Welcome to the odo newsletter—a free, weekly newsletter on AI for product builders. We'll cover:

• AI-Created Art Ruled as Not Copyrightable

• OpenAI Enhances Customization and Safety

• Meta Raises the Bar for Text and Speech Translation

• …and more!

News of the week 🗞️ 

⚖️ Judge Rules AI-created Art is Not Copyrightable

A federal judge upheld a finding from the US Copyright Office that the copyright of AI-created art cannot be protected. Stephen Thaler had sued the Office for rejecting his attempts to copyright an AI-generated image. US District Judge Beryl Howell emphasized that copyright law has never been granted to work that was “absent any guiding human hand,” stating, ”human authorship is a bedrock requirement of copyright."

Looking pretty gray…In many instances, the line between human-generated and technology generated art is not always black and white. We expect this debate to continue until there is a common framework for defining this gray line. Keep in mind, we are still not great at telling whether a piece of work is AI-generated or human-generated, complicating the matter a tad bit.

⚡ OpenAI Enhances Customization and Safety

It's been a busy week for our friends at OpenAI. The company announced fine tuning capabilities for GPT-3.5 Turbo, allowing developers to customize this powerful model to improve steering, make output format more reliable, and customize tone, while saving costs. OpenAI also launched ChatGPT Enterprise, which offers enterprise-grade security and privacy—emphasizing that prompts and customer data will not be used to train OpenAI models—alongside other benefits.

Defensive move against competition…Since OpenAI launched Chat-GPT last year, a flurry of open-source and closed LLMs entered the chat. These recent improvements will likely make a step-wise change for customers by offering more functionality (through fine-tuning) or addressing barriers for adoption (security and privacy through ChatGPT Enterprise).

🗣️ Meta Releases A Multilingual and Multitask Translation Model

Continuing a furious pace of releases, Meta published a new AI model SeamlessM4T for realtime translation that works with both text and speech in almost 100 languages. This is the first time a single model is able to do all of those tasks. Meta claims this reduces errors and delays while increasing efficiency and quality. Try out the demo yourself.

Lowering the barrier to accessbility…This release signals a world where realtime translation is much more accessible to us product developers. Translation has compelling use-cases for connecting people across language barriers. Additionally, we're excited at the prospect of using this type of tool to improve the accessibility of our products. It will lower the barrier to show interfaces in the user's preferred language and provide speech interfaces for those with vision impairments.

AI product highlight ✋ 

There's buzz about how generative AI is coming for your job. But did you know that you can use generative AI to find jobs? If you're in the market, give AutoApplyAI a try—it will auto-fill applications, making it appear as if you applied.

Disclaimer: We are not getting paid to highlight the product. We just think it’s cool and want to share!

Before you go 💨 

Are you interested in saving the world by helping defend US government's critical digital infrastructure? The Defense Advanced Research Projects Agency (DARPA) is hosting an AI Cyber Challenge (think of it like a hackathon), alongside Anthropic, Google, Microsoft, and OpenAI.

There's nearly 💸$20 million💸 in prize money. If you win, don't forget that you heard this from us!