AI just officially took our jobs… I hate you Devin

Fireship
13 Mar 202404:13

Summary

TLDRThe video script discusses the advancements in AI in software engineering, highlighting the emergence of Devon, an AI developed by Cognition Labs that can perform tasks like a human software engineer. Devon accesses developer tools, solves problems by searching the web, and iterates on code until it meets the requirements. It outperforms other AI tools on GitHub benchmarks, and while it's not yet ready to replace human engineers, its capabilities are growing. The video also touches on the potential future where AI could significantly change the programming landscape, and humorously suggests a gene therapy pill as a solution to staying relevant in the face of AI advancements.

Takeaways

  • 👨‍🌾 The speaker's family history involves being replaced by robots in various industries over three generations.
  • 💻 The speaker chose programming as a career to avoid being replaced by automation, but fears that even this field is at risk.
  • 🐍 'Co-pilot' and 'GP4' are tools that have started to automate parts of coding, leading to concerns about job security for software engineers.
  • 🚀 Devon, an AI developed by Cognition Labs, is being hailed as the first AI software engineer capable of performing tasks like a human engineer.
  • 🔍 Devon uses developer tools like a terminal, browser, and code editor to solve problems by searching the internet and iterating on code until it finds a solution.
  • 📈 The AI has shown promising results on the software engineering benchmark, scoring significantly higher than its predecessors like GPT-4.
  • 🏆 Despite its improvements, Devon is not yet proficient enough to replace most software engineers, although its performance is expected to improve over time.
  • 🌐 Tools like Devon and Google's Alpha Coder are pushing the boundaries of what AI can achieve in software development, with Alpha Coder already outperforming human programmers in certain tasks.
  • 🤖 The potential future of software development could involve AI tools like Devon creating multiple solutions for tasks, from which humans can choose the best.
  • 💡 Silicon Valley's interest in replacing programmers with AI is evident through the backing of several startups in this field.
  • 💥 The speaker humorously suggests a gene therapy pill as a solution to staying competitive in a world where coding might become obsolete.

Q & A

  • What was the occupation of the speaker's great-grandfather?

    -The speaker's great-grandfather worked on a farm.

  • What happened to the speaker's great-grandfather's job?

    -His job was replaced by robots.

  • What was the occupation of the speaker's grandfather?

    -The speaker's grandfather worked in a car factory in Detroit.

  • How did the speaker's grandfather's job end?

    -He was also replaced by robots.

  • What was the speaker's father's occupation?

    -The speaker's father became a McDonald's chef.

  • What happened to the speaker's father's job?

    -His job was replaced by robots as well.

  • Why did the speaker choose to get into programming?

    -The speaker got into programming to avoid being replaced by robots, as coding was considered the new literacy.

  • What is Devon, as described in the script?

    -Devon is an AI software engineer developed by Cognition Labs, capable of performing tasks similar to human software engineers.

  • How does Devon solve problems?

    -Devon solves problems by Googling for solutions, creating a plan of action, accessing API documentation, writing code, and iterating until it solves the original problem.

  • What is the current success rate of Devon on the software engineering benchmark?

    -Devon currently has a success rate of 13.86% on the software engineering benchmark.

  • What is the significance of Devon's performance compared to GPT-4?

    -Devon's performance is significantly better than GPT-4, which only solves GitHub issues 1.74% of the time, indicating a huge improvement in AI's capability for software engineering tasks.

  • What is the potential future development for AI tools like Devon?

    -In the future, AI tools like Devon could potentially perform tasks in parallel, generating multiple solutions to problems and selecting the best ones, much like Google's Alpha Coder.

  • What is the speaker's humorous solution to the potential obsolescence of coding skills due to AI?

    -The speaker jokingly suggests a gene therapy pill that rewires the brain for optimal coding performance, with side effects including hallucinations and physical death.

Outlines

00:00

🚜 The Evolution of Automation and the Rise of Devon

This paragraph discusses the historical replacement of human labor by automation, starting with the speaker's great-grandfather on a farm, through the grandfather in a car factory, to the father as a McDonald's chef. The speaker then transitions into programming to avoid being replaced, only to find that AI tools like Devon are emerging to potentially replace software engineers. Devon is an AI developed by Cognition Labs that utilizes the same developer tools as humans, capable of solving problems by searching the web and iterating on code until it finds a solution. The AI's performance on the software engineering benchmark is notably higher than GPT-4, and it's already performing real jobs on platforms like Upwork.

Mindmap

Keywords

💡Robots

In the context of the video, robots symbolize the automation and technological advancements that have historically replaced human labor in various industries, such as farming and manufacturing. The great-grandfather being replaced by robots on a farm illustrates the shift towards automation and the impact it has on job displacement.

💡Detroit

Detroit is historically known as the hub of the American automotive industry. In the video, it represents the industrial era where the grandfather worked in a car factory, which was later also subject to automation through robots.

