me him ever again. He led me astray, not because he wasn’t smart, but because his flow does not match how I work. I don’t know what every step to his flow is, but the few that I tried, I did not like the output at all.

Chris Davis 20:21
I’m like, here, have it. Take it back. Get out of here. This is ridiculous, right?

Chris Davis 20:27
But this is all part of the exploration and finding your way. We’re not just taking or you shouldn’t just be taking somebody’s word of how a thing is done and just doing it that way. And that’s what’s happening. You’re joining groups and they’re like, hey, look, just use this prompt.

Chris Davis 20:46
Use this. And I’m like, well, hold on, let me look at that prompt. Why? Maybe I’m not building that way.

Chris Davis 20:53
I’ll start with what you’ve given me, but I’m not ending there. I need to own this understanding in this intelligence. That’s what helps me make sure that I can always stay ahead of the AI model. Not ahead in the world.

Chris Davis 21:11
I like to stay ahead in the world too, but I always need to be able to check the model. And lastly, I’ll say this. You still need to be smart. Smart in your prompting.

Chris Davis 21:23
Smart in your prompting. A lot of you just don’t know how to prompt. And you don’t learn by getting swiped files and people saying, hey, use all my prompts. You actually learn by just watching what.

Chris Davis 21:36
How other people prompted. I learned so much when someone shares their, their chat thread and I’m like, oh, I would have never thought to ask AI that. Wow. You know, again, sharing.

Chris Davis 21:48
This is again, the, the. The. My community is what I call a sharing space. Because where else are you going to get this?

Chris Davis 21:55
Most people are going to charge for it. Like, join my community for $10 a month, 99, and I’ll share. We share freely. Now.

Chris Davis 22:02
There are, there, there are benefits to becoming a paid member, but you can, you can go far with just a member account that. That is of no cost, right? And I say all this to say this. Imagine if you don’t do anything and you go ask AI and you have no way to know when it’s lying and when it’s not.

Chris Davis 22:35
It’s. It’s the. It’s. The scariest thing to me is to ask AI about something I have no clue about and think that its response is the gospel.

Chris Davis 22:46
If I don’t have a clue about something, I will ask. It will start to educate me, but then I’ll start poking holes at what it’s saying. And then when I get to a point, I’ll go do my other research, bring that information back. You get what I’m saying?

Chris Davis 22:58
I’m not relying solely on AI you still have to be smart. And the easiest way to start to understand this is just ask it questions you already know the answer to. And you’ll see how it’s answering. You’ll see what it omits out.

Chris Davis 23:17
You’ll see how it lacks context in its response. You’ll see how sometimes its inability to have any level of emotion is skewing the answer. There are times where, you know, I. One of the prompts I like to give AI is, hey, I’m about to say some stuff.

Chris Davis 23:34
Do not psychoanalyze me. Do not be a therapist, and do not advise me. Just make it clear what I’m trying to say. One of the best prompts, everyone.

Chris Davis 23:46
And then I just word vomiting and say, well, what you’re saying is boom, boom, boom, boom, boom. I’m like, yeah, that is what I was trying to say. Some of you need to do that before you send an email. Some of you need to do that before you have a conversation with somebody, right?

Chris Davis 23:59
But, but once you know and you start to learn how AI, how AI actually works, you’ll get a better output. If you’re not getting the output that you want, it’s not solely on AI. I used to think that and I used to yell at my LLM. I don’t yell anymore.

Chris Davis 24:15
I realize it’s me. The context I’m providing, the way I’m instructing. I take ownership of it. These mod, these models are artificially intelligent.

Chris Davis 24:25
Now, don’t get me wrong, they are not real smart, like real world smart. They are artificially intelligent, true to the name. So let’s not get that twisted. I don’t want to give them more credit because I have wasted a lot of times on that artificial intelligence when I need a real intelligence, when I needed actual intelligence, the real AI, Right?

Chris Davis 24:48
So don’t, don’t get it mixed up, everyone. You need actual intelligence to get the most out of artificial intelligence. Okay. So I hope that was helpful just to guide you in this journey.

Chris Davis 25:02
If you find yourself stuck, if you find yourself hitting a wall, there’s something you don’t know. There’s something you don’t know. And usually you’re not going to know it by asking AI. You still need people, you still need community, and you still need books and learning and research.

Chris Davis 25:23
It’s still required. Everybody, do not let this AI age and phase. And maybe it’s not a phase. This new way of living dumb you down.

Chris Davis 25:34
Okay? And I’ll say this in closing to go back to my original story. This person was trying to get a app launched and what AI was telling them was, oh, it’s time for you to bring in a developer. And there are times where you do need to bring in a developer.

Chris Davis 25:53
That is true. However, I couldn’t agree until I could see what did you prompt for it to say that? Because this may just be a gap in you knowing how to prompt, you know, and sometimes we could be prescriptive, get frustrated and say is, do I just need a developer to do this? And I would be agreeable, like, yeah, and here’s why you needed developer.

Chris Davis 26:15
And it’s like, whoa, whoa, whoa, no, you just didn’t know enough to ask it the right questions to prompt it appropriately to get further. Without a developer, there is a time where you will need a developer, especially those of you that are bringing building SaaS solutions. But for me, no, the internal software that I’m building, there’s no need for a developer. I just need it to be smarter, that’s all.

Chris Davis 26:36
So hopefully this was helpful in your journey in AI and app building and just overall, the day and age that we’re in with automation, AI technology, hopefully this helps keep your brains on. Please keep your brain on. If AI isn’t making you more intelligent, you’re losing and you’re using it wrong. AI should be adding to your intelligence.

Chris Davis 27:04
You should not be adding to its intelligence. Only right when you look at the end of the day and you weigh it, you should be smarter. You should be more smarter than AI because we are training these models on our information, right? They need to train us now.

Chris Davis 27:21
They need to give us intelligence. And it’s, and it’s an intentional approach to say, okay, you said this. Let me go cross reference it. Let me go talk to somebody, let me go see where this breaks.

Chris Davis 27:35
Stay, Stay intelligent. Stay intelligent, everyone. All right, hopefully this was of value. I want to thank you.

Chris Davis 27:44
For everybody who continues to listen, I am seeing an uptick and I can’t be more thankful. So for those of you who are sharing, those of you who are having converse, more educated conversations now, people are telling me, hey, this episode that you recorded helped me have a conversation at my job and people respected what I was saying. That pleases me. It is one of the reasons why I do this is to transfer intelligence from me to you.

Chris Davis 28:13
And I am always grateful for your listenership. With that being said, going to go ahead and sign off. So until next time I see you all online, automate responsibly, my friends.