0:01 Hey guys, how you doing? Uncle Steph
0:04 here. So, why shouldn't developers fear
0:09 AI? Why is we see a lot of fear? Now,
0:11 why is this fear irrational in the
0:14 medium and long term? Well, first of
0:17 all, AI based development though
0:19 productive and useful yada yada yada.
0:21 I've talked about this before. It's not
0:23 perfect. It's not trivial. You're not
0:25 going to have some random business
0:29 person, some HR department person
0:31 being able to develop
0:35 really good functional apps with AI
0:37 unless they put a lot of effort into it.
0:39 AI will make you more productive just
0:41 like React
0:44 made front-end web app development much
0:48 much easier. Right? VanillaJS,
0:51 it's capable, but if you want to do what
0:54 React does, it makes you, you know, 10
0:56 times more capable. AI is the same
0:59 thing. It speeds up the process. It
1:03 makes it easier, but it still requires a
1:04 lot of work and there's a lot of
1:06 planning and there's a lot of complexity
1:09 with AI development. Don't fool
1:11 yourself. this irrational fear in my
1:14 opinion about AI killing all jobs. A lot
1:17 of it's hype. A lot of it's because our
1:19 lizard brains, our lower brains, link
1:21 below, I have a whole course on it. It's
1:24 designed to overemphasize, to
1:27 artificially magnify
1:29 potential threats.
1:31 AI is perceived as a potential threat
1:33 because it does a lot of stuff, but it
1:35 makes a lot of mistake.
1:37 If you look at some of my previous
1:38 videos, you see people are now being
1:42 hired to clean up what is called AI slop
1:46 code. It's like messy code.
1:51 So, we've seen RAD protest prototyping
1:53 tools before. RAD applica rapid rapid
1:55 application development. RAD is short
1:59 for rapid application development.
2:02 We have seen over the years, over the
2:05 last couple decades, in fact, several
2:07 iterations of that, several versions of
2:11 tool sets that allow you to quickly
2:14 build applications or more quickly build
2:18 applications. Most of them fall short.
2:19 That's why you don't hear about them
2:21 much these days. Some of them are there.
2:24 typically uh when you have rad tools and
2:29 I consider AI uh a high level rad tool
2:31 they're good at very narrow scopes if
2:33 you go outside of that you're going to
2:35 have some problems and that's what we're
2:38 seeing with AI so AI don't get me wrong
2:39 it's much better than previous
2:41 iterations previous versions of these
2:44 type of tools but it's still that and
2:47 the fundamental problem with AI is that
2:49 it doesn't think you see AI doesn't
2:51 think it doesn't have that capacity to
2:54 go, hm, this here doesn't make any
2:56 sense. It would just go off on these
2:59 tangents. AI makes assumptions. It
3:02 doesn't have that ability to reason.
3:04 I know they have AIS where they call
3:06 reasoning AI and it looks like it might
3:09 be reasoning to a certain extent but
3:13 it's reasoning level is
3:16 at the end of the day based on my
3:19 experience working with the latest GPT
3:23 I train a an AI fitness coach to
3:27 automate some of the uh time consuming
3:30 processes in uh when you're when you're
3:33 getting healthy. Anyway, so I trained
3:34 this thing for about three and a half,
3:36 four months now.
3:38 And it's like training a small puppy
3:41 dog. It really is. You have to give it a
3:44 context. You have it to give it a solid
3:45 context, but even though you give it a
3:49 context, it'll still start dropping
3:51 turds all over the place. It's it's it's
3:54 it's it's like, why did you do that? You
3:55 know, it's like I told you not to. So,
3:57 you have to deal with what are called
4:02 edge cases. Basically, you give your AI,
4:04 you give it a a context, and it does a
4:06 bunch of stuff automatically, and you're
4:07 going, "Oh my god, I didn't have to
4:09 explicitly code this, right? I just
4:12 prompted it in."
4:14 But then it will just like it will go be
4:17 working pretty good and all will go off
4:18 on this tangent. You're like, "What the
4:21 hell is that?"
4:23 It's like having a self-driving car.
4:24 It's it's it's to the point where it's
4:26 taking it like they've been at it for
4:28 what Tesla's been at it for many years
4:29 where it's just now getting to the point
4:32 where it could drive itself. Driving is
4:34 not new as complex as you know building
4:36 apps and figuring out business
4:38 processes. One of the things I like to
4:39 say, if you're halfway there, you're
4:44 nowhere, right? So an AI could take you
4:45 quickly towards your goal in terms of
4:48 developing application, some sort of
4:51 workflow, no question. But it's still
4:53 you still need to intervene as a human.
4:55 You still need to intervene and deal
4:58 with stuff. So for example, going back
5:02 to my custom AI, I call Brad Fit the AI
5:05 fitness coach. Very powerful, very
5:06 useful. after three and a half, four
5:09 months of training and I still have to
5:11 deal with these random edge cases, these
5:12 where it just goes off in these little
5:16 tangents and uh just ruins it. Still
5:18 very useful. So the key to working with
5:21 AI from my experience anyway and
5:24 speaking to AI experts is you have to
5:27 keep the scope of what your chatbot
5:28 would do, what your agent would do, you
5:31 have to keep it very narrow.
