📆 Hairy Ethical AI Issues + Live Training Invite

A digest of Gen AI learnings and learning opportunities

In this week’s newsletter, I’m digging into one of the biggest ethical issues I’ve uncovered in my recent work upskilling into Generative AI, but first:

If you’re new to the world of generative AI and LLMs, one of the more fundamental aspects of working with them involves fine-tuning LLMs.

Lucky for us genAI newbies, tomorrow SingleStore is sponsoring a free training on this exact topic!

📆 This week on the calendar: Using Google Vertex AI to Fine-Tune LLM Apps

Join us tomorrow live to learn more about Using Google Vertex AI to Fine-Tune LLM Apps!

If you're a developer, data scientist, or machine learning enthusiast looking to revolutionize your LLM applications... It's time to stop scrolling and start soaring! 🚀

Join us on Tuesday, September 5th, at 10:00 am PDT for an exclusive event where we'll unveil the unmatched power of Google Vertex AI!

This isn't your average tech talk; it's a transformative experience featuring cutting-edge presentations and jaw-dropping, hands-on coding demonstrations.

👩‍💻👨‍💻 Meet Rockstar Speakers: Madhukar and Alex Peng. They're not just talking the talk; they've walked the walk by successfully implementing Vertex AI in groundbreaking projects.

Here’s a sneak-peek into what you’ll get:

• Deep Dive into Vertex AI: Grasp the nuts and bolts of Google Vertex AI and discover how to apply its core components to LLM applications.

• Hands-On Coding Experience: Forget the yawn-inducing slide decks; witness and participate in a live coding demo, where the potential of Vertex AI will unfold before your eyes.

• Optimization Masterclass: Uncover the secrets to fine-tuning LLM applications. You’ll walk away with actionable techniques and tools that make a difference.

• Insider Insights: Exclusive tips and best practices from experts who've been there, done that, and are now willing to share their blueprint for success.

The best part? Whether you're just starting out in machine learning or you're an experienced developer, this event has something incredibly valuable for everyone.

Spots are filling up faster than you can say "machine learning"!

Click the link below to reserve your seat and catapult your journey into leveraging Google Vertex AI for LLM applications.

👉 Sign Up Here 👈

Hairy Ethical AI Issues

Over the last week, I’ve been taking the Generative AI with LLMs class on Coursera… and well - I’ve really been loving all that I’ve learned in that power punched course.

Yesterday I was learning more about reinforcement learning with human feedback (RLHF), which is a method for fine-tuning LLMs in order to minimize the chance the model will produce “toxic” or “harmful” content.

I’m going to share my learnings of the process below, but I want to raise one point first.

Hey, generative AI outputs aren’t perfect, but their builders have their hearts in the right place when it comes to ethical issues, and I see that generally reflected in the model outputs I get on a near daily basis.

I’ve used generative AI applications pretty darn heavily over the last 5 months… and I am incredibly impressed by the design engineers’ and product peoples’ ability to have launched products that are more or less producing unbiased and harmless content. Of course, there are exceptions, which I will discuss in later newsletters and in my upcoming books and courses, but for now… let me just say: Hey, generative AI outputs aren’t perfect, but their builders have their hearts in the right place when it comes to ethical issues, and I see that generally reflected in the model outputs I get on a near daily basis.

Here’s the process that’s used to collect and prepare human feedback for use in fine-tuning LLMs.

Fine-Tuning LLMs with Human Feedback

Here’s the basic process for (RLHF):

1. Choose an initial model.

2. Use a prompt dataset to generate multiple model completions.

3. Establish alignment criteria for the model.

4. Have humans rank the model’s output based on this criteria.

5. Gather all human feedback.

6. Average and distribute feedback across multiple labelers for a balanced view.

7. Feed that into the LLM to fine-tune its output so that they more closely align with human values.

Hairy Ethical Concerns

Addressing fairness in AI is challenging due to diverse global beliefs.

Gen AI companies say that they are selecting human labelers from a diverse pool, but… are they addressing their own personal biases in that selection process?

It's difficult to form a consensus when belief systems—religious, political, or otherwise—often conflict, with completely unreconcilable differences.

With AI set to revolutionize various sectors, it's crucial to include diverse perspectives in its development to avoid perpetuating unfairness.

Early intervention is essential for a future that benefits everyone - but I do not recall myself or anyone that I personally know getting to have a seat at the decision-making table. Considering that these technologies are in the process of upending the digital world in irrevocable ways, and that these changes will impact the lives of my children and generations forth, the fact that the voices of everyday people, like me and you, are not at all considered … it doesn’t seem right IMHO.

Wanna learn from the course too? Be my guest! Generative AI with LLMs class on Coursera…

The weekly freebies (or low-cost) spotlight

I have two fun features for you here this week:

1. You’re invited to join our free 100 Days Of Generative AI learning challenge.

Join the LinkedIn Group here to get started.

2. Break into data science with coaching and support from Dr. Adam Ross Nelson.

Adam is offering a sale right now where you can get all of his books and companion courses on how to break into the data science field, all for less than $50! If you’re still a data science newbie and looking to make your break, check these out!

Happy learning and I hope to see you inside our free 100 Days To Gen AI Challenge!

Yours Truly,

Lillian Pierson

Shop | Blog | LinkedIn

PS. If you liked this newsletter, please consider referring a friend!

Disclaimer: This email may include sponsored content or affiliate links and I may possibly earn a small commission if you purchase something after clicking the link. Thank you for supporting small business ♥️.