- NATURAL 20
- Posts
- Grok AI Models Top Leaderboards
Grok AI Models Top Leaderboards
PLUS: Google Hires Shazeer for Gemini, Salesforce's Autonomous AI Agents and more.
Power your competitive advantage with intelligent automation from ELEKS
ELEKS' intelligent automation service transforms your business operations through data-driven solutions. We automate complex tasks, streamlining processes to increase productivity and reduce operational costs. Our tailored solutions adapt to your changing needs and help you unlock new growth opportunities by freeing your team to focus on high-value tasks.
The result? Enhanced customer satisfaction, improved client retention, and a stronger market position.
Today:
Grok AI Models Top Leaderboards
ElevenLabs Expands Globally, Adds Languages
Perplexity AI to Introduce Ads
Google Hires Shazeer for Gemini
AI21 Launches Jamba 1.5 Update
YouTube's New AI Recovery Tool
Salesforce's Autonomous AI Agents
Abacus.ai Launches Dracarys Models
Grok 2 Large is Smart, Uncensored and has "DANGEROUS POTENTIAL"...
The latest advancements in artificial intelligence, particularly highlight the achievements of the Grok 2 AI model and its smaller version, the Grok Mini. Grok 2 has achieved second place in the Chatbot Arena leaderboards, closely competing with giants like Google and OpenAI, while Grok Mini secured the fifth position.
These models, developed by xAI, have significantly improved in speed and efficiency due to a complete rewrite of their inference stack. Despite these advances, concerns about the potential dangers of AI persist, echoed by tech leaders like Elon Musk. The article showcases the model's capabilities through various logical and coding challenges, demonstrating both strengths and areas for improvement.
ElevenLabs’ AI Reader app can now narrate text in 32 languages
ElevenLabs' AI Reader app, initially launched in the US, UK, and Canada, is now globally available and supports 32 languages including Spanish and Mandarin. This app allows users to listen to various text formats like PDFs and articles on iOS and Android devices.
It features AI-generated voices, even mimicking deceased celebrities with permissions from their heirs. The app is free but will include a premium version in the future, with promises of a generous free plan.
Perplexity AI plans to start running ads in fourth quarter as AI-assisted search gains popularity
Perplexity AI, an AI search startup, is set to introduce ads in its search app during the fourth quarter amidst increasing popularity and past plagiarism controversies. The company, now valued over $1 billion, will use a revenue-sharing model with publishers, compensating them when their content generates ad revenue.
The app, downloaded over two million times, processes 230 million queries monthly, mostly from the U.S. Perplexity's upcoming ad model will target sectors like technology and health, offering varied ad placement options. This move highlights AI's growing role in reshaping information access, challenging traditional search engines like Google.
Google Makes Former Character.AI CEO Shazeer a Co-Leader of Gemini AI
Noam Shazeer, the former CEO of Character.AI, has rejoined Google as the co-technical lead of the Gemini AI project. Alongside established Google AI researchers Jeff Dean and Oriol Vinyals, Shazeer will help develop Gemini, Google’s major artificial intelligence initiative designed to compete with OpenAI's GPT models. This strategic move highlights Google's ongoing commitment to leading innovations in AI technology.
AI21 debuts Jamba 1.5, boosting hybrid SSM transformer model to enable agentic AI
AI21 has upgraded its Jamba model to version 1.5, enhancing the hybrid Structured State Space (SSM) transformer model. This update introduces the Jamba 1.5 mini and large models, which aim to integrate the strengths of transformers and SSM to improve AI performance and accuracy. New features include JSON mode for structured data, citation mode for accountable outputs, and a large context window.
These advancements support the development of agentic AI systems, capable of sophisticated data handling and providing transparent sourcing. AI21 offers these models under an open license, alongside commercial support and partnerships with major cloud platforms.
YouTube has a new tool to help restore hacked channels: an AI chatbot
YouTube has introduced a new AI chatbot to assist creators in recovering hacked accounts. This tool, currently available only in English and limited to select creators, guides users through securing their Google logins and reversing unauthorized changes on their channels.
Although the AI chatbot is a step towards better support, it doesn't address the broader issue of limited direct communication with YouTube, which has been a longstanding complaint, especially among smaller creators. The intent is to eventually make this feature accessible to all YouTube users, aiming to streamline the account recovery process.
Meet Einstein SDR and Einstein Sales Coach: Two New Autonomous AI Sales Agents to Scale Your Sales Team
Salesforce introduced two new AI sales agents: Einstein SDR Agent and Einstein Sales Coach Agent, set to be available in October. The Einstein SDR Agent automates the initial engagement with prospects, handling inquiries and setting meetings, thus allowing sales teams to focus on deeper customer interactions. Unlike typical chatbots, it uses enhanced AI to provide personalized, data-driven responses. The Einstein Sales Coach Agent helps salespeople by simulating realistic buyer interactions for training purposes, giving feedback based on comprehensive data analysis.
These tools aim to boost productivity by enabling sales teams to concentrate on complex deals while routine tasks are automated. Accenture plans to use these agents to enhance their sales effectiveness and manage more deals efficiently.
Open source Dracarys models ignite generative AI fired coding
Abacus.ai has launched Dracarys, a new family of open-source large language models (LLMs) tailored for coding tasks, named after the command for dragon fire in "Game of Thrones." Unlike its general-purpose LLM, Smaug, Dracarys is designed to optimize coding through a fine-tuning process dubbed the "Dracarys recipe." This process enhances the coding capabilities of models such as Qwen-2 72B and LLama-3.1 70B, evidenced by improved scores on LiveBench benchmarks.
