Welcome to the odo newsletter—your weekly digest on AI for product builders.

Be in the Know 👀 

• Stability AI Launched Stable Audio: Stability AI launched a new service that generates short music and sound effects based on simple text prompts. Give it a try!

• Roblox Launched an AI Chatbot: The new Roblox Assistant allows creators to build virtual worlds through text prompts.

• Salesforce Launched More AI Products: Salesforce launched two new products: Einstein Copilot, a conversational AI assistant to assist users' workflow; Einstein Copilot Studio, a platform to build AI-powered apps within Salesforce.

• China's Baidu Launched Ernie Bot: Ernie Bot can generate text and images in response to questions and prompts by users. It would be interesting to see how it navigates conversations about censored or politically sensitive topics. These are some early reviews.

• Adobe Firefly Now Available: Adobe's flagship AI product is generally available to its Enterprise clients.

• Free Audio Books via Project Gutenberg: Project Gutenberg teamed up with Microsoft and MIT to create thousands of free audiobooks. You can now listen to an AI reading Romeo and Juliet.

News Deep Dive 🤿 

⚖️ OpenAI’s Response to Copyright Lawsuits from Prominent Authors

OpenAI responded to two class-action lawsuits from authors who claimed that ChatGPT was illegally trained on improper copies of their books.

OpenAI argued that copyright law is supposed to protect the way authors express ideas, instead of “the underlying idea itself, facts embodied within the author's articulated message, or other building blocks of creative.” It cited a copyright case on Google Books where “statistical information” about a book such as word frequencies and syntax patterns were determined to be beyond the scope of copyright protections. Hence, OpenAI claimed that creating “wholesale copies [of a work] as a preliminary step to develop a new product” is not an infringement of copyrights.

It seems far-fetched to claim that every single output from OpenAI constitutes copyright infringement of all the inputs that the models have been trained on. But what if the output is a close approximation of one of the inputs? How does one define the degrees of approximation? It should be interesting to see how the arguments play out in the court, especially given the potential economic and societal impact of such guardrails.

If the outcome is too favorable to AI companies, we may undermine economic incentives to create new and unique output. What kind of long-lasting impact might it have on the advancement of art, and what it means to be human? If too favorable to the authors, we may hamper advancements of technologies that can have transformative impact on society's wellbeing.

Product Resource ☀️ 

Lenny Rachitsky interviewed Oji Udezue, CPO of Typeform, who has extensive experience in product-led-growth from his time at Twitter, Atlassian, and Calendly. Oji shared tips on how to identify the “sharpest” problems to work on, and how to use “forest time” to provide clarity on your own work. We enjoyed the podcast very much, and wanted to pass it along!

Job Postings 💼 

• Product Manager, Hebbia AI (AI productivity suite)

• Senior Product Manager, AI Assistants and Authoring Experience at Databricks

Disclaimer: The job postings listed in this newsletter are for informational purposes only. We are not endorsing these positions.

Before you go 💨 

Want a taste of the future? Coca cola just launched Coca-Cola Y3000, a soda co-created between humans and AI to taste like the year 3000 👀. Is this just another “slap AI to anything?” Or will we be mind-blown by the taste of the future?

If anyone gets their hand on the drink, please let us know!