😮Coders make an AI that replaces them

ALSO: EU passes first AI laws

Estimated Read Time: 10 minutes

Well it happened. Coders have created an AI that can do their job for them. And software engineers are NOT happy about it. And the EU is passing massive AI laws that will shake the entire industry.

Today’s Horizon

• An AI engineer codes programs for you. 

• OpenAI is sued… again. 

• The EU passes world’s first AI laws.  

• Google makes a new videogame AI.  

• ChatGPT tips that’ll change your prompts forever.

AI for Jobs (3 minutes)

😱Coders made an AI that does their job for them

What happened: Oh boy, it’s starting. Remember when everyone said you should learn to code? That future doesn’t look too bright anymore.

A startup called Cognition surprise revealed their new AI model; Devin, “the first AI software engineer”. The model claims to code entire projects from start to finish in minutes flat. Cognition revealed Devin on X, where software engineers have begun to freak out in the comments. Congrats coders, you replaced yourselves.

The details: 

• Devin can create entire websites, programs or fix bugs on his own.

• Devin learns, plans, tasks and writes the code on his own with no oversight… and he’ll finish in minutes 💣.

• Prompts are spoken in simple english. Ask for a task and he’ll do it.

• Devin can train his own AI minions to perform world domination tasks with him.

Why it matters: Devin is in the early stages. He can’t create at the level of top software engineers. Software engineers will have to use him to help, but not replace themselves. But in ten years AI programs may completely replace most computer jobs in exchange for overseeing the AI itself. This is a just a taste of the future. Learn to use these tools or perish when everyone is replaced.  

Further Details:

• Scott Wu, the CEO of Cognition, was a child prodigy who learned to code at the ripe old age of nine 😲.

• He then competed in multiple math competitions and aced them all.

• Scott is well known as a very competitive and skilled programmer.

• Devin is meant to be a coding companion for software engineers, rather than an outright replacement.

• This is the second company Scott has created, the first being Lunchclub, another AI focused company.

Takeaway: Ironically we’re getting closer to Sam Altman’s vision of a 1 person billion dollar company. While the ground floor is in danger, those who use AI might have found a serious advantage here.

In addition companies like PayPal have thrown $21 million behind Cognition’s efforts. Now is truly the time the invest in AI companies.

Stocks

📈AI Company Stocks Today

As of 03.14.2024 market close.

Headlines

🚄SILVER BULLETS

The New York Times (NYT) is suing OpenAI for copyright infringement. NYT claims that ChatGPT, copied NYT articles. Accusations have flown back and forth, but both sides agree that ChatGPT can bypass paywalls (stop paywalling NYT, it’s annoying). Does any of this sound familiar? It’s the same claim artists make about Midjourney and Dalle, since those AI models were trained on other artist’s art without their permission. Not too long ago YouTube channel Corridor was sued for making this anime video with AI (it looks awesome).

Company TogetherAI evaluated at $1.25 billion in SalesForce led funding round. THe company primarily offers easy access to powerful GPUs for processing AI. The new funding will increase their computing power while they research new architecture. It seems the most reliable AI companies to invest in right now are those focused on computing power and infrastructure, not one-off AI tools.

Studies show most LLMs think the same way. The study attempted to learn the trajectory of most neural networks (what powers AI) and found that most of them share strong similarities. This suggests that most LLMs we have now will eventually be very similar to one another.

TikTok might get banned in the United States (more general news). The US House of Representatives voted to pass a bill that might ban TikTok nationwide. The company would be banned unless they spun off from parent company ByteDance. The reason? The Chinese government partly owns and has access to TikTok, and so the US has deemed TikTok a security threat. Time will tell if this bill gets any traction.

Future of AI (2 minutes)

🧑‍⚖️EU passes groundbreaking AI regulations 

What happened: The European Parliament approved sweeping legislation to regulate artificial intelligence called The AI Act. The aim? Set global standards for safe AI use. This law is the first in the world to comprehensively restrict AI for the betterment of humanity. It has not fully passed yet.

Why this happened:

• European Parliament approved the world’s first comprehensive framework for safe AI development.

• The law was met with overwhelming approval (another sign AI trust is at an all time low).

• The law applies stricter rules depending on the level of risk the AI system poses.

• The act will force copyright compliance.

• Low level AI must be more transparent (stuff like chatbots).

• Only big, powerful AI systems will likely receive full risk assessments.