Dracarys models are available on Hugging Face and as part of Abacus.ai's enterprise offerings, promising enhanced coding performance for developers and enterprises while keeping data private and off public APIs.
🧠RESEARCH
The study introduces Sapiens, a versatile model for processing human images, excelling in pose estimation, body-part segmentation, depth estimation, and surface prediction. Sapiens achieves high accuracy by learning from over 300 million human images and can be easily customized for different tasks.
The paper surveys advanced methods for controlling text generation in large language models, emphasizing customization and safety. Techniques like fine-tuning and prompt engineering enhance model reliability and applicability, focusing on real-world uses and addressing challenges in fluency and practical implementation.
Show-o is a transformative transformer model that merges multimodal understanding and generation, efficiently handling diverse inputs and tasks like visual question-answering and text-to-image generation. It performs on par or better than specialized models, offering a flexible and robust solution for next-generation foundational models.
The paper introduces xGen-VideoSyn-1, a model that generates realistic videos from text descriptions. Leveraging latent diffusion models and a novel video variational autoencoder, it efficiently handles long video sequences. This model, tested over 13 million video-text pairs, generates high-quality videos up to 14 seconds long, showing excellent performance compared to existing text-to-video models.
DreamCinema introduces a groundbreaking AI framework for film production, allowing users to generate 3D characters and manipulate camera movements easily. This innovation makes film creation more accessible, seamlessly integrating 3D character generation and cinematographic elements to produce high-quality cinematic content. The framework’s effectiveness in enhancing film creativity and production is confirmed through extensive testing.
🛠️TOP TOOLS
Ideogram - Advanced text rendering, photorealistic image generation, and Magic Prompt for detailed and beautiful images
D-iD Video Translate - Translate your videos into multiple languages using the power of AI
Google AI Studio - Explore prompt ideas for working with the latest Gemini models.
CreativAI - Create amazing blog entries, attention-grabbing advertisements, interesting emails, and super cool social media posts in just a few seconds
EVF-SAM-2 - Early Vision-Language Fusion for Text-Prompted Segment Anything Model
📲SOCIAL MEDIA
We've added support for LaTeX rendering as a feature preview.
Claude can now display mathematical equations and expressions in a consistent format.
— Anthropic (@AnthropicAI)
5:08 PM • Aug 22, 2024
🗞️MORE NEWS
OpenAI joins opposition to California AI safety bill
OpenAI criticizes California bill SB 1047 for potentially hindering AI growth and innovation in the state. Silicon Valley divided over the bill's impact on AI start-ups and California's position in the AI sector. Despite amendments, critics argue the bill still imposes unrealistic requirements on AI developers and startups. FINANCIAL TIMES
Lightweight Champ: NVIDIA Releases Small Language Model With State-of-the-Art Accuracy
NVIDIA introduced a compact AI language model, Mistral-NeMo-Minitron 8B, which matches the performance of larger models but is efficient enough to run on local devices. This smaller model, designed using advanced techniques like pruning and distillation, enhances accessibility for developers with limited resources and offers high accuracy and lower operational costs. NVIDIA
AI-powered coding pulls in almost $1bn of funding to claim ‘killer app’ status
AI-driven coding assistants attract nearly $1bn in funding, with companies like Replit and GitHub Copilot leading the way. Investors pour money into AI coding tools as software engineering is transformed by AI technology. GitHub Copilot, an AI coding assistant, writes better code than the average developer, contributing significantly to GitHub's revenue growth. FINANCIAL TIMES
Meshy-4 brings sci-fi level AI to 3D modeling and design
Meshy, an AI startup, has launched Meshy-4, a 3D modeling tool that enhances the quality of AI-generated 3D models with cleaner geometry and a more efficient workflow. This latest version aims to revolutionize design across industries like gaming and architecture by allowing more control over texturing and modeling, speeding up design processes, and reducing costs. VENTUREBEAT
China Uses Amazon Cloud to Get High-End US Chips, AI
Chinese entities have circumvented U.S. chip export bans by accessing advanced AI chips and technologies through Amazon Cloud services, according to Reuters. This method exploits a loophole as it doesn't directly violate U.S. export regulations, which only restrict physical transfers. The U.S. government is considering tighter rules to close this loophole amidst concerns about the potential military applications of these technologies. NEWSMAX
Google DeepMind staff call for end to military contracts
Nearly 200 employees at Google DeepMind have signed a letter urging the company to end its military contracts, particularly highlighting concerns over AI technologies being used in warfare and surveillance, including a contentious project with the Israeli military. The letter calls for the establishment of a governance body to prevent military use of DeepMind's AI, reflecting ongoing tensions within Google over the ethical implications of such partnerships. Despite these concerns, there has been no substantial response from Google leadership. THE VERGE
What'd you think of today's edition? |
Learn AI with us. Let’s Build the Future Together. |
Hello fellow AI-obsessed traveler, Over the past 2 years, as we’ve grown to over 250,000 subscribers between the YouTube Channel and this newsletter, we've received an overwhelming number of requests for one specific thing. While the newsletter helps keep you up to speed with AI news, many of you have asked for the next step: to learn how to actually apply AI in your work. Today we’re finally announcing the solution with NATURAL 20, the community for like-minded AI learners. As a loyal newsletter reader you are getting access at the lowest price it will ever be: JOIN NATURAL 20 AI UNIVERSITY TODAY What you get: * Tutorials by experts across various AI fields. * Daily tutorials by Wes Roth about the latest use cases. * Building Autonomous AI Agents to Automate Your Life and Business (NEW!) * A network of the top 1% of early AI adopters. * Access to community-only resources and software. * And many more features rolling out soon. |
Reply