•
December 19, 2024
•
4 minutes
Hello, Niuralogists!
Welcome to this week’s edition, where we dive into the latest advancements in artificial intelligence. We’ll explore how these innovations are shaping various aspects of our lives—from the workplace and business to policies and personal experiences. This issue highlights some fascinating updates, including worries about AI getting out of control, new media creation tools, content creator protections, ChatGPT’s step back in time, and more.
For more in-depth coverage, keep reading…
Former Google CEO Eric Schmidt voiced concerns about the future of AI, cautioning that its self-improving capabilities could become “dangerous.” Speaking on ABC News, Schmidt emphasized the need to unplug systems if they reach a point where they define their own objectives. “When the system can self-improve, we need to seriously think about unplugging it,” he said.
Schmidt, co-author of Genesis, a book exploring AI’s power and ethical challenges, highlighted the difficulty of maintaining a balance between innovation and preserving human values.
Addressing China’s advancements, Schmidt noted that while the U.S. once led AI development, China has quickly closed the gap and may surpass American capabilities. He advocated for U.S. leadership in AI and stronger intervention to establish guardrails, suggesting that AI systems should police themselves rather than relying solely on human oversight.
Google has unveiled Veo 2 and Imagen 3, its latest advancements in AI video and image generation, now available through VideoFX, ImageFX, and the new Labs experiment, Whisk.
Veo 2 sets a new benchmark in video realism, understanding physics, human movement, and cinematography. It can produce 4K videos in diverse styles, capturing intricate details like camera angles, lens effects, and natural environments. Veo 2 includes SynthID watermarking to ensure AI-generated content remains identifiable, supporting responsible use.
Imagen 3 enhances image generation with sharper compositions, richer textures, and better prompt adherence. It excels in diverse styles, from photorealism to abstract art, and is now accessible in over 100 countries via ImageFX.
Google also introduced Whisk, a playful tool that combines images with AI descriptions, enabling users to remix and create custom designs. These innovations empower creators to push the boundaries of digital storytelling.
OpenAI has launched a new landline integration for ChatGPT, bringing conversational AI to traditional telephones. This feature allows users to access ChatGPT's voice assistant directly via landline calls, making AI assistance more accessible for those without smartphones or reliable internet connections.
The landline integration supports tasks like scheduling, answering FAQs, and providing information in real-time. It leverages OpenAI's advanced voice recognition and response systems to ensure clear, accurate, and natural interactions.
This development expands ChatGPT's reach, bridging the gap between cutting-edge AI and underserved communities reliant on traditional communication methods. OpenAI’s move signals a step toward democratizing AI access, ensuring its benefits extend to a broader audience. The feature is available starting today, with plans for global rollout in 2024.
Perplexity AI Inc., the AI startup challenging Google in search, has closed a $500 million funding round, tripling its valuation to $9 billion. The round, led by Institutional Venture Partners, highlights growing investor confidence in generative AI's potential to transform online search.
Founded in 2022, Perplexity distinguishes itself with real-time information and tools like internal file search and finance-focused features for stock and earnings data. Its user base exceeds 15 million, and it recently launched revenue-sharing partnerships with major publishers, addressing plagiarism concerns.
Despite rapid growth, Perplexity faces stiff competition from OpenAI, Microsoft, and Google, who are integrating conversational AI into their search tools. Backed by Jeff Bezos, Nvidia, and SoftBank, the company aims to redefine search while expanding its services and partnerships.
OpenAI introduced major updates aimed at developers, including the release of OpenAI o1, a cutting-edge reasoning model that enhances multi-step task handling with advanced features like function calling, Structured Outputs, and vision capabilities. The new model significantly improves efficiency, reducing latency by 60% compared to its predecessor.
The Realtime API saw upgrades like WebRTC integration, enabling real-time voice applications with reduced costs and better performance. Pricing for GPT-4o audio tokens dropped by 60%, and the new GPT-4o mini offers a highly cost-effective alternative.
Developers can now leverage Preference Fine-Tuning to tailor models based on user feedback, ideal for subjective tasks. Additionally, official Go and Java SDKs make OpenAI’s tools accessible across more programming environments. These advancements empower developers to create more efficient, personalized, and scalable AI-driven solutions.
📬 Receive our amazing posts straight to your inbox. Get the latest news, company insights, and Niural updates.
A new benchmark, FACTS Grounding, introduces an online leaderboard to assess the factual accuracy of language models' long-form responses. The benchmark evaluates models on their ability to generate responses grounded in user-provided context documents of up to 32k tokens.
Evaluation follows a two-phase process: responses are disqualified if they fail to address the user request and are then judged for factual accuracy based on adherence to the provided context. Scores aggregate results from multiple automated judge models to reduce bias.
The leaderboard supports public and private participation to encourage collaboration while maintaining integrity. This tool is a significant step in improving LLMs' capability to produce factually accurate and contextually relevant long-form outputs. Explore the leaderboard on Kaggle to participate and track progress.
YouTube has partnered with Creative Artists Agency (CAA) to develop tools that give creators more control over AI-generated content depicting their likeness. This initiative grants top talent, including award-winning actors and athletes, early access to technology for detecting and managing AI-generated content featuring their faces or voices on YouTube.
The collaboration allows CAA's clients to provide valuable feedback to refine these tools, ensuring they align with creators' needs. The technology includes streamlined removal requests through YouTube's privacy complaint process, empowering creators to protect their digital presence.
This partnership is part of YouTube’s broader effort to build a responsible AI ecosystem. By collaborating with industry leaders like CAA, YouTube aims to empower artists while balancing the ethical challenges and creative opportunities of AI.
📸 Google Imagen 3 is Google’s highest-quality text-to-image model, capable of generating images with even better detail, lighting, and fewer artifacts.
🎨 Google Whisk generates images by using other images as prompts for a subject, scene, and style to create personalized visuals.
☎️ NewOaks AI Phone Agent is an AI phone agent that can listen, understand, and speak in real-time to automate inbound and outbound calls.
🧠 Findr unlocks infinite digital memory with your AI second brain.
📫 MagicMail is an AI email generator that turns text prompts into fully styled and ready-to-send HTML emails.