Why it matters: This is big. It’s the first major law restricting AI, and it will be the stepping stone to more legislation in other countries. AI is amazingly useful… but it’s also dangerous. There needs to be some way to prevent AI from being used destructively. Time will tell if these laws work to our benefit or detriment.

Further details:

• The legislation still needs the final approval from the Council of the European Union. This is expected to be a formality… the law is pretty much set.

• The law will come into force 20 days after its publication. Phased implementation could last until 2027.

• It will take time for other countries to adopt similar laws, if they do at all.

New AI (2 minutes)

🎮Google made an AI videogame companion

What happened:  Google has introduced SIMA, an AI videogame companion that is not designed to brute-force win. Instead SIMA is designed to follow instructions and cooperate. And it’ll work in most 3D games, not just one in particular.

The details: 

• SIMA is not focused on winning (unlike most game AI). Instead the AI is about cooperation and obediently obeying instructions.

• Some of the games SIMA is currently tested with are No Man’s Sky, Valheim, and Goat Simulator 3 🧐.

• The goal is to create an AI that understands human language and can be a helpful partner in games.

• DeepMind is the powerhouse behind SIMA.

• SIMA could be useful outside of games one day.

• The AI learns to see objects and play games the way humans do.

• SIMA has a long ways to go before reaching human-level understanding.  

• Every game may someday have a version of SIMA as game AI.

Why it matters: SIMA shines a positive light on what AI can do for us. Most AI models are focused on productivity. As a result they either help or replace human jobs. Then everyone panics or hates the AI because they’re scared AI will replace them (it probably will). SIMA shows we have more lighthearted uses for AI. I hope we see more projects like this.

If you play videogames, just know this might be the future of the medium. Which is exciting considering most videogame AI is dumb as bricks. The game industry is even bigger than the movie industry. We may find Google’s AI (or something like it) in some of the most culturally influential games in the future.

Learning (1 minute)

🧠What is prompt engineering?

You may have heard of prompt engineering. It’s a term to describe someone whose primary job is to design high quality prompts to accomplish very specific goals. Some sites have reported job salaries of over $300K a year for prompt engineers. It’s crazy.

And it’s… a bunch of nonsense. 

While it’s true the prompt quality massively impacts an AI’s output, prompts are not magical or mysterious. If a prompt works… you copy and paste it. Nobody owns the prompt because it’s just text. And they’re not that difficult to create (we have tips below).

Recently we’ve seen people using AI to create its own prompts. Those high paying prompt engineering jobs were just the result of hype. Prompt engineering is not a long lasting job or career.

Prompting (2 minutes)

🤖ChatGPT Prompt “Engineering” Tips

Here’s some general rules and guides for prompting in ChatGPT.

Be very specific in your request. The biggest reason people don’t get the prompt results they want is because they aren’t specific enough. The AI doesn’t know enough about what you want.

It’s no different than giving an employee an instruction. If you’re vague then the employee either needs to ask for clarification (AI won’t do this) or they do the task wrong (this is what AI will do). You have nobody but yourself to blame.

Asking for a detailed bulleted list of something? Well, what defined “detailed”? What details do you want? List them out.

Tired of the AI formatting your responses wrong? Tell it how to format the output. Be as specific as you can.

Provide context. Going back to the employee example, if the worker is on his first day of the job he won’t know what your lingo or commands mean. The employee needs context to understand your instructions.

AI is just as blind, so provide as much context as possible. This includes background information. If you told this command to a stranger, would they understand everything?

Ask for examples or explanations. Anytime you ask a question, try asking for 3 (specific number) real world examples. Or X number of real world applications.

Not only does this help you understand the answer better, but it helps the AI to formulate a clearer response to questions.

Productivity

🛠️AI Tools

• TaxGPT. It’s still tax season so you know what that means… AI powered tax assistants!

• Story.com an AI story/video generator.

• Docsdna. AI powered document organization.

📕FREE RESOURCE OF THE DAY

Anthropic has released a prompt database for their new AI model, Claude 3. If you use Claude at all then you need to see this.

Check it out here. 

AI Generated Images

🪓Famous people as vikings

Source: Midjourney

We’re at your beck and call!

Got feedback? Rate today’s newsletter by clicking below! 

Got more personal feedback? Fill in the box when you rate us and give us your criticism and feedback.

Thank you for reading! 

❤️Share the (ARTIFICIAL) Love!

Got a friend who wants to appear as smart as you? An AI fanatic or someone shaking in their boots of getting replaced? Share us by clicking the link below and you’ll get our love ;)