💡McDonald's Chef

The term McDonald's Chef refers to the role of the father in the video who worked at McDonald's, which was then replaced by robots. This highlights the pervasiveness of automation across different sectors, including the fast-food industry.

💡Programming

Programming is the process of writing code to create software, applications, or to control machines. In the video, the individual turns to programming as a means to avoid being replaced by robots, considering it a safe profession due to its complexity and the demand for software development.

💡Co-pilot

Co-pilot, in the context of the video, refers to an AI tool that assists in coding by writing code or providing suggestions. It represents the evolving nature of AI in the software development field, where AI is becoming capable of performing tasks traditionally done by human programmers.

💡Devon

Devon is an AI software engineer developed by Cognition Labs, which has the ability to interact with developer tools and solve programming problems. It signifies a significant leap in AI technology, as it can perform tasks similar to human software engineers, raising questions about the future of programming jobs.

💡GitHub

GitHub is a platform used by developers to store, manage, and collaborate on code. In the video, it is used as a benchmark to test the capabilities of AI tools like Devon in solving real-world programming issues, highlighting the practical application of AI in software development.

💡Alpha Coder

Alpha Coder is a tool mentioned in the video that uses an AI to generate thousands of possible solutions to coding problems and selects the best one. It represents the parallelization of code generation, which is a step beyond the serial approach of tools like Co-pilot and Devon.

💡Upwork

Upwork is a freelancing platform where individuals can post jobs and receive bids from freelancers. In the video, it is used as an example of how AI tools like Devon could potentially disrupt the freelancing market by providing solutions to posted jobs, further emphasizing the impact of AI on employment.

💡Silicon Valley

Silicon Valley is a region in California known for its high-tech innovation and venture capital investments. In the video, it represents the hub of technological advancement and the driving force behind the development of AI tools like Devon, which aim to replace human programmers.

💡AI Ethics

AI Ethics refers to the moral principles and guidelines that govern the development and use of artificial intelligence. The video touches on this by discussing the potential for AI to replace human jobs and the ethical considerations that come with such advancements.

Highlights

A century ago, the user's great-grandfather worked on a farm before being replaced by robots.

The user's grandfather worked in a Detroit car factory, which was later automated.

The user's father was a McDonald's chef, who was also eventually replaced by automation.

The user chose programming to avoid being replaced, considering coding as the new literacy.

The emergence of AI tools like Co-pilot and GP4, which can write code autonomously.

Devon, an AI that can use developer tools to solve problems like a human software engineer.

Devon can create a plan of action based on product manager requirements.

Devon accesses web resources and APIs to write and debug code in an infinite loop until the problem is solved.

Devon's performance on the software engineering benchmark is significantly higher than GPT-4's.

While Devon is impressive, it is not yet capable of fully replacing software engineers.

The potential future where AI tools like Devon can generate multiple solutions for a single problem.

Devon is already performing real jobs on Upwork and earning money.

Cognition Labs, the developer of Devon, is backed by significant Silicon Valley investors.

Other stealth startups like Magic and Dodev are also developing AI software engineering tools.

The ambition of Silicon Valley to replace programmers with AI, with significant financial incentives.

The current limitations of AI tools in handling complex problems and their potential to be counterproductive.

Optimistic views on the future of AI in programming, with the goal of making everyone a programmer.

The hypothetical solution to the automation of coding: a gene therapy pill for optimal coding performance.

