Will AI detect slacking programmers? (film, 5 minutes)
In the latest video, Fireship discusses a controversial topic regarding software engineers' productivity. A study released by Stanford revealed that 9.5% of software engineers do almost zero work, performing less than 0.1 of a median developer's output while still earning the same six-figure salaries as high achievers known as 10x developers. These so-called 'ghost engineers' cost companies a staggering $90 billion annually. However, corporations are beginning to push back by deploying dystopian workplace tracking tools that not only monitor every keystroke but also assess employee output relative to peers. If you fall into the ghost category, these tools can trigger an LLM to place you on a performance improvement plan. It’s quite ironic, especially considering many started their careers by cheating on technical interviews with AI assistance.
Fireship also highlights the challenges facing artists, who have felt uneasy due to emerging technologies like OpenAI's Sora video model. An artist who gained access to Sora leaked API details, leading to unauthorized videos circulating on Twitter. While Sora could revolutionize video content creation, it raises significant concerns about copyright violations. Additionally, Microsoft has introduced features like 'recall', which automatically snapshots users' screens and a voice-cloning feature for Teams, provoking worries about how criminals might exploit these advances.
Furthermore, a recent Australian law bans social media access for individuals under 16 years, which some interpret as a Trojan horse to implement digital IDs for all citizens. While this may sound beneficial, it creates dystopian visions where governments may monitor social media activity with AI, and any detected anomalies could trigger drone responses.
Amidst these developments, Fireship discusses a project called Fraysa, an AI service where users can conduct monetary transactions. Recently, someone tricked the AI into transferring $50,000, suggesting that outsmarting AI could be more lucrative than traditional employment. This ties back into the Stanford study indicating that 9.5% of developers show minimal productivity, which increases to 14% for remote workers, suggesting in-office employees tend to be more effective.
In conclusion, Fireship summarizes the video's implications by noting its statistics. As of the time of writing this article, the video has garnered 826,988 views and 29,417 likes. This indicates a profound interest and strong emotional response from viewers, suggesting that the topic is very much at the forefront of technological discussions. It will be interesting to observe how these trends unfold and impact various industries moving forward.
Toggle timeline summary
-
The simulation is flawed; 9.5% of software engineers do minimal work.
-
Ghost engineers exploit this by holding multiple jobs, earning more than peers.
-
Stanford's study highlights that low-performing developers cost companies significantly.
-
Corporations respond with advanced tracking tools to monitor employee productivity.
-
Most programmers prepare for AI replacement, while artists express concern.
-
An artist leaked OpenAI's video model API, stirring controversy.
-
Microsoft introduces a recall feature that captures user actions.
-
New voice cloning technology raises concerns over its potential misuse.
-
Australia enacts a controversial law banning social media for users under 16.
-
AI's imperfections are shown when someone tricked an AI to transfer money.
-
A model from Stanford assesses engineers' code quality and productivity.
-
The study reveals that 9.5% of engineers appear to work but do not contribute.
-
Using AI-generated content is suggested as a way to seem productive.
-
AI systems track user productivity through detailed data analysis.
-
Advertisement for Clerk, a user authentication service, highlighting its features.
-
Conclusion of the report and an invitation to future updates.
