Latest / DevQuestions with Tim Corey / 250. Can AI Do a Developers Job?

Transcript

250. Can AI Do a Developers Job? · DevQuestions with Tim Corey

• 0:00
  SPEAKER_01: Can AI do a developer's job?
• 0:03
  SPEAKER_01: Can it replace junior developers?
• 0:05
  SPEAKER_01: Is it even worth becoming a developer if AI can do it better?
• 0:09
  SPEAKER_01: Let's talk about it in today's episode of DevQuestions.
• 0:20
  SPEAKER_00: Welcome to the Dev Questions Podcast with Tim Corey.
• 0:24
  SPEAKER_01: Software development is more than just writing code.
• 0:27
  SPEAKER_01: So let's talk about the rest of it.
• 0:28
  SPEAKER_01: Specifically, let's talk about AI and developers.
• 0:32
  SPEAKER_01: I'm gonna give you the direct answer up front and then explain why I'm so confident in my answer.
• 0:38
  SPEAKER_01: The answer is no.
• 0:40
  SPEAKER_01: Not it depends or maybe.
• 0:42
  SPEAKER_01: The answer is no.
• 0:44
  SPEAKER_01: So let's talk about why.
• 0:46
  SPEAKER_01: If you define a software developer's job as someone who writes code, then yes, AI can do a developer's job.
• 0:54
  SPEAKER_01: But that's not what a developer's job is.
• 0:57
  SPEAKER_01: A developer's job is to solve problems.
• 1:01
  SPEAKER_01: And that's where AI falls short.
• 1:03
  SPEAKER_01: An AI takes an existing pattern that it found and applies it to the problem at hand.
• 1:08
  SPEAKER_01: If you ask it to draw an eight-side shape, it will look up this training data and find a bunch of examples and then create an example based upon the examples that it found.
• 1:19
  SPEAKER_01: But if you count the sides and the shape, it might not be eight sides.
• 1:24
  SPEAKER_01: Why?
• 1:24
  SPEAKER_01: Because it didn't actually count the sides.
• 1:27
  SPEAKER_01: It just drew a shape that looked like the examples it was given.
• 1:31
  SPEAKER_01: That's why when you ask it to draw a nine-sided shape, it really struggles.
• 1:35
  SPEAKER_01: Now, if you asked a toddler to draw a nine-sided shape, they'd probably figure it out at some point in the four, five, six-year-old range.
• 1:44
  SPEAKER_01: Because what they would do is keep drawing sides and count them until they got to nine, as long as they can count the nine.
• 1:51
  SPEAKER_01: So why is an AI in some ways dumber than a toddler?
• 1:56
  SPEAKER_01: Because it's not actually creating something, it's just pattern matching.
• 2:02
  SPEAKER_01: So let's talk about software development.
• 2:05
  SPEAKER_01: Let's say you want to build an app for student enrollment at a college.
• 2:09
  SPEAKER_01: That seems pretty simple, right?
• 2:11
  SPEAKER_01: Well, first of all, it's not.
• 2:13
  SPEAKER_01: But in fact, AI has probably seen the code for dozens of similar apps.
• 2:19
  SPEAKER_01: But that also raises the question: why are there dozens of similar apps?
• 2:24
  SPEAKER_01: Because they all have different requirements, right?
• 2:28
  SPEAKER_01: Otherwise, why would we have any new apps?
• 2:31
  SPEAKER_01: Because everything's been invented at least once, right?
• 2:34
  SPEAKER_01: So the reason why is because of those differing requirements.
• 2:38
  SPEAKER_01: And that's where AI is going to fall short.
• 2:42
  SPEAKER_01: A viewer, this is a little side note, but a viewer posted recently a comment under one of my videos, but the comment was meant for a different video.
• 2:49
  SPEAKER_01: The comment was about comparing the outcomes from running events over the course of recent history.
• 2:58
  SPEAKER_01: Their point was that even something as simple as the 100-meter dash, it changes all the time.
• 3:04
  SPEAKER_01: We see that one person holds the record, but things weren't equal compared to now.
• 3:10
  SPEAKER_01: The track was different, the shoes were different, the crowd noise was different, the weather was different, the wind at their back versus at their front was different.
• 3:18
  SPEAKER_01: There are millions of variables that are different from one race to the next.
• 3:24
  SPEAKER_01: Yet our brains see them as equal on the surface.
• 3:27
  SPEAKER_01: The same thing is true with software development.
• 3:30
  SPEAKER_01: There are thousands, tens of thousands, hundreds of thousands of variables that go into what an application needs to do, the circumstances around the creation, and the team putting it together.
• 3:42
  SPEAKER_01: An AI doesn't account for all of those variables.
• 3:44
  SPEAKER_01: And so it doesn't understand what they mean, even.
• 3:48
  SPEAKER_01: So it doesn't evaluate the circumstances, it just tries to pattern match.
• 3:54
  SPEAKER_01: Now, with that being said, AI is absolutely a powerful tool and it has a place in software development.
• 4:03
  SPEAKER_01: I highly recommend that you learn to use AI well.
• 4:07
  SPEAKER_01: Just be sure that you are using it well.
• 4:11
  SPEAKER_01: Ask it for help with things that you understand.
• 4:15
  SPEAKER_01: Don't let it teach you, don't let it lead you.
• 4:18
  SPEAKER_01: Don't let it make decisions for you.
• 4:21
  SPEAKER_01: It's a tool, not a replacement for your brain.
• 4:25
  SPEAKER_01: But if you use it well, it can be one of your most powerful tools for creating software.
• 4:30
  SPEAKER_01: So just like high-level programming languages, WYSIWYG editors, no code solutions, and all the other developer replacing tools that came before it, AI is just another tool for faster code completion.
• 4:45
  SPEAKER_01: Get the most out of it, but don't be afraid of it.
• 4:49
  SPEAKER_01: All right.
• 4:50
  SPEAKER_01: AI is not coming for your job, but developers who can use AI well are going to be the better developers moving forward.
• 4:58
  SPEAKER_01: All right, so thanks for listening.
• 5:00
  SPEAKER_01: And as always, I am Tim Corey.
• 5:02
  SPEAKER_00: Thank you for joining us for this episode of Dev Questions.
• 5:05
  SPEAKER_00: When you're ready to learn to think and code like a professional developer, head over to imtimcore.com and enroll in a course.