Meta's AI Secrets Just Leaked

Advertise here | 10-min Read

Welcome to our Monday edition!

In today’s menu:

• AI Inspirational Quote

• Leaked: Meta AI Training

• Build Clean HTML & CSS Like a Pro

• Viral Thread

• OpenAI Smartest AI Lies More

• Top Rated AI Tools

• Byte Sized news

• Meme

• How to Create Conversational Branches to Explore Ideas

AI INSPIRATIONAL QUOTE

LATEST NEWS

Leaked: Meta AI Training

Meta’s internal AI training secrets leaked in court documents

A high-profile lawsuit against Meta has revealed internal documents detailing how the company trains its AI models, including Llama. These documents highlight a process called Ablation, used to test which types of data improve model performance. Some of these experiments even involved illegally distributed books.

• Meta tested its AI using copyrighted books from LibGen to measure performance impact

• Adding science, tech, and fiction books improved benchmark scores by up to 6%

• Over 100 ablation tests were run during Llama’s development, but the results stayed private

These leaks reveal how AI models like Llama rely heavily on real-world written content, including possibly unauthorized materials. As researchers push for fair compensation and data transparency, these documents could lead to major changes in how AI companies handle training data and credit original content creators.
Continue reading…

Learn AI in 5 Min a Day

Start learning AI in 2025

Everyone talks about AI, but no one has the time to learn it. So, we found the easiest way to learn AI in as little time as possible: The Rundown AI.

It's a free AI newsletter that keeps you up-to-date on the latest AI news, and teaches you how to apply it in just 5 minutes a day.

Plus, complete the quiz after signing up and they’ll recommend the best AI tools, guides, and courses – tailored to your needs.

Sign up to start learning.

Build Clean HTML & CSS Like a Pro

Viral Thread

OpenAI Smartest AI Lies More

OpenAI's new reasoning models show better performance, but with a big accuracy problem

OpenAI recently introduced two new reasoning models, o3 and o4-mini, that outperform earlier versions in coding and math tasks. But tests show they also generate more hallucinations, making up facts, than previous models like o1, o1-mini, and even GPT-4o.

• o3 hallucinates on 33% of PersonQA questions, double the rate of o1 (16%) and o3-mini (14.8%)

• o4-mini performs worse, hallucinating in 48% of PersonQA cases

• OpenAI admits it still doesn’t fully understand why hallucinations increase as reasoning improves

While these models are smarter in some areas, they also make more false claims. That’s a big issue for users who need accurate outputs. It shows that progress in AI doesn’t always mean better reliability, highlighting the need for features like web search integration to keep responses grounded in truth.
Continue reading…

Top Rated

Google Whisk 2.0: Turn images into eight-second animated clips

404tomb: Digital tombstones for dead startups

Flow: A FOSS expense tracker that focuses on privacy and UX

Mirror: Track habits, log activities, build streaks & view insights

Controol: Focus on what you can spend, not what you already did

Ship Faster, Make more Money

Build Shopify apps with Gadget, the all-in-one development platform with an AI code assistant.

Get everything you need to ship in minutes, bundled in one tool:

• Full-stack infra: Node.js backend, Postgres DB, and React frontend

• Serverless & performant: Automatically hosted and scaled as needed Instant

• Shopify connection: OAuth, webhooks, rate limiting all handled

• Expert AI assistant: trained on your infra, code, and backend best practices

Code, debug, and scale faster than ever with Gadget.

Build your first app for free

8020AI Picks

Space: SpaceX launches cargo Dragon to ISS with additional crew supplies
Read more…

Crypto: Gold hits new ATH at $3,390, is Bitcoin close behind?
Read more…

Gaming: Grand Theft Auto Fans Want to See 1 Big Change in GTA 6 Online
Read more…

Don't miss: Google is in more danger than ever of being broken up
Read more…

How to Create Conversational Branches to Explore Ideas

In this tutorial, you’ll learn how to use Google AI Studio’s new branching feature to explore different ideas by creating multiple conversation paths from one starting point, without losing context.

Step-by-step:

1. Go to Google AI Studio and choose your preferred Gemini model from the dropdown.

2. Start a conversation. When you reach a point where you want to try a different direction, stop there.

3. Click the three-dot menu (⋮) next to any message and select “Branch from here.”

4. Switch between branches using the “See original conversation” link at the top.

Create branches at key moments to test different AI responses to the same problem, no need to start over.

Hey there! Quick ask 🙋‍♂️

If this email landed in your 'Promotions' or 'Spam' folder, could you move it to your Primary Inbox?

I work hard to deliver top news, tips, and tutorials to save you time. But if these emails get lost in the shuffle, we both miss out.

By moving it to your Primary Inbox, you’ll never miss a beat on the latest updates. Let’s stay in the loop and keep growing together! 🚀

FEEDBACK

How was your experience with 8020AI today?

Login or Subscribe to participate in polls.

If you have specific feedback or anything interesting you’d like to share, please let us know by replying to this email.

👋 THAT’S A WRAP

SPONSOR US

Get your business in front of over 66k+ AI professionals

Our newsletter is read by thousands of tech professionals, investors, engineers, managers, and business owners around the world. Get in touch today.

Or Email: [email protected]