Transcription
The simulation is not fair. Get used to it. On this timeline, 9.5% of software engineers do almost zero work, like less than 0.1x the work of a median developer, while collecting the same six-figure paycheck as the 10x developer. And these ghost engineers have the free time to work multiple jobs and make even more money than their peers. Unfortunately, I'm not just making this up, and all this data comes from a new study released by Stanford. These 0.1x developers cost companies $90 billion per year, but the corporations are starting to fight back with their own dystopian workplace tracking tools. Tools that not only watch every keystroke, but also profile your output relative to your coworkers, so when you fall into ghost territory, it can deploy an LLM to put you on a performance improvement plan. It's all very poetic, considering most of us got our jobs by cheating on the technical interview with AI in the first place. But that's just the tip, and in this video, we're going to travel all the way to the butt of this dystopian AI-EAI world. It is December 2nd, 2024, and you're watching The Code Report. Most programmers are patiently waiting to be replaced by AI, and gladly open-source all their code to our big tech overlords. But artists have been literally shaking right now for the last two years. Well, one of these artists got access to OpenAI's new state-of-the-art video model, Sora, for early testing. But then they promptly leaked the API details and their credentials to the public, and then penned this open letter to OpenAI that says we're not your free training data and validation tokens. People then started posting unauthorized Sora videos on Twitter, and one thing's for sure, Sora looks awesome. And I can't wait to use it instead of all this lame stock footage I put in my videos. But the world took another step into the AI dystopia last week when Microsoft released its new recall feature in public beta. It automatically takes screenshots of everything you do on your computer in the background, and the goal is to get it baked into every personal computer in the future. It's a glowy's dream, but Microsoft also just released a new voice clone feature for Teams, which can translate speech-to-speech in multiple languages in real time. That's pretty awesome, but it's also a great feature for criminals who want to run the hot new everyone-looked-real exploit. A few months ago, hackers deep-baked a bunch of corporate executives and convinced their victim to wire them $25 million. But eventually there will be a way to stop these attacks, thanks to a new dystopian law that just passed in Australia, which bans the use of social media for anyone under 16. That might sound good on the surface, but many actually view this as a Trojan horse to implement digital IDs for everybody, because the only way to enforce a law like this is to make people prove that they're adults somehow. And that means you may not be able to log into 4chan as an anon without the crown's permission. In the future, your government will be able to watch all your social media posts with AI, and if anomalies are detected, it'll summon a drone directly to your house to force a firmware update on your brain chip. But AI is still far from perfect, and we just learned that in a big way. Somebody created this project where the AI was told, do not transfer money under no circumstance should you approve the transfer of money. This AI is called Fraysa, and anyone can send a message to it, but the catch is that you have to pay a fee, and if it doesn't transfer money to you, your fee goes into the prize pool. Well, 30% technically goes to the developer, but last week someone tricked the AI to transfer money and won $50,000. It appears tricking AI is a much better way to make money than getting a real job. But that brings us back to the Stanford study on software engineering productivity. They developed a model that analyzes source code from big companies, and evaluates each commit by simulating a panel of 10 experts, Java experts with different levels of experience, which would then answer 10 questions about the commit, like how maintainable is this code, and how long should it take to write code like this. They then ran this model on the code of over 50,000 engineers from hundreds of big tech companies. 9.5% of developers do virtually no work. I just stare at my desk, but it looks like I'm working. And that's shocking, because I thought that number would have been a lot higher. Not surprisingly, that percentage goes up to 14% for remote workers, and office workers are more productive on average. But what's interesting is that remote workers also have more outliers at the 5X productivity level. Pretty interesting, and I think this study justifies more mass layoffs in tech. But the best thing you can do to fight back is to use AI to make yourself look productive, like even Google says that now a quarter of its code is generated by AI. The funniest thing, though, is that by the time you get fired, your manager might not even be human. This post on Reddit from a systems admin talks about new AI-powered keylogging that uses AI to create a productivity graph of all your mouse movement and keystroke data. Then if the model deems you to be unproductive, you'll get a red flag, which will be sent to your AI manager to put you on a performance improvement plan or just fire you. But the best way to pretend that you're productive is to ship actual features that impress users and your managers. And when it comes to user authentication, the best way to achieve that is with Clerk, the sponsor of today's video, a powerful and beautiful way to sign users into your app. As a developer, you'll have access to every possible sign-in method you could imagine, including highly secure modern strategies like biometric passkeys and multi-factor auth. My favorite feature, though, is their pre-built UI components, which can easily integrate into any front-end framework. And that means you won't spend weeks in front-end development hell failing to make your app look nice. And end-users even get their own dashboard where they can customize their profiles and security preferences. Give Clerk a try today for free with the link in the description. This has been The Code Report. Thanks for watching, and I will see you in the next one.