5 Min Read • Intensity: ☕️ ☕️

Good morning Caffeinators ☀️

Despite recent setbacks, is Google still leading the AI race?

I am excited to share this week’s top GPT headlines and fun tips! Today's tip is as delightful as the whipped cream atop your icy cappuccino 🍦

Today’s Starting Sips:

• The AI Legend Leaves, but the warning lingers ⚠️

• The rise of open source is toppling titans 🌄

• Midjourney + GPT = Creative Dream Team 🌟

AI Godfather's Wake-Up Call 🚨

Geoffrey Hinton, aka “Godfather of AI“, has made groundbreaking contributions to deep learning. He laid the foundations for today's LLM giants: GPT, Bard, LLaMa, and more.

An AI Legend's Parting Words 💬

Upon leaving Google, Hinton warned about AI technology's dangers. He urges the tech world to prioritize ethical practices and responsibility in AI development.

There’s no beef between Hinton and Google. However, there are concerns.

Hinton's Key Worries:

1. Weapon of mass destruction

2. Disinformation

GodFather’s Concern: Weapon of Mass Destruction 🗡️ 

While LLMs haven't been used in warfare yet, Hinton anticipates the potential for mass destruction.

I agree, especially with cyber attacks. How? Think about LLM as an enhancement tool. It helps writers write faster; coders code faster; and, unfortunately, hackers hack faster and better.

If we do not develop better safety measures for LLMs, malicious individuals could use them to inflict harm.

GodFather’s Concern: Disinformation 🗣️

Disinformation is already prevalent. We see thousands of AI-generated photos and "art" flooding the internet daily, including this one 👇️ 

The main concern lies in manipulating the general public with nearly indistinguishable misinformation or ‘fake news’.

Fake news is already rampant without AI, so it will only worsen… by a lot. Between LLMs and AI-generated art, the situation could quickly spiral out of control.

Balancing AI Ethics & Innovation ⚖️

Hinton's warning emphasizes the need to balance AI innovation with ethical considerations. As AI becomes increasingly embedded in our lives, addressing its potential risks is crucial.

If the creator of AI is telling us to pause, maybe we should stop and pay more attention to the whys and hows.

I say that because currently, no one on Earth could explain how LLM works. Not even Sam Altman himself (founder of ChatGPT).

— Sources

The Underdog Outpacing Google and OpenAI 🚀

The Leaked Document: "We Have No Moat" 📃

On May 4th, a Senior Engineer at Google leaked a document titled "We have no moat". It suggests that Google and OpenAI's LLMs are falling behind.

Open source community is closing the quality gap fast! In just 3 weeks, they were able to train an LLM that achieves 92% of ChatGPT’s performance model.

A Wholesome Programming Community 🤗

For the uninitiated, open source refers to code that is published and available for everyone.

Yes, the programming community is wholesome. Sharing is caring. And they're making incredible progress:

1. Run LLMs on smartphones

2. Fine-tune personalized AI in a single evening

3. Limitless and free art models

4. Vicuna: 90% of ChatGPT but 100,000x cheaper (not an exaggeration)

These open-source breakthroughs have Google and OpenAI realizing they don't possess a secret weapon.

Reality Check: How Close Are We? 🧐

While open-source projects like Vicuna show promise, there's still a significant gap between them and products like GPT-3.5 (ChatGPT).

Wait… Didn’t I just say they were close? Let me explain. While Vicuna asserts possessing 92% of ChatGPT's abilities, this is only applicable under specific benchmark tests.

Additionally, turning a research project into a public product requires tackling legality, security, scalability, advertising, and more 🐙 

We haven’t even compared Vicuna with the best LLM yet — GPT4. How good is GPT4 compared to others? Subscribe and stay tuned!

Btw, I got approved by OpenAI. I’m now in their beta pool for GPT4’s API! Join me and enjoy more interesting articles in the future!

Redefining Value in the AI Arena 🏆

In the future, customers will not pay for GPT or Bard anymore. Why? Because there will be equivalent open-sourced LLMs with similar quality and FREE!

Google and OpenAI will need to rethink their strategies:

1. Collaboration: Join forces and learn from open-sourced projects

2. Integration: Emphasize seamless compatibility with third-party platforms

3. Efficiency over size: Zero in on smaller, more effective models to bring down the cost

Google and OpenAI may not open-source their LLMs yet, but collaborating with open-source projects could provide valuable insights and benefit both sides of the equation.

— Sources

Tip of the day

Combine GPT with Midjourney 🎨

Midjourney is an AI art tool that generates images based on your descriptions. You describe a photo, wait 30 seconds, and voila!

The key to getting amazing results is to provide detailed, imaginative descriptions. As a STEM person, it is hard for me to write figuratively. That’s where GPT comes in!

I asked GPT to act as a prompt engineer. Its job is to elaborate my inputs to full, coherent prompts for Midjourney. It works too well!

I said, “An abstract art of loneliness “. GPT replied:

Feed that into Midjourney, and…

And, what about this one? Can you guess the prompt for this art?

You are right! It is “companionship”. The opposite of loneliness. Don’t worry, I wouldn’t have ended on a sad note 😁 

And here is the prompt GPT generated for Midjourney:

To-Go Cup 🥤

• AI GodFather’s warning 1: LLM could become a weapon of mass destruction without safeguards 🔫 

• AI GodFather’s warning 2: Rising disinformation risks during elections 🗳️

• Open source LLMs progress concerns Google engineers 😨 

• Major LLM companies may need to collaborate with open-source communities for strategic success 🤝 

• Spark creativity by combining GPT with other AI tools, like Midjourney 🖼️ 

If you enjoyed this article, I think you will like the GPT Job Apocalypse too!

Until next time,

Your friendly barista and ☕️ GPT!