OpenAI's New Fact-Checking AI Model: Strawberry

OpenAI is poised to unveil its next major product, Strawberry, an AI model with self-fact-checking capabilities, according to reports.

• Strawberry is expected to be released as a stand-alone product, but will be integrated with ChatGPT.
• This model is reportedly superior to other top-tier generative AI models, including OpenAI's GPT-4o, in programming and math.
• However, Strawberry comes with a drawback: it's significantly slower, taking 10-20 seconds to respond to a single query.

The Generative AI Race Heats Up

The emergence of Strawberry highlights the intense competition in the generative AI space.

• OpenAI is facing pressure to deliver results, having invested billions in AI training and staffing.
• Investors and potential backers are eager for a return on their investment.
• Despite the urgency, releasing an unpolished model like Strawberry, especially with a higher price tag, may be risky.

Apple's AI Progress: A Focus on ChatGPT Integration

Apple has been making significant strides in AI integration, but its approach seems to be primarily focused on ChatGPT.

• Apple Intelligence, Apple's suite of AI-powered services, is heavily reliant on ChatGPT for its functionality.
• Although Apple previously hinted at collaborations with other AI models, including Google's Gemini, there have been no concrete announcements.
• This raises questions about the status of these potential partnerships and whether Apple's investment in OpenAI has impacted its relationships with other AI players.

Apple's AI-Powered Visual Search: A Step Forward

Apple's new iPhone 16 features "visual intelligence," which essentially combines reverse image search with text recognition.

• This feature is powered by collaborations with third-party companies, including Google.
• Apple is making a noticeable effort to integrate AI into its devices and services.

Audible Leverages AI for Audiobook Generation

Audible, Amazon's audiobook platform, is using AI trained on professional narrators' voices to create new audiobooks.

• This raises questions about the ethics of using AI to generate content based on human voices.
• Narrators will be compensated for audiobooks created using their AI voices on a title-by-title basis.

Research Paper of the Week: AlphaProteo for Protein Binding

Google's DeepMind has developed AlphaProteo, an AI model that predicts protein binding to target molecules.

• AlphaProteo generates candidate proteins that bind to a molecule at a specified binding site.
• In tests, AlphaProteo outperformed previous methods in creating protein binders with higher binding affinity.
• This model has successfully developed a binder for a protein associated with cancer and complications arising from diabetes (VEGF-A).

Model of the Week: OLMoE - Open Source Mixture-of-Experts Model

The Allen Institute for AI (AI2), along with Contextual AI, has developed OLMoE, a 7-billion-parameter mixture-of-experts (MoE) English-language model.

• OLMoE outperforms models like Meta's Llama 2, Google's Gemma 2, and Mistral's Mistral 7B in several applications and benchmarks.
• The model and its code are openly licensed, making it accessible for research and development.

Apple's Event: AI Focus and Uncertain Partnerships

Apple's recent event focused on new iPhones, Apple Watch models, and app announcements, with Apple Intelligence being highlighted.

• ChatGPT is a key component of Apple Intelligence, but there were no new announcements regarding AI partnerships beyond the OpenAI deal.
• This raises questions about the status of previously confirmed plans to collaborate with other AI models, including Google's Gemini.
• The lack of clarity about these partnerships suggests ongoing uncertainty, potential setbacks, or perhaps unintended consequences stemming from Apple's investment in OpenAI.