Transcripts

play00:00

100 years ago my great-grandfather

play00:02

worked on a farm then he got replaced by

play00:04

robots my grandfather went on to work in

play00:06

a car factory in Detroit he got replaced

play00:08

by robots my father went on to become a

play00:10

McDonald chef but he too was replaced by

play00:12

robots I didn't want to suffer the same

play00:14

fate so I got into programming they said

play00:16

coding is the new literacy and everyone

play00:18

needs to Learn Python but then co-pilot

play00:20

started writing a little bit of code gp4

play00:22

started writing a lot of code but then

play00:24

yesterday Devon burst onto the scene an

play00:26

AI that takes over our tooling to write

play00:28

code in an infinite feedback loop in

play00:29

today to video we'll find out if this

play00:31

little named Devon is the final death

play00:33

blow for software engineering or if it's

play00:35

yet another hyped up pile of garbage

play00:36

designed to milk what's left of the VC

play00:38

money in this latest AI hyp cycle it is

play00:40

March 13th 2024 and you're watching the

play00:42

code report the software Engineers right

play00:44

now are absolutely terrified begging the

play00:46

developers of Devon to stop what they're

play00:48

doing human software Engineers over at

play00:50

cognition Labs have developed a machine

play00:52

that they're calling the first AI

play00:53

software engineer it has access to all

play00:55

the same developer tooling that you do

play00:57

like a terminal a browser a code editor

play00:59

and so on and that means it can solve

play01:00

problems like a real software engineer

play01:02

by Googling until you find a blog or

play01:04

stack Overflow answer that you can copy

play01:06

and paste into your code a product

play01:08

manager can give Devon some requirements

play01:09

and it will respond by creating a plan

play01:11

of action then it goes to the web to get

play01:13

some API documentation then it starts

play01:15

writing some code then it runs it it

play01:17

gets an error it fixes its code and

play01:19

continues doing this in an infinite Loop

play01:21

until it solves the original problem

play01:23

it's similar to tools like Auto GPT

play01:25

where an llm is the base technology but

play01:27

you give it the ability to perform

play01:28

actions and then react to the feedback

play01:30

that you get from those actions Devon is

play01:32

already smashing other tools on the

play01:34

software engineering Benchmark which is

play01:36

a test that determines if AI can solve

play01:37

real world issues on GitHub generally

play01:39

speaking AI sucks really hard at this

play01:41

Benchmark GPT 4 only gets the job done

play01:44

1.74% of the time while Devon is

play01:46

currently sitting at 13.86% that's a

play01:49

huge Improvement but a tool that can

play01:51

only solve GitHub issues 133% of the

play01:53

time is not quite good enough to replace

play01:54

most software Engineers today although

play01:56

this number will likely improve in the

play01:57

future if things are going to get really

play01:59

lit though when tools like this start

play02:00

doing everything in parallel like

play02:02

Google's Alpha coder which currently

play02:03

beats 85% of competitive programmers and

play02:06

it does that by using an llm to generate

play02:08

thousands of possible solutions in

play02:10

parallel then picks the best one in the

play02:12

future you'll be able to post a job just

play02:14

like you would on upwork and 10 minutes

play02:15

later Devon will come back with a

play02:17

thousand different apps and you can

play02:18

choose the best one in fact Devon is

play02:20

already doing real jobs on upwork and

play02:22

getting paid for it so I guess that

play02:24

makes it official they took our jobs

play02:26

your job your job your job can learn new

play02:30

technologies it can build and deploy

play02:32

apps it can fix bugs and existing code

play02:34

bases train its own AI models train its

play02:36

own AI models train its own AI models

play02:38

and do all kinds of other stuff pretty

play02:40

impressive and the company itself is

play02:42

backed by Peter the's Founders fund

play02:43

along with the Callis and brothers from

play02:45

stripe and a bunch of other Heavy

play02:46

Hitters from Silicon Valley and

play02:48

cognition is not the only stealth

play02:49

startup in this game there's also magic

play02:51

dodev which has been funded with over

play02:53

100 million and has a shared investor of

play02:55

elad Gill who's also made investments

play02:57

into mistal and perplexity AI it's naive

play02:59

to think think that Silicon Valley

play03:00

doesn't want to replace programmers they

play03:02

absolutely do and whoever does it first

play03:04

is going to make a ton of money but

play03:06

there's really no point in being a dumer

play03:08

for one there's a very high probability

play03:10

that this technology will be like

play03:11

self-driving cars or 3D printing which

play03:13

promise to revolutionize the world but

play03:15

in the grand scheme of things never

play03:16

delivered on those promises currently

play03:18

the only thing that Devon can do is bang

play03:20

out crappy little demos and anyone

play03:22

that's used AI tools over the last year

play03:24

know that they're extremely useful for

play03:25

simple problems but as soon as you get

play03:27

into more complex problems they become

play03:29

counterproductive and end up hurting

play03:30

your code more than helping it however

play03:32

there's a lot of smart people who think

play03:34

that problem will be solved like Jensen

play03:35

Wong recently made this bombshell

play03:37

statement it is our job to create

play03:39

Computing technology such that nobody

play03:42

has to program everybody in the world is

play03:44

now a programmer this is a miracle of

play03:46

artificial intelligence but even if

play03:48

Devon succeeds and coding becomes an

play03:50

archaic skill there is a solution to

play03:52

that depressing outcome and that of

play03:53

course is the one daily pill I'm working

play03:55

on with fiser it's a dimethyl tryptamine

play03:57

based gene therapy that rewires your

play03:59

brain for optimal coding performance

play04:01

common side effects may include headache

play04:02

diarrhea geometric hallucinations ego

play04:04

death and physical death so don't let

play04:05

the robots take your job ask your doctor

play04:07

if getting fir ship pilled is right for

play04:09

you this has been the code report thanks

play04:11

for watching and I will see you in the

play04:12

next one

Rate This

5.0 / 5 (0 votes)

Ähnliche Tags
AI EngineeringSoftware AutomationDevon AIGitHub BenchmarkFuture of CodingJob DisplacementSilicon ValleyInvestment TrendsProgramming SkillsTech Industry
Benötigen Sie eine Zusammenfassung auf Englisch?