🍓OpenAI’s ‘Strawberry’ AI is coming...

Welcome back to a new edition of AI Odyssey!

TL;DR

• Adobe’s AI video tool is coming: Adobe’s Firefly will soon let users create AI-generated videos, built on licensed content for copyright safety.

• OpenAI’s ‘Strawberry’ model nears release: OpenAI’s new AI model, “Strawberry,” handles complex reasoning, setting them up for a competitive edge.

• Google’s AI makes podcasts from your notes: Google’s NotebookLM can now turn research notes into AI-generated podcast-style audio summaries.

• California’s AI bill divides tech: California’s AI bill could lead on regulation, but it’s dividing Silicon Valley over innovation vs. safety.

Sponsored by Pipedrive

Sell smarter and faster with Pipedrive AI.

Winning deals is as simple as the sales tools you use. Pipedrive’s AI-powered CRM helps you streamline success and make data-driven decisions so you can focus on one thing: selling.

Generate next-level insights with Pipedrive’s AI Sales Assistant. Uncover patterns, spot high-potential deals and identify priority actions to stay one step ahead.

Need help writing creative sales emails? Pipedrive’s one-click email creation saves you time and enhances your outreach. Enter any email prompt and generate compelling, personalized communications in seconds. Then, when the replies come rolling in, pinpoint key information with instant email thread summaries.

Explore Pipedrive AI today for smarter, faster sales.

1: Adobe previews Firefly for video creation

Adobe is set to release its generative AI video tool, Firefly, later this year, expanding its AI capabilities beyond still images.

Why it matters: Like its Firefly tools for images, Adobe’s video model is built on licensed content, ensuring a "commercially safe" option for professional creators. This is crucial for users seeking creative tools that respect copyright.

Driving the news: The Firefly video model will offer text-to-video and image-to-video options, allowing editors to generate cinematic video content. It will include controls for adjusting settings like camera angles and motion, with future integration into Premiere Pro via a "Generative Extend" feature.

What they're saying: Adobe’s digital media CTO, Ely Greenfield, emphasized that Firefly’s training respects content rights and integrates seamlessly into Adobe’s workflows, enhancing both creativity and efficiency for video editors.

2: OpenAI nears release of ‘Strawberry’ AI model with reasoning abilities

OpenAI is close to releasing its new AI model, codenamed "Strawberry," which can handle human-like reasoning tasks. This new model aims to solve multi-step problems, such as complex math and coding questions, marking a major step in AI's development.

Why it matters: AI reasoning is considered a key advancement, and competitors like Google and Anthropic are also working on similar capabilities. Strawberry’s release could give OpenAI an edge in the competitive AI landscape as it seeks to raise billions in funding.

What’s new: Strawberry’s "chain of thought" prompting allows it to pause and process related prompts before responding, improving accuracy in areas where models like ChatGPT currently struggle. For example, Strawberry can correctly determine that 9.9 is greater than 9.11, a task that often confuses existing AI systems.

What’s next: A limited release of Strawberry could happen as soon as this week, with OpenAI continuing to refine its ability to solve multi-step problems. The model has already impressed during internal demos, solving complex word and chemistry problems.

3: Google’s AI turns your notes into fake podcasts

Google’s NotebookLM app now lets AI create podcast-style discussions based on your research. The feature uses two AI hosts to "banter" about your notes, making connections between topics and summarizing your documents.

Why it matters: This AI-generated audio feature offers a fresh way to engage with your research. It builds on NotebookLM’s existing text-based summaries but takes it a step further with a conversational, podcast-like experience.

How it works: Using the Gemini AI model, the hosts summarize your material, adding lighthearted commentary. While the audio can make research more fun, there are quirks—like spelling out certain words or awkward phrases. Plus, the tone might feel off for serious topics like cancer or war.

Limitations:

• Takes a few minutes to generate.

• Only available in English.

• Accuracy isn’t always perfect.

You can try it by selecting "Audio Overview" within NotebookLM and seeing what the AI podcasters come up with for your notes!

4: California's AI bill is still dividing Silicon Valley

California's AI regulation, SB 1047, has split Silicon Valley, with Governor Gavin Newsom facing a September 30 deadline to sign or veto the bill.

Why it matters: With no federal AI regulation, California’s decision could set a national precedent. The bill requires big AI developers to run safety tests, assess risks, and implement kill switches for large models.

The divide: Supporters like Elon Musk say regulations are needed to ensure safety, while critics, including OpenAI and Meta, argue the bill could stifle innovation and push companies out of the state.

What’s next: If signed, California will lead in AI regulation. If vetoed, the debate over how to manage AI's rapid growth continues.

That’s a wrap! See you again for a fresh dose of AI Odyssey soon. 😎