5:33 The more narrow and specialized the
5:36 agent, more narrow and specialized the
5:39 uh chatbot, your AI, the better off it's
5:42 going to be, the less likely it's going
5:44 to do some crazy stupid stuff that's
5:46 going to ruin everything.
5:50 If you look at Open AI, so they came out
5:54 with GPT4 and I trained my my agent, not
5:58 my agent, I changed I trained my fitness
6:01 coach AI based on GPT4 because Grock
6:04 didn't have capability of a GPT had
6:06 gemini didn't have capability of a GPT
6:09 head and it's very good on four. I got
6:10 it working after three and a half months
6:15 or so. So then five comes out JPT5. Now,
6:18 the companies that own these AI, they
6:20 they're trying to tell us, "Oh my god,
6:22 AGI is around the corner. Oh my god,
6:25 Mimosa, they come out with five and it
6:28 breaks Brad fit.
6:31 It breaks months of work. Not horribly,
6:33 but it breaks it pretty good." And I'm
6:36 like, damn. So, what does that tell you?
6:38 That tells you that it's not a perfect process.
6:40 process.
6:42 several years ago when I was looking at
6:44 AI a little bit. So I picked up a book
6:47 on it. The chapter I was concerned about
6:51 was a chapter on training models
6:53 and the author of a book was apparently
6:55 a top AI guy. He was saying and I've
6:58 gotten confirmation from other people
7:00 saying that AI training model training
7:05 is more of an art than it is a science.
7:08 That's a big red flag. When people start
7:11 telling you something is an art,
7:14 more art than science, that tells you
7:16 the process is at best imperfect. So,
7:18 when you saw they went from GPT4 to five
7:22 and they busted poor Brad Fit, my AI
7:23 fitness coach, the world's greatest
7:27 living fitness coach. That's an AI. Um,
7:30 that tells you they got problems. Now,
7:32 not just me. I have a friend of mine and
7:36 some of his uh some of his u AI which
7:39 was based on GPT4 it got busted the hell
7:41 out of it in GPT5 and they had to roll
7:43 back and it's not so it's not just me
7:45 anyway. So there you go. It's complex.
7:47 We have different levels of complexity
7:50 uh or different types of complexity
7:53 shall I say with AI than you have with
7:55 software development.
7:57 And it's just another tool in your tool
8:00 set as a developer, just like React is,
8:02 just like native JavaScript is, just
8:04 like uh WordPress is. If you're a
8:07 developer, you may client wants a a vlog
8:10 or a blogging site, if you will. Makes
8:12 sense to use WordPress or some other
8:15 content management system,
8:18 etc., and so on. So, yeah, don't fear
8:21 the AI. It is not a threat to developers
8:23 in any way and in fact it's a huge
8:26 opportunity in my opinion. So if I were
8:29 you, you learn your foundations of code
8:31 which uh I would recommend the web stack
8:32 with that will give you the most
8:35 opportunity. But it doesn't really
8:39 matter if you do C or C++ or Java or you
8:41 do JavaScript, Python, PHP, my
8:43 recommendation, HTML, CSS. You learn
8:46 that, learn to work with AI, become an
8:48 AI expert. You got to look at, you know,
8:49 some experienced developers will say,
8:52 okay, I got these tools. One tool is
8:54 called JavaScript.
8:56 Another tool is called Python. Another
8:58 tool is called SQL.
9:00 Another tool is called, I don't know,
9:02 C#, whatever. Right? These are all tools.
9:04 tools.
9:08 React as a tool. WordPress is a tool. I
9:11 would argue Wix that uh build their
9:15 website is a tool. So are AIs. Gemini is
9:18 a tool. Grock is a tool. Anthropic code
9:20 is a tool. GPT is a tool. They're all
9:22 tools. You just got to learn to use
9:25 them. Back in the 90s when the web was
9:27 new and I would go in to see clients and
9:30 they would say what is a website Steph I
9:32 would explain to them what a website is
9:33 and I get some gigs when I get other g
9:37 anyway adopt my point is adopt the new
9:40 adopt the new so learn the web
9:41 start working with the AIs understand
9:44 what Gemini could do for you what GPT
9:46 could do for you what code could do for
9:49 you what Grock could do for you
9:52 understand the tools and uh also
9:53 understand the language
9:57 So for example with my fitness app AI
9:59 allows me to do things I could not do
10:02 with code and it's allowing me to do
10:05 things that I could do with code but it
10:06 does it much faster. But on the other
10:09 hand there are still things that I need
10:11 to know the web stack and coding and
10:13 onboarding and all this kind of stuff
10:15 and architecture
10:17 databases theor database theory. There's
10:18 these things they still need to know to
10:21 make this thing work with its ultimate
10:24 vision. With its ultimate vision. So
10:27 yeah, AI is not a threat to developers.
10:29 It's just a new tool set. Leverage it.
10:33 Those who embrace the change are going
10:35 to do fantastic.
10:37 I'm old compared to you probably. I'm
10:40 ancient. I'm old enough to know that you
10:44 should never use Ruby. And
10:46 I'm embracing the change. You should too.
10:47 too. [Music]
Chrome Extension