When AI Took the Wheel: The New Engine of Everyday Life

Hey there!

Much of the public discourse around AI focuses on risks like deepfakes, autonomous weapons, or the existential threat of AGI. In his TED Talk at TEDAI Vienna in October 2024 published just recently, Thomas Wolf, co-founder of Hugging Face, shifts the conversation toward AI's potential as a useful, well-controlled tool.

Wolf asks: What if AI simply becomes a well-integrated tool? Like the internet, AI could soon become indispensable in daily life, influencing both personal and professional routines. He predicts AI will mediate most of our tasks, automating work, facilitating communication, and handling creative processes appealing to fundamental human needs like reducing mental burdens and providing entertainment.

For AI to truly work for us, it needs to be reliable and adaptable. That's why Wolf highlights the importance of open-source AI, models that can run on personal devices and be improved by a global community, offering stability over time.

For everyday users, the key takeaway is simple: as AI becomes more embedded in our lives, choose systems that offer transparency, community-driven support, and resilience. Just as we evaluate the privacy and reliability of internet services, we must do the same with AI tools to ensure they are both transformative and safe.

Now, let's get to the news!

• Generative AI Tools Updates
  • OpenAI Deep Research System Card
  • Anthropic Claude 3.7 Sonnet and Claude Code
  • Google Research AI Co-Scientist
  • Luma Labs Video to Audio
  • Google AI Coding Assistant
  • Amazon Alexa+
  • ElevenLabs Scribe
• Other News
• Curated Gems
• From Learn Prompting Team

Generative AI Tools Updates

OpenAI Deep Research System Card

OpenAI has released a system card for its Deep Research, a new agentic AI capability designed for conducting multi-step research across the internet. It can autonomously search, interpret, and analyze vast amounts of text, images, and PDFs to complete complex research tasks. The model is built on an early version of OpenAI o3, optimized for web browsing and data analysis.

Key Features

• Autonomous web research: Searches, reads, and synthesizes information from multiple sources.
• Interprets diverse formats: Handles PDFs, images, and structured data.
• Python-based analysis: Executes calculations, creates visualizations, and processes datasets.
• Dynamic reasoning: Adjusts research direction based on new findings.

Access

Pro users can make up to 100 queries per month. Recently, OpenAI has extended the availability of 10 Deep Research queries per month to all Plus, Team, Enterprise, and Edu users.

Anthropic Claude 3.7 Sonnet and Claude Code

Anthropic has unveiled two new AI models, Claude 3.7 Sonnet and Claude Code. These models blend rapid response generation with deep, multi-step reasoning, enabling them to tackle complex problem-solving and streamline software development processes.

Key Features

• Hybrid reasoning & extended thinking: Claude 3.7 Sonnet fuses quick output with a deliberate, step-by-step analytical process. It features an extended thinking mode (available on paid tiers) that allocates extra "thinking tokens" for deeper multi-stage reasoning, with users even able to view the model's internal thought process.

• Expanded context window: Supports up to 200,000 input tokens and can produce outputs up to 128,000 tokens (beta), allowing it to handle lengthy documents and detailed conversations effectively.

• Agentic coding capabilities: Claude Code is optimized for coding, autonomously navigating and modifying codebases. It can execute tasks such as file editing, test running, and version control integration based solely on natural language prompts, streamlining the software development lifecycle.

• Performance improvements: Benchmarks indicate enhancements including over 10% better management of complex workflows, a 30% boost in summarization accuracy, and a 24% improvement in information retrieval precision.

Access & Deployment

• API & cloud integration: Both models are accessible via Anthropic's API and are integrated with major cloud platforms like Amazon Bedrock and Google Cloud's Vertex AI.

• Interactive web interface: Users can also experiment with these models in real time through an interactive web interface, which supports their extended reasoning and coding functionalities.

• Pricing: The models are priced at $3 per million input tokens and $15 per million output tokens, ensuring a cost-effective solution for both research and professional development.

Google Research AI Co-Scientist

Google Research has introduced its new AI Co-Scientist, a system built on Gemini 2.0, that acts as a collaborative research partner. It generates novel hypotheses, evaluates them through simulated scientific debates, refines ideas via tournaments and evolution, and performs fact-checking using web search and other tools.

Key Feature: Multi-Agent Architecture

The system is composed of several specialized AI agents that mirror key steps in the scientific method:

• A Generation agent proposes new hypotheses by reviewing relevant literature and data.
• A Reflection agent serves as a peer reviewer, assessing the quality and novelty of the proposals.
• A Ranking agent pits ideas against each other in simulated debates, while an Evolution agent refines the top suggestions.
• A Proximity agent helps with the search for relevant literature.
• A Meta-review agent performs a final check of the proposed solutions.

Access

For now, access to the tool is limited. Research organizations can gain early access by joining Google's Trusted Tester Program, where they can provide feedback and help further refine the system.

Luma Labs Video to Audio

Luma Labs has launched Video to Audio feature for its Dream Machine platform. This new capability allows users to easily generate synced audio tracks for AI-created videos, addressing a long-standing gap in the AI video production landscape.

Key Features

• Contextual audio generation: The system analyzes video content to produce sound effects or ambient audio that matches the visuals.
• Dual mode options: Users can either let the AI automatically select and generate the best-fitting audio or provide custom text prompts to guide the audio creation process.
• Seamless integration: The feature appears as a new "Audio" button alongside existing options like "Extend" and "Enhance" in the Dream Machine interface.
• Real-Time Processing: Audio generation happens in real time after the video is created, making the editing process more efficient.

