• NATURAL 20
  • Posts
  • Gemini Ultra 1.0 | Is THIS the GPT-4 Killer? We ran a BATTERY of tests, here is the SHOCKING result.

Gemini Ultra 1.0 | Is THIS the GPT-4 Killer? We ran a BATTERY of tests, here is the SHOCKING result.

PLUS: Amazon Joins AI Consortium, Microsoft and Sarvam AI Team Up and more…

Today:

Gemini Ultra 1.0 | Is THIS the GPT-4 Killer? We ran a BATTERY of tests, here is the SHOCKING result.

Google just dropped Gemini, waving goodbye to Bard and bringing in the big guns with Gemini Ultra 1.0. It's a showdown between Gemini and GPT-4, comparing their chops on jokes, proverb interpretation, and even art creation. Gemini's part of Google's One AI Premium plan, costing $20 a month with a free 2-month trial and some hefty storage perks.

Gemini shows promise, especially with its ability to generate snappy answers and handle tasks with a blend of precision and speed. While it's early days, Gemini's stepping up as a solid contender to GPT-4, suggesting we're in for an exciting AI face-off.

Google Joins Effort to Help Spot Content Made With A.I.

Google is teaming up with other tech giants to establish digital credentials for content, akin to nutrition labels. These labels would reveal how A.I. has influenced media creation or alteration. The move follows a similar initiative by Meta. With trust in online content increasingly important, Google aims to integrate these credentials into its products. 

Lawmakers and tech watchdogs seek to address disinformation fueled by realistic A.I.-generated content. Though efforts like watermarks aim to verify content origin, challenges persist. Adobe also emphasizes the urgency of ensuring content authenticity in the digital era.

Amazon joins US Artificial Intelligence Safety Institute to advance responsible AI

Amazon has joined the U.S. Artificial Intelligence Safety Institute to advance responsible AI. With AI's transformative potential, Amazon is committed to its responsible use, collaborating with governments and customers. They're part of a consortium to develop trustworthy AI and contribute $5 million for safety evaluation tools. 

This aligns with VP Harris's initiative and President Biden's Executive Order. Amazon aims to address challenges of generative AI, ensuring safety and security. They endorse flexible best practices for managing AI risks and invest in future AI standards. Their involvement reflects dedication to responsible AI and customer welfare.

OpenAI Shifts AI Battleground to Software That Operates Devices, Automates Tasks

OpenAI is shifting gears in the AI race, focusing on software that controls devices and handles tasks automatically. After the success of ChatGPT, which sparked imitations, the company is gearing up to launch a potentially groundbreaking product. They're working on agent software that can take over a user's device to perform complex tasks, like transferring data or filling out expense reports. 

This software would mimic human actions across various applications, triggered by user commands. With rivals like Google catching up, OpenAI is hustling to stay ahead in the evolving landscape of AI technology.

Microsoft partners with India's Sarvam AI for voice-based genAI tools

Microsoft is teaming up with Indian startup Sarvam AI to create voice-based AI tools. This collaboration, announced during CEO Satya Nadella's visit to India, aims to expand Microsoft's reach in the country. Sarvam AI, known for its work on AI models at Indian Institute of Technology, Madras, will leverage Microsoft's cloud services to develop solutions for Indic languages. 

With a $41 million funding boost, the startup plans to utilize Azure OpenAI Service for its projects. This move aligns with Nadella's initiative to provide AI skilling opportunities to 2 million Indians by 2025, reflecting Microsoft's commitment to the Indian market's growth.

🧠RESEARCH

1X Tech's latest feat: androids trained to perform tasks like patrolling and home chores by learning from visual data, no human remote control needed. They're expanding their team to push the boundaries of robot autonomy. It's like a techy mix of DIY meets AI, where robots learn from watching and doing, aiming to ace a variety of gigs from warehouse work to social interactions, all in real-time. 

Nvidia's Canary-1B, a powerful tool with 1 billion parameters, is revolutionizing automatic speech-to-text recognition (ASR) by supporting four languages: English, German, French, and Spanish. It doesn't just stop there; it also offers translation capabilities between these languages, both ways, and can handle text with or without punctuation and capitalization. This innovation opens up new possibilities for seamless communication and content creation across different languages. 

 Large Multi-View Gaussian Model (LGM) is a new method for creating high-resolution 3D content from text prompts or single-view images. By utilizing multi-view Gaussian features for efficient representation and an asymmetric U-Net for processing, LGM generates high-fidelity 3D models within seconds. This approach significantly improves the resolution and speed of 3D content creation, achieving high-resolution outputs at 512 within just 5 seconds​.

This paper presents a breakthrough in chess AI, demonstrating a 270M parameter transformer model trained on 10 million chess games and achieving grandmaster-level play without search algorithms. It surpasses traditional engines and even AlphaZero in certain aspects by learning from annotated action-values of the Stockfish 16 engine. The model showcases the potential of scale in machine learning, achieving high performance in chess solely through supervised learning, without the need for domain-specific adjustments or explicit search strategies​

"BiLLM: Pushing the Limit of Post-Training Quantization for LLMs" introduces BiLLM, a novel 1-bit post-training quantization method tailored for pretrained large language models (LLMs). It efficiently compresses model weights to reduce computational and memory demands while maintaining high performance. By selecting salient weights and employing a binary residual approximation strategy, along with an optimal splitting search for non-salient weights, BiLLM achieves high-accuracy inference across various LLM families, significantly outperforming state-of-the-art quantization methods.

🛠️TOP TOOLS

Letterly - Turn your speech into well-written text

Together - focused on fostering collaboration and teamwork among individuals and groups, using AI-powered tools to enhance productivity and communication.

Colossyan -  AI video platform for workplace learning. It transforms text into AI videos, auto-translates to multiple languages, and offers various AI avatars.

Bitskout - AI-powered plugins to extract data from documents and emails for automation. Users can integrate these plugins with tools like Asana, Zapier, or Power Automate. 

Lazy - designed to streamline note-taking and information organization.

📲SOCIAL MEDIA

🗞️MORE NEWS

News Corp in ‘advanced negotiations’ with AI companies over access to content, CEO says

News Corp CEO Robert Thomson announces advanced negotiations with AI firms for content access, prioritizing negotiation over litigation. Media concerns over compensation for AI training content persist globally. Thomson favors agreements over lawsuits, citing the prevalence of content counterfeiters in the AI realm. News Corp aims for global precedents benefiting journalism. THE GUARDIAN

AI-Generated Voices in Robocalls Are Now Illegal

The FCC bans AI-generated voices in US robocalls, targeting scammers. Expanded TCPA covers such calls, effective immediately. FCC chair warns against using AI to deceive vulnerable individuals or misinform voters. Move follows identification of Life Corporation behind recent Biden imitations. Existing laws like TCPA empower FCC to combat AI threats. WIRED

Pivotal Founder Rob Mee’s New Startup Is Using AI To Drag Ancient Mainframes Into The Cloud

Rob Mee, previously of Pivotal, launches Mechanical Orchard, targeting legacy mainframe systems' migration using AI. With $24 million funding, it achieves $10 million revenue in 16 months. Leveraging open-source tech and AI, it modernizes systems faster and cheaper, attracting interest despite challenges from big players like Accenture. FORBES

New AI Circuitry That Mimics Human Brains Makes Models Smarter

A new type of transistor, called a moiré synaptic transistor, mimics human brain function, integrating memory with processing to enhance AI efficiency. Developed by researchers like Mark Hersam at Northwestern University, this breakthrough paves the way for energy-efficient AI systems capable of more complex decision-making, reducing costs and environmental impact. SCIENTIFIC AMERICAN

AI is mixing up how liquor brands make and sell their spirits

The liquor industry leverages AI for personalized recommendations and marketing strategies. Platforms like What’s Your Whiskey analyze consumer preferences, while companies like Diageo and Pernod Ricard use AI for targeted advertising and optimizing sales tactics. AI's role extends to packaging design and production efficiency, shaping the future of liquor brands. FORTUNE

London Underground Is Testing Real-Time AI Surveillance Tools to Spot Crime

The London Underground has conducted a trial using real-time AI