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

Be in the Know 👀 

• OpenAI to Unveil DALL-E 3 Soon: DALL-E 3 will be built natively within the ChatGPT interface. The sample images look much more vivid than its predecessor. It will be interesting to see how it stacks up against companies like Midjourney and Stable Diffusion.

• Microsoft to Roll Out Microsoft Copilot: Microsoft Copilot will be available in Windows 11 starting today (Sep 26). It will bring AI-powered experiences to apps like Paint, Photos, and Clipchamp. Bing will support the latest DALL-E 3 model from OpenAI. Microsoft 365 Copilot will be available in November.

• TikTok Launches Labels for AI-generated Content: Creators can start labeling TikTok content that has been wholly or significantly generated by AI as such. TikTok plans to support automatic labeling soon. With this, TikTok is the first major social media platform to support AI-generated labels.

• Proprio Uses AI to Assist Surgeons with Spinal Operations: Seattle-based Proprio started training surgeons to use its FDA-cleared device to navigate spinal surgeries. The technology is supposed to help surgeons map and visualize the surgery site.

• Google and the Department of Defense Build AI-powered Microscope to Spot Cancer: Augmented Reality Microscope, or ARM, could become a useful tool for pathologists who want a second opinion when identifying cancer.

• OpenAI launched Red Teaming Network: OpenAI formalized its efforts to assess and mitigate risks in their AI models with outside experts.

• GitHub Copilot Chat beta is now available to all: All individual users (not just Business users) can now have access to GitHub Copilot Chat across Visual Studio and VS Code.

News Deep Dive 🤿 

🧙‍♀️ Google's Bard Extension Sounds Great…But What is it like in Reality?

Google has just announced Bard Extensions, which integrates with Google apps and services to help you find the most relevant information. According to their press release, Bard can find information from tools like “Gmail, Docs, Drive, Google Maps, YouTube, and Google Flights and Hotels…even when the information you need is across multiple apps and services.” Given how many people use Google's apps and services in their day-to-day, the extension sounds like it could become a very powerful personal assistant. If executed well, this could give Google a huge leg up compared to competitors that are purely focused on LLM technologies, like OpenAI.

I (Yooni) became really excited, and decided to give Bard a simple task.

I thought this would be fairly straightforward, given that 1) I gave explicit dates to search ("2022") and 2) provided keywords (like weddings and registry) that are strongly associated with what I'm searching for and not likely to randomly appear in other emails.

Results: F. 🤦

The only search result that Bard returned was one wedding that I attended in 2023, and it didn't even show me how much I gave in registry. It failed to show any weddings I had attended in 2022, despite the fact that when I search for “weddings 2022” in Gmail, I find a number of wedding invites and confirmation emails.

Google seems to have a long way ahead in making Bard a truly useful personal assistant for people. It's interesting that Google decided to roll out the product so early on, given its obvious shortcomings. We're excited to see how much and quickly Google can improve upon Bard. For now, we're likely going to still rely on good ole' search for finding information.

Product Resource ☀️ 

Amazon's Working Backwards process is a pretty well-known artifact. Before building the product, you can write an internal press release announcing the finished product to your to-be customers. This helps you crystalize what customer problems you are solving and how your customers would describe the solution. Ian McAllister, a former GM at Amazon, has shared the template and a mock example here if you want to take a look.

Job Postings 💼 

• Senior Product Manager, Siri and AL/ML at Apple

• Product Manager, Generalist at Anthropic

• Senior Product Manager, Artificial Intelligence & Human Health Research at Mount Sinai

• Lead Product Manager, AI Applications at Zoom

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

Before you go 💨 

Having trouble prompting LLM to do things? A recent study has shown that telling LLM to “take a deep breath” or encouraging it to “think step by step” has led to an increase in scores for solving math problems. Maybe humans and LLMs are not too different after all?