OpenAI's 'Strawberry' Agent, Perplexity revenue share, Thrive AI Health // AI Inside #26

Introduction

Hello everybody and welcome to yet another episode of AI Inside. The show where we take a look at the AI that is sprinkled into so many different places in the world of technology. I'm Jason, and we’re here to shine a light on it. As always, I'm joined by my co-host Jeff Jarvis. Jeff recently went back to the roots of typography, which is completely Jeff Jarvis. However, we're back now delving into some exciting AI news and developments.

OpenAI's 'Strawberry' Project

OpenAI is again making headlines with its new project, Strawberry. According to Reuters, Strawberry models will not just answer queries but also autonomously navigate the internet to perform deep research and penetrate science and math questions beyond current model capabilities. This development aligns with agentic AI, enhancing the ability to think ahead and search for necessary information.

OpenAI has also indicated it might be nearing the second level of five defined steps toward achieving Artificial General Intelligence (AGI). The five levels are:

Chatbots: Conversational language AI.
Reasoners: Human-level problem-solving.
Agentic: Systems that can take actions.
Innovators: AI that can aid in invention.
Organizations: AI performing entire organizational work.

While it's an exciting sentiment from OpenAI, critics like Benedict Evans urge caution, questioning whether these pre-defined steps and sequences hold any definitive basis.

Thrive AI Health

In collaboration with Ariana Huffington’s Thrive Global, OpenAI announced Thrive AI Health, an AI-powered health coaching startup trained on peer-reviewed science and user data. However, using AI for health recommendations comes with risks. Critics argue that while it might provide motivational advice effectively, handling medical data and advising accordingly needs more scrutiny and a robust system of checks and balances.

Washington Post's Climate Answers

The Washington Post has launched 'Climate Answers,' an AI chatbot integrating the publication's articles about climate change. Utilizing services like Chat GPT and Meta Llama, the chatbot aims to provide responses backed by journalistic rigor. This development seems to be a cautious attempt to explore AI in journalism, focusing on credible information dissemination.

Last month, the Post also rolled out an AI summarization tool on about 10% of its stories. The engagement with AI in editorial processes is being closely monitored to gauge effectiveness and utility.

Microsoft and Apple Backtrack

Both Microsoft and Apple have stepped away from holding board seats at OpenAI. Regulatory scrutiny and antitrust concerns are significant factors influencing this decision.

Perplexity's Revenue Share

Perplexity AI is starting a revenue share program with web publishers next month. They plan to run ads with search queries and share the revenue with publishers contributing source input for answers. This model differs from OpenAI's approach, aiming for equitable merit-based revenue distribution.

Selective Indexing and AI's Web Impact

Vincent Schmalbach's article highlights Google's shift to selective indexing, defaulting to non-indexing unless content shows genuine need, authority, and uniqueness. This change seems essential in maintaining quality amid growing AI-generated content but poses challenges for new publishers to gain visibility.

Ziff Davis' AI and Editorial Integrity

Ziff Davis, a media group incorporating Mashable, Lifehacker, and PCMag, signed a new contract ensuring that staff won’t be laid off or have salaries reduced due to generative AI. Editorial review by humans remains a crucial part of this new AI-driven content process.

Google's Gemini and AudioShake with Disney

Google's Project Gemini, teased at Google I/O, is now being tested, allowing users to create video presentations using their Google Docs and Slides. Meanwhile, Disney collaborates with AudioShake to decompose its music catalog using AI, especially useful for older recordings where multi-tracks aren’t available.

Conclusion

From advancements in OpenAI's agentic models to innovative uses of AI in health coaching and journalism, the landscape of AI is rapidly evolving. Much like the ancient struggle between innovation and traditional methods, today’s AI breakthroughs present exciting yet cautious optimism.

Keywords

AI Inside
OpenAI Strawberry
Agentic AI
Thrive AI Health
Washington Post Climate Answers
Perplexity AI Revenue Share
Selective Indexing
Google Gemini
AudioShake
Disney Music Catalog

FAQ

Q1: What is Project Strawberry by OpenAI? A: Project Strawberry aims to develop AI models capable of autonomous internet navigation, performing deep research, and answering complex queries in science and mathematics.

Q2: How is OpenAI defining its pathway to AGI? A: OpenAI has defined five levels leading to AGI: chatbots, reasoners, agentic systems, innovators, and organizations performing comprehensive tasks.

Q3: What does Thrive AI Health aim to achieve? A: Thrive AI Health aims to use peer-reviewed science and personal health data to provide AI-powered health coaching.

Q4: What is Perplexity AI’s new revenue share model? A: Perplexity AI plans to run ads with search queries and share the revenue with web publishers contributing source input for the AI-generated answers.

Q5: What impact does Google's selective indexing have? A: Google’s shift to selective indexing aims to maintain quality in search results by only indexing content showing uniqueness, authority, and genuine need, posing challenges for new publishers.

Q6: What is Ziff Davis' strategy regarding AI? A: Ziff Davis has signed a contract ensuring no layoffs or salary reductions due to generative AI and mandates human editorial review for all AI-driven content creation.