🗞 OpenAI's "Strawberry" Model: Next-Gen AI with Advanced Problem-Solving 🍓

🔎 The Latest on the AI Frontier:

• OpenAI's "Strawberry" Model: Next-Gen AI with Advanced Problem-Solving 🍓

• Nvidia Launches NIM Agent Blueprints for Streamlined Enterprise AI Development 🛠️

• Cerebras Unveils Ultra-Fast AI Inference Service to Challenge Industry Giants ⚡

• Google Meet Introduces AI-Powered Note-Taking with Gemini 📝

• Character.AI Founder Returns to Google to Co-Lead Gemini Project 🔄

• Anthropic Reveals System Prompts Behind Claude AI Models 🔍

• Amazon's AI-Enhanced "Remarkable Alexa" Set for October Launch 🗣️

• Amazon Expands AI Shopping Assistant Rufus to Indian Market 🛒

• Perplexity AI Explores Data Provider Integrations for Enhanced Capabilities 💼

• California's AI Regulation Bill Sparks Debate in Tech Industry ⚖️

• Survey Shows Hiring Professionals' Approval of AI-Assisted Resume Writing 📄

• Dexcom Introduces Stelo: First OTC Glucose Monitor for Prediabetics 🩺

• UK Government Initiates AI Development for Reducing Teacher Workloads 🏫

• McKinsey's Tech Trends Report Raises Questions About AI in Forecasting 📊

• Plaud.AI Unveils NotePin: Simplified AI Wearable for Conversation Transcription 🎙️

• AI Recipe Generator Struggles to Accurately Recreate Restaurant Dishes 🍳

• Elon Musk Announces Plans for Free Emergency Starlink Cellular Service 🛰️

• Apple Announces CFO Transition as Luca Maestri Steps Down 🍎

🍓 OpenAI reportedly preparing to launch advanced "Strawberry" AI model with enhanced problem-solving abilities. Link

• The new model, previously known as Q*, is said to excel at complex tasks like solving novel math problems and developing marketing strategies.

• Strawberry reportedly scores over 90% on the MATH benchmark, significantly outperforming GPT-4's 53% and GPT-4o's 76.6%.

• The model's development was linked to last year's brief leadership chaos at OpenAI, with some researchers expressing concerns about its potential risks and AGI capabilities.

🚀 Nvidia launches NIM Agent Blueprints to accelerate enterprise AI development. Link

• Nvidia introduced pretrained, customizable AI workflows called NIM Agent Blueprints to help enterprises quickly create AI applications using one or more AI agents.

• The blueprints include sample applications, reference code, and deployment tools for key use cases like customer service, drug discovery, and data extraction from PDFs.

• Major tech companies like Accenture, Dell, and HPE will help deliver and support these blueprints, with Nvidia planning to release new ones monthly to address additional enterprise AI needs.

💨 AI startup Cerebras launches ultra-fast inference service, challenging industry leaders. Link

• Cerebras introduced an AI inference service claiming to be 10-20 times faster than competitors like Nvidia, processing up to 1,832 tokens per second compared to around 300 for others.

• The service leverages Cerebras' CS-3 computer with the WSE-3 chip, offering "100x higher price-performance" at $0.60 per token versus $2.90 from average cloud providers.

• Beyond speed, Cerebras argues its service enables more complex, "agentic" AI applications and improves accuracy through faster multi-turn processing and chain-of-thought prompting.

📝 Google Meet adds AI-powered note-taking feature using Gemini. Link

• New "take notes for me" tool uses Gemini AI to generate summaries of key discussion points during video calls.

• Summaries are saved as Google Docs and can be automatically shared with attendees or added to calendar events.

• Currently limited to English language and desktop/laptop meetings for Google Workspace customers with specific AI add-ons.

🤝 Former Character.AI founder Noam Shazeer rejoins Google to co-lead Gemini AI project. Link

• Shazeer will serve as a technical lead on Gemini, alongside Jeff Dean and Oriol Vinyals.

• Google reportedly paid billions to bring Shazeer and some colleagues back, also striking a licensing deal with Character.AI.

• The move resembles similar talent acquisitions by Amazon and Microsoft from AI startups, raising potential regulatory scrutiny.

🗃 Anthropic reveals system prompts behind Claude AI models, offering rare transparency. Link

• The company shared commands that guide Claude 3.5 Sonnet, Claude 3 Opus, and Claude 3 Haiku models.

• Prompts include instructions on handling controversial topics, admitting limitations, and avoiding apologies or stereotyping.

• Sonnet has the most prompts, followed by Opus and Haiku, reflecting their different capabilities and use cases.

• Anthropic plans to regularly update the public on changes to these system prompts, marking an unusual move in the AI industry.

🗣️ Amazon's AI-powered "Remarkable Alexa" set to launch in October, with subscription required. Link

• The revamped Alexa with generative AI capabilities is now expected to roll out in October, slightly later than initially planned.

• Subscription for "Remarkable Alexa" could cost up to $10 per month, offering advanced features like content generation and more natural conversations.

• This move comes as tech giants race to integrate generative AI into virtual assistants, with Apple and Google also working on AI upgrades for Siri and Google Assistant.

