Justin's AI Newsletter

The most interesting content, news, and insights in AI each week.

--

Hey friends 👋,

I am excited to announce that our startup, CassidyAI, is no longer in stealth mode. CassidyAI has been my focus for the past four months, and I truly believe that we are building one of the most powerful AI products for businesses.

Our goal with Cassidy is to make it easy to automate parts of your business using AI. With our no-code tooling, you can build custom AI assistants that are trained on your knowledge and data to automate functions across teams.

Currently, I am personally onboarding every company that joins, so I'm giving away early access slowly. My philosophy is to move slowly to move fast. We already have some amazing early customers, but there's a lot more to come. If you join the waitlist, don't worry if you don't hear from me right away. You'll have access soon enough.

---

STARTUP ADVICE

🚀 Moving Slow to Move Fast*

The Key to Early Stage Startups

For the past 4 months, I have been in stealth mode with Cassidy. This wasn’t for fear of someone stealing our idea, but because we wanted to do things right. Our goal was simple: to develop a product that genuinely solves users’ problems when they use it. And the reality is that, you need to make 100 customers happy before you can make 1,000,000 customers happy.

A lot of AI companies make this mistake. They rush out new tools that make big claims but can’t deliver. Whether it’s because the tech isn’t ready, or they just haven’t figured out how to make a great product yet, you can’t start scaling until you get it right.

Don’t get me wrong, you have to move fast. It’s exactly why we’re ready to publicly introduce Cassidy now. But it’s vital not to overlook the importance of those early days, taking care to ensure your first customers are genuinely happy. There’s no shortcut to genuine success; it’s all about building something people truly want.

---

TECH INSIGHT

🔒 Privacy is More Important Than Ever

Recently, over 100,000 ChatGPT accounts were breached and credentials were leaked onto the dark web from around the globe. Though it is important to note that OpenAI was not directly at fault for the incident (the breach stemmed from malware on users' computers), the event underlines the necessity of maintaining a high standard of data protection when using AI products.

The reality is that as we spend more time interacting with AI tools, they will become a trove of private and proprietary information. For both large enterprises and startups, it's important that the products they use provide a certain degree of data protection.

AI thrives on context, meaning that the more information it receives, the better it performs. Instead of worrying about what we can or can't give AI products, we should simply trust that our data is secure.

---

💡 This Week's Recommendations

🔗 The EU’s AI Act, Explained: The EU has recently introduced numerous regulations regarding AI. This brief overview provides a good summary of what they entail. Essentially, the regulations establish boundaries around certain use cases of AI, including face recognition and biometric analysis. They also prevent the creation of AI models that can generate illegal content and ensure published summaries of copyrighted content used to train models.

🔗 Meta announces new AI model that can replicate the voices of loved ones, but says it's too risky to release: Meta has developed a new generative AI model called Voicebox that can replicate voices and speak six languages. However, the company has decided not to release the model publicly due to the potential risks of misuse. Meta believes the technology could have many exciting use cases, such as allowing visually impaired people to hear messages from friends in their voices or enabling users to speak in foreign languages in their own voice. However, the risk of misuse in scams is significant.

🔗 Google’s “RoboCat” will power new general purpose robots: Google's DeepMind has developed an AI agent called RoboCat that can learn to operate different robotic arms and take on different tasks with as little as 100 examples. RoboCat can create its own training data to help sharpen its performance and learn much faster than other models. The potential applications of AI in the field of robotics are truly astounding.

🔗 The Next Token of Progress: 4 Unlocks on the Generative AI Horizon: The article highlights four innovations in generative AI that will unlock new possibilities soon: improved steering, enhanced memory capabilities, tool usage, and multimodality. These will enable better control over LLM outputs, personalization and tailored outputs, taking action on extracted information, and more complex and diverse outputs. These innovations will have significant implications on the future of AI.

---

Thanks for tuning in,

Justin

​Justin's Full Content Library | Previous Newsletters​

113 Cherry St #92768, Seattle, WA 98104-2205
​Unsubscribe · Preferences​