Access

The Video to Audio feature is available in beta for all Dream Machine users at no extra cost. Users can access it via the Dream Machine platform on both iOS and the web. Existing Dream Machine accounts and projects are automatically updated with this new feature.

Google AI Coding Assistant

Google has launched a free AI coding assistant, Gemini Code Assist for individuals, to help developers write, debug, and review code.

Key Features

• Generous usage limits: Offers up to 180,000 code completions per month, far exceeding the limits of competitors like GitHub Copilot.
• Powered by Gemini 2.0: Optimized for coding, it supports all public domain programming languages and includes a massive context window of 128,000 tokens, ensuring better understanding of codebases.
• IDE integration: Available as an extension for popular development environments, including Visual Studio Code, JetBrains IDEs, and GitHub (with integrated AI-powered code reviews).
• Natural language interaction: Developers can use plain language prompts to generate code, debug, and request code reviews, making the tool highly user-friendly.

Access

Accessible immediately for individual developers via a personal Gmail account, no credit card is required. You can install Gemini Code Assist in your preferred IDE (Visual Studio Code, JetBrains, or as a GitHub app) and start coding with AI support right away.

Amazon Alexa+

Amazon has unveiled Alexa+, an AI-enhanced upgrade to its popular Alexa voice assistant. This next-generation service uses advanced generative AI to provide more natural, personalized, and proactive assistance.

Key Features

• Generative AI capabilities: Powered by advanced AI models from Amazon and Anthropic, Alexa+ can perform tasks such as ordering rides, booking restaurants, planning events, creating recipes, and even generating bedtime stories. It can analyze emotional cues, making conversations more natural and personalized.

• Enhanced integration & functionality: Works across a range of Amazon Echo devices, including Echo Show models (8, 10, 15, and 21), and will also be available via a mobile app and web browser. Integrates with partner services like Uber, OpenTable, Ring, Grubhub, and others for comprehensive smart home management.

• Subscription model & pricing: Alexa+ is priced at $19.99 per month for non-Prime members. Amazon Prime members can access it at no additional cost.

• Future rollout: The service will start rolling out in the U.S. in the coming weeks through an early access program, with gradual expansion to other regions and devices.

Access

Alexa+ is free as an added benefit with an Amazon Prime subscription. It is available for $19.99 per month.

Initially available on eligible newer Echo devices, particularly the Echo Show series, with plans to extend to other compatible devices and platforms.

ElevenLabs Scribe

ElevenLabs has launched Scribe, a state-of-the-art speech-to-text model designed to deliver the most accurate transcriptions available. Scribe is engineered to handle multiple speakers and provide precise timestamp support, setting a new benchmark in audio transcription quality.

Key Features & Updates

• High accuracy: Scribe leverages advanced AI and deep learning techniques to achieve unparalleled transcription accuracy.
• Multi-speaker support: It can differentiate and transcribe dialogues involving multiple speakers seamlessly, making it ideal for meetings, interviews, and podcasts.
• Timestamp integration: Every segment of the transcription is time-coded, enabling easy navigation through audio content.
• Robust performance: Designed to perform reliably even in noisy environments, ensuring high-quality transcriptions across diverse settings.

Access

Scribe is available via ElevenLabs' platform and API. Developers and content creators can integrate this tool into their workflows through the ElevenLabs website.

Other News

• Pika Labs launched Pikaswaps, an innovative video editing tool that lets users replace any item or character in a video using images or text prompts, transforming video content on the fly.

• Veo 2 unveiled its pricing, setting the cost at 50 cents per second (approximately $30 per minute and $1800 per hour).

• Google AI Studio introduced branching chats, allowing users to explore multiple ideas in parallel.

• Perplexity announced its new web browser, Comet, coming soon.

• Microsoft announced that its AI-powered features, Copilot Voice and Think Deeper, are now available for free with unlimited access.

• Perplexity updated its app with a new voice mode.

Curated Gems

Top 8 Online AI Red Teaming Courses [Free & Paid]

Discover the top 8 AI red teaming courses to master vulnerabilities, adversarial testing, and securing AI systems, perfect for all skill levels.

From Learn Prompting Team

Join our 6‑week Masterclass on AI Security and learn from top experts in Generative AI, Cybersecurity, and AI Red Teaming.

In addition to Sander Schulhoff, you'll gain exclusive insights from:

• Pliny the Prompter: Renowned AI Jailbreaker, known for bypassing major AI model defenses.
• Johann Rehberger: Ex‑Microsoft Azure Red Team leader, expert in advanced attack vectors like ASCII Smuggling and AI‑powered C2 attacks.
• Joseph Thacker: Principal AI Engineer who's uncovered over 1,000 vulnerabilities across leading platforms.
• Akshat Parikh: Former AI security researcher, celebrated for his elite performance in global bug bounty competitions.
• Richard Lundeen: Microsoft's AI Red Team lead, at the forefront of innovative security strategies for AI systems.
• Sandy Dunn: Seasoned CISO with 20+ years in cybersecurity, specializing in risk governance for LLM applications.

Secure your spot today and elevate your AI security expertise with hands‑on training and real‑world projects!