Welcome to The Upgrade

Welcome to my weekly newsletter, which focuses on the intersection of AI, media, and storytelling. A special welcome to my new readers from CBS, UC Berkeley, UC Irvine, and many other top organizations — you’re in good company!

In today’s issue:

• The Week’s Top AI Stories 📰

• 🎓 AI Fundamentals now 30% Off!⚡️

• 🎙️The Big Interview: Founder William Shields on growing an AI agency and multi-agent frameworks

The Week’s Top AI Stories

Top AI Headlines

• Google CEO Calls AI Chatbot Responses Biased and Unacceptable — The Wall Street Journal

• Elon Musk Sues OpenAI and Sam Altman for Violating the Company’s Principles — The New York Times

• Welcome to the Era of BadGPTs — The Wall Street Journal

• Apple may soon be feeling the heat from investors over its slow rollout of buzzy AI products — Business Insider

• Robotics startup Figure raises $675 mln from Microsoft, Nvidia, OpenAI — Reuters

Regulation & Policy

• Political consultant behind fake Biden robocalls says he was trying to highlight a need for AI rules — AP

• Salesforce's top AI ethics leader says she's 'optimistic' on the path to U.S. regulation — CNBC

Ethics & Safety

• Police open criminal investigation into Beverly Hills middle school AI-nude photos incident — Yahoo News

• Google abandoned "don't be evil" — and Gemini is the result — Nate Silver (Opinion)

• AIs get better at maths if you tell them to pretend to be in Star Trek — The New Scientist

• Stanford study outlines risks and benefits of open AI models — Axios

Legal & Copyright

• Generative AI Is Challenging a 234-Year-Old Law — The Atlantic

• OpenAI says New York Times 'hacked' ChatGPT to build copyright lawsuit — Reuters

In the Workplace

• All the tech layoffs are because AI is like ‘corporate Ozempic’ — Fortune

• Apple Gives Up on Apple Car, Moves Team to Generative AI — Futurism

• Why Google's 'woke' AI problem won't be an easy fix — BBC

🎓Grab your discounted seat now! 

Ready to level up your AI knowledge? Want to become the AI lead on your team? Join me and an amazing group of marketing, communications, and journalism professionals this spring! ⚡️

The next cohort starts on Monday, March 11th at 7pm ET / 4pm PT. Eight spots are left, so don’t wait! Grab your discounted ticket using the code below — valid until Friday, March 1st, 2024!

30% OFF NOW WITH “THEUPGRADE30”

💡AI Founder William Shields on AI Agents

William “Wes” Shields is the founder of Synthminds, an agency partnering with organizations to strategize, design, build, and implement custom AI solutions that solve complex business challenges.

Note: This interview has been edited for brevity and clarity. Disclosure: Sythminds and The Upgrade are partnering on several projects, including one with MindStudio.

Peter: Can you share Synthminds' backstory and how you started in AI, considering your machine learning and computer science expertise?

Wes: I shy away from calling myself an expert because AI is constantly changing. I'm more of an AI explorer. The pace at which new research and tools are released is overwhelming, making it essential to stay humble and curious.

Peter: I like your perspective. The entire field is moving super fast right now. It’s hard to keep up even for those of us watching it daily full-time. How did you start exploring AI, and what led to the creation of Synthminds?

Wes: Synthminds came to be through a collective curiosity for AI among like-minded individuals. My background is in business and operations research, which led me to machine learning and computer science. After noticing a gap in my professional application of these skills, I sought out others passionate about AI. This was around the time ChatGPT was released in 2022. We began sharing our research and insights online, which eventually led me to freelance on Upwork and start SynthMinds, without any grand plans initially. However, momentum started building from there.

Peter: How did your momentum translate into growth for Synthminds?

Wes: Our growth was pretty organic. My podcast partner, Goda Go, gained a significant YouTube following, and another partner, Joe Rosenbaum, created a popular prompt that got OpenAI's attention. This visibility, along with our efforts on Upwork and engagements in AI conferences, significantly expanded our client base, including Fortune 50 companies. We were recognized in Upwork's expert vetted program, marking us as OpenAI experts, which opened more doors for us.

Peter: It sounds like Synthminds has had an unconventional path to success. How is your team structured?

Wes: We have about 10 to 12 core contributors, who are part-time but deeply involved in our projects. While we're moving towards more full-time positions, Gota, for instance, creates content full-time for her YouTube channel, which benefits us all. Our journey has been unique, and we're still exploring what it means to be an AI integrator, doing both training and some software development.

Peter: Tell me about the consulting and AI development projects you're working on at Synthminds, Wes.

Wes: We're handling a mix of projects, about half focused on integration work. This involves taking existing AI tools or APIs, wrapping them to meet specific business needs of our clients, often using low code and no code platforms like Mind Studio, which we're affiliated with. This approach allows individuals from various backgrounds to automate AI processes and potentially start their own businesses. Beyond integration, we're involved in bespoke development, creating SaaS platforms and deploying models for universities, alongside prompt engineering consulting. We've also collaborated on a unique project, a polymorphic multi-engined orchestration framework, which is essentially a comprehensive software suite for creating an entire AI organization.

Peter: That last one sounds incredibly complex... Can you break down what that framework entails?

Wes: Essentially, it's about building AI workflows that can make decisions and perform tasks with minimal human input. Currently, most AI tools require users to define logic explicitly. Our multi-agentic framework aims to understand context and user goals, then automatically creates the logic path, collaborating with multiple AIs to achieve those goals. It's very forward-thinking and not something we introduce to all clients due to its complexity.

Peter: Fascinating. That sounds like AutoGPT.

Wes: It's similar in concept. Auto GPT involves multiple AIs collaborating, where one AI generates a prompt for another to execute. However, our framework aims for a higher level of orchestration, ensuring each AI understands its role within the system, enhancing efficiency and effectiveness in achieving set objectives.

Peter: Who do you typically work with on these projects?

Wes: Our clients span various industries, including tech, aviation, and microprocessor technologies, as well as large accounting and consulting firms seeking AI training and integration solutions.

Peter: I know your team also does a lot of live, cohort-based trainings too. Tell me about that.

Wes: Wes: We're partnered with Uplimit.com, created by some original founders of Coursera. They started Uplimit after realizing the low completion rates of Coursera courses, with only about 10% of people finishing what they start. Uplimit aims to tackle this with an AI-enhanced learning management system (LMS) that offers courses in a cohort style, featuring live instruction sessions, group interactions via Slack, office hours, study groups, and projects. This approach has significantly improved course completion rates to about 85%.

Peter: That sounds incredibly engaging.

Wes: Absolutely. The community aspect of Uplimit is key, especially for our asynchronous courses. We organize them in blocks, incorporating weekly office hours and class-specific Discord or Slack groups. This supports our exploration into the vast potentials of AI. We offer a range of no-code AI courses, starting with "AI and ChatGPT for Everyone," which is ideal for beginners. For those looking to dive deeper, we have an advanced prompt engineering course led by my partner Joe, focusing on maximizing the potential of language models.

Peter: What about courses for more advanced topics?

Wes: We're excited about upcoming courses, including one on Image Generation that Goda and I are developing. Additionally, this summer, we're launching an intro and an advanced data analytics course with ChatGPT. I aim to condense a two-and-a-half-year graduate degree in operations research into a five-week course, focusing on asking the right questions without involving math or code. Our approach has garnered interest from the National Operations Research and Management Sciences Society, where I'll be presenting a paper this year.

Don’t be shy—hit reply if you have thoughts or feedback. I’d love to connect with you!

Until next week,

Peter

Psst… Did someone forward this to you? Subscribe here!