Welcome back to AI for 30! This week's edition is crammed with nuggets of AI wisdom and practical insights. We're kicking things off with a quick tutorial on leveraging AI to craft killer PowerPoint presentations—effortless efficiency at its best. Then, dive into the world of AI agents with a session from Andrew Ng, who breaks down their potential to transform how we work. We'll also dissect Stanford's 2024 AI Index Report, serving up the critical data and trends that are sculpting the AI field. On top of that, we're spotlighting Meta's latest innovation, Llama 3, setting new benchmarks in AI capabilities. And, of course, we’ll wrap up with our quote of the week to send you off inspired for the week ahead.

On the Agenda This Week:

• How to prompt a great PowerPoint.

• AI Agents - What is that? Andrew Ng Explains

• Breakdown of the Stanford 2024 AI Index Report

• Big News In AI: Meta Introduces Llama 3 

• Quote of the week

How to prompt a great PowerPoint in a snap.

This is a basic prompt to generate code to create a slide deck. Any additional context can be added to enhance the content.

Step 1: Use any chatbot you want: I used ChatGPT-4.

Step 2: Enter the prompt into the chatbot.

Step 3: Change the subject and number of slides to fit your needs. You can also add additional [placeholders like this that are interchangeable]

I want you to write me VBA code for a PowerPoint presentation about [subject]. You must fill in all the text with your knowledge, with no placeholders. I need [number of slides] slides.”

[subject] = Rolling Truckload consolidation 

[number of slides] = 7

Step 4: Copy the VBA code generated as the output.

Step 5: Open Powerpoint

• On the View tab, choose Macros.

• In the Macro dialog box, type a name for the macro.

• In the Macro in list, click the template or the presentation where you want to store the macro.

• In the Description box, type a description for the macro.

• Click Create to open Visual Basic for Applications.

• Paste the code we got from ChatGPT and press the play button, and you will get a PowerPoint.

Step 6: Watch the video I made - it might be easier if you’re a visual learner.

Agentic workflows with Andrew Ng during a discussion with Sequoia Capital. I know what you’re thinking; what does that even mean?

Andrew Ng, founder of DeepLearning.AI and Coursera and the leader of Google Brain, speaks at Sequoia Capital's AI Ascent about what's next for AI agentic workflows and their potential to significantly propel AI advancements.

What are AI agents or agentic workflows? Several different AI models perform individual tasks to help complete a more significant task. Think of them as a business with other employees performing different job roles.

A more detailed example might be using AI agents to help you prepare for a party. One agent lists and orders supplies; the other model makes the invite list, while the following model designs and sends invites. Together, they work toward one goal: planning a party.

Additional takeaways:

• Traditional vs. Agentic Workflows: We're moving past the days of one-and-done AI outputs. Traditional AI workflows operate on a single-shot basis—put in a prompt, get out a result. But agentic workflows? These systems iterate, react, and evolve like a pro tackling a problem.

• Enhanced Output Quality: Talk about leveling up—agentic workflows integrate feedback loops. This means the AI reviews and refines its outputs continuously.

• Application in Coding: Where do these workflows shine? In coding, for starters. AI in agentic workflows can write code, run tests, spot bugs, and fine-tune its work—all on a loop.

• Importance of Iterative Processes: Andrew Ng’s got it right—iterative processes are crucial. They bring AI workflows closer to human-like problem-solving, enhancing accuracy and adaptability for complex tasks.

• Future of AI Development: Looking ahead, it’s clear that agentic workflows are paving the way. They’re setting the stage for AI to edge closer to accurate generalized intelligence.

• Impact on AI Deployment: There’s a significant shift ahead in how we use AI. Agentic workflows focus on ongoing interaction and continuous improvement, moving away from one-off deployments.

The 2024 AI Index Report from Stanford University's Institute for Human-Centered AI.

_{This is quite the read—it's packed with insights and serves as a compass for where AI is heading. Here’s what you need to know about this year's edition, which is more robust and detailed than ever:}

You can also skim the report for yourself here

• Expanding Horizons: This latest report dives deep. It goes beyond the usual data, adding fresh perspectives on the costs of AI training and the tech’s ripple effects across science and medicine. For anyone involved in AI, from policymakers to developers, this report is a treasure trove of information crucial for navigating the future.

• Technological Strides and Stumbles: The advancements in AI are nothing short of impressive. GPT-4 and Gemini show capabilities that match—and sometimes exceed—human performance in various benchmarks. But it’s not all smooth sailing. The report candidly discusses the challenges, particularly AI’s struggle with complex reasoning and reliable fact-handling.

• Financial Trends and Regulatory Shifts: Private investment in AI might be cooling off, but the bucks flowing into generative AI are heating up like never before. Also, if you’re keeping an eye on the regulatory scene, the report notes an uptick in AI-related regulations. As AI becomes a more significant part of our lives, everyone from governments to the public.

• The Public’s Pulse: Speaking of attention, public awareness and concern about AI are rising. People are increasingly tuned into how AI might affect their privacy, jobs, and the truth about the information they consume. This growing concern shapes how AI is discussed in public forums and how regulations are framed.

Big News: Meta Introduces Llama 3

This week, there is some big news in AI. Specifically, Meta has rolled out an exciting new feature across its platforms. Here's what you need to know about the launch of Llama 3:

• Platform Integration: The new AI chatbot is integrated into the search bars of Facebook, Instagram, and WhatsApp.

• Powered by Llama 3: This AI assistant uses Meta's latest and most advanced model, Llama 3, acclaimed as the most powerful free and open-source model currently available.

• Key Features:

  • Interact with Meta AI via your feed, chats, and search bar.

  • Visit the browser version at meta.ai for more functionalities.

  • Access real-time web information.

  • Create custom GIFs and generate images as you type prompts.

• Technical Specs: Llama 3 has 70 billion parameters, with a future version planned to have 400 billion.

• Performance Claims: Meta suggests Llama 3 can compete with leading Google, Anthropic, and Mistral models across several benchmarks.

• Open Source Advantage: As an open-source model, Llama 3 can be adapted and fine-tuned by other companies, potentially matching the performance of GPT-4.

• User Base Impact: Nearly 3 billion daily users of Meta’s platforms will now have access to advanced AI tools, broadening AI interaction to a massive audience.

• Strategic Implications: By making Llama 3 open source, Meta not only fosters widespread adoption of its products but also pressures companies' business models relying on closed AI systems.

Meta's introduction of Llama 3 is set to redefine user interaction with AI and potentially alter the competitive landscape in AI technologies.