Here's what caught my eye in the world of AI this week:

Deep Byte:

1.  In the AI space, everyone's buzzing about what will be the first killer Gen AI app. Google and Microsoft have integrated large language models (LLMs) into their products, but this feels more like "bolt-on" AI. If I had the definitive answer, I'd be building it now!

   Benedict Evans offers an intriguing perspective in his piece, "Building AI Products." He emphasizes that successful AI products should cater to users, not expect them to adapt to the technology. Over the last 50 years, consumer computing has thrived by moving toward the user, not by making them learn command/template shortcuts. Who remembers WordPerfect? It's a perfect example of how not moving toward the user can lead to obsolescence.

AI applications must be intuitive and user-friendly to drive widespread adoption. Look for AI that enhances the user experience without requiring a steep learning curve. A company that is IMHO is doing a great job is Notion.

Notion AI (Affiliate) is revolutionizing how we organize and manage information. It leverages AI to automate tasks, suggest content, and improve workflow efficiency.

• Automation: Handles routine tasks, freeing up time for more strategic work.

• Content Generation: Assists in creating documents, notes, and plans with ease.

• Strategy: Moving Towards the User

When implementing AI in your company, prioritize tools that are easy to use and enhance the current workflow. Avoid complex solutions that require extensive training or change how employees work.

• User Training: Invest in training programs that make the transition smooth.

• Regularly collect and incorporate user feedback to refine AI tools.

2. How LinkedIn went for from artisanship to industrial scale AI. Great Video talk and deck.

The Data is staggering:

• LinkedIn used to write over a petabyte of data per day, now they write over 2 petabytes of data per day. They went from 3 trillion to 5 trillion rows per day that are updated within this system. The peak has also increased. Now we ingest up to 80 gigabytes per second of mostly feature data, and upward of 150 million rows per second at peak.

3. Here are the three key takeaways from the podcast:

• Modality and Model Choice: Choose models based on the specific modality (text, image, audio) needed. For example, text models for text processing.

• Self-hosting vs. API Models: Self-hosting offers control over data privacy and cost but requires significant infrastructure and access to GPUs. API-based models are easier to use but may have privacy concerns.

• RAG Applications: Effective Retrieval Augmented Generation (RAG) apps rely on well-designed data pipelines and embedding searches rather than the specific generative model or vector database used.

Sponsored By: AI Quality Conference: Elevate Your AI Game

LAST CHANCE to join us on June 25th in San Francisco for the AI Quality Conference! Rub shoulders with top experts from Cruise, NVIDIA, Google, UBER, and more —practitioners who are shaping the future of AI. Gain firsthand insights into building rigorous, reliable, and scalable AI solutions. Network where it counts and transform your knowledge into action. Don't miss out—use this exclusive code to save $300 on your ticket “JTA300” ! Let me know if you are coming, I love to meet you.

4. We previously discussed LLMOps in a previous newsletter, I thought it would be helpful to take a step back. Here is an excellent primer on CI/CD , which is crucial to understand, even if you're not implementing all the details. A solid foundation in these concepts allows you to create scalable, transferable solutions. Learning how to build, automate and delegate solutions is key to avoiding long-term support burdens and ensuring you can move on to your next challenge seamlessly.

5. Verizon’s strategy to navigate the "Wild West" of AI development responsibly.

6. Friday Funnies 🤣 

Send me your best AI meme and I will post the Top 5 and put it to a vote by our readership! I am sure anyone can do better! Winner gets a bottle Moet Champagne! Have to get at least five memes for there to be a competition.

7. Autonomous Vehicles Are Great at Driving Straight But humans still do better at turns and in the twilight

Autonomous vehicles excel at driving straight, but humans still outperform them in complex scenarios like turns and twilight. Despite occasional fatalities, banning them isn't the solution. Voltaire's "Perfection is the enemy of the good" from Philosophy 101 resonates here; waiting for perfection means more lives lost. According to IEEE Spectrum, while autonomous vehicles face challenges, they continually improve and often surpass human drivers in safety. We had a fruitful LinkedIn discussion on this, emphasizing progress over perfection. In business, where lives aren't at stake but brand and reputation are, does this principle still hold?

What are your thoughts?

8. General Intelligence (2024)

True expertise lies in explaining complex concepts in a way that anyone can understand. The author, J. Becker, an employee at OpenAI, shows his passion as he delves into what is required to create a generally intelligent agent, highlighting three key ideas:

1. Interaction and Observation: Embodiment—the ability to perceive and interact with the natural world.

2. Robust World Models: Mechanisms for quick, accurate inference, akin to human intuition or “System 1 thinking.”

3. Deep Introspection: Mechanisms for reasoning or “System 2 thinking.”

According to Becker, these advancements will help us achieve AGI in just a few years. A must-read to gain a better understanding of AGI as defined by Becker.

👇 If you like my newsletter I think you will also like 👇

Learn AI in 5 Minutes a Day

AI Tool Report is one of the fastest-growing and most respected newsletters in the world, with over 550,000 readers from companies like OpenAI, Nvidia, Meta, Microsoft, and more.

Our research team spends hundreds of hours a week summarizing the latest news, and finding you the best opportunities to save time and earn more using AI.

Sign up with 1-Click