🛒 Amazon launches AI shopping assistant Rufus in India, expanding its generative AI features. Link

• Rufus, first introduced in the US six months ago, is now available to select customers on the Amazon.in mobile app, with wider rollout planned in coming weeks.

• The AI assistant helps with product discovery, answering queries, comparing products, and providing recommendations.

• This move follows competitor Flipkart's launch of a similar assistant, Flippi, in October 2023, as e-commerce giants race to integrate AI features.

🔍 Perplexity AI reportedly developing integrations with data providers like FactSet. Link

• Code analysis suggests Perplexity is working on an "Integrations" feature, with FactSet as a potential first connector.

• FactSet is a financial data provider, indicating Perplexity may offer access to extensive financial datasets and analytics.

• The integration could significantly enhance Perplexity's capabilities, though it's unclear if it will be limited to enterprise accounts or widely available.

🤖 California's AI regulation bill divides tech industry, sparking debate on innovation and safety. Link

• SB 1047, which passed the state Senate 32-1, requires AI developers to comply with certain rules before developing models, with the Assembly facing an Aug. 31 deadline for approval.

• Tech companies are split, with Anthropic cautiously supporting the bill after changes, while OpenAI opposes it, claiming it could "stifle innovation."

• Critics argue federal regulation is more appropriate, while supporters emphasize the need for action given Congress's inaction, highlighting California's role as a first mover in tech policy.

📄 Majority of hiring professionals approve using AI for resume writing, survey finds. Link

• 64% of hiring professionals surveyed by Adobe approve of job seekers using AI to help write resumes, while 28% of job seekers admit to already using AI for this purpose.

• Key resume red flags identified include spelling/grammar errors, lack of relevant experience, frequent job changes, unexplained employment gaps, and overly casual language.

• Best practices recommended by hiring professionals include: limiting resumes to one page (55%), using PDF format (75%), and emphasizing unique skills (29.68%).

🩺 Dexcom launches Stelo, first over-the-counter glucose monitor for prediabetics. Link

• Stelo is a $99 wearable sensor that provides 24/7 glucose monitoring for people with prediabetes or Type 2 diabetes not using insulin.

• The device has a 15-day battery life, is waterproof, and doesn't require fingerstick tests or a prescription.

• An accompanying app offers personalized insights, spike detection, and tracking of how food, exercise and sleep affect glucose levels.

🏫 UK government launches initiative to develop AI tools for reducing teacher workloads. Link

• The Department for Education (DfE) is granting AI models special access to its resources to train on official assessments, curriculum guidance, and teaching materials.

• A £3m "content bank" will be created to help developers train AI models to generate accurate, high-quality content like workbooks and lesson plans.

• An additional £1m competition will award developers for creating AI tools to assist teachers with feedback and marking tasks by March 2025.

📊 McKinsey's 2024 tech trends report sparks debate on AI's role in trend forecasting. Link

• McKinsey's report highlights AI, digital future, compute/connectivity, engineering, and sustainability as key trends.

• When asked to assess the list, AI assistants Gemini and Claude suggested additional trends and areas for improvement.

• The article questions the value of traditional trend reports when AI can quickly generate similar lists, while noting that all trend forecasts are hypotheses rather than certainties.

📌 Plaud.AI unveils NotePin, a simplified AI-powered wearable for note-taking. Link

• NotePin is a $169 device focused solely on transcribing conversations using LLM-based technology.

• It offers 300 minutes of free transcription monthly, with a $79/year Pro Plan providing 1,200 minutes and additional features.

• The device follows Plaud's earlier Plaud Note, which sold 200,000 units, positioning the company in the growing AI wearables market alongside competitors like Humane.

🍳 AI-powered recipe generator falls short in recreating restaurant dishes. Link

• SideChef's RecipeGen AI app, designed to turn photos into recipes, struggled with accuracy, missing key ingredients like strawberry butter and ricotta in a brunch dish recreation.

• The app performed better with a homemade sweet potato gnocchi photo, identifying 90% of ingredients correctly, but added unnecessary complexity to the cooking process.

• While potentially useful for inspiration or deciphering restaurant meals, the app may have limited appeal for skilled home cooks who prefer creative freedom in the kitchen.

🛰️ Elon Musk announces Starlink will offer free emergency cellular service worldwide, pending approval. Link

• SpaceX's "direct to cell" Starlink technology aims to provide emergency connectivity in cellular dead zones for all mobile phones.

• The service would be free for emergency use, with Musk stating "Can't have a situation where someone dies because they forgot or were unable to pay for it."

• SpaceX is seeking FCC approval to launch the service with T-Mobile this fall, facing opposition from rival carriers AT&T and Verizon over potential interference concerns.

🍎 Apple announces CFO transition, with Luca Maestri stepping down after successful decade-long tenure. Link

• Longtime CFO Luca Maestri will step down on January 1, 2025, to be replaced by Kevan Parekh, current VP of financial planning and analysis.

• During Maestri's tenure, Apple's net income more than doubled and stock value increased ninefold.

• Maestri will remain with Apple in a reduced role overseeing Corporate Services teams, reporting to CEO Tim Cook.

Have any feedback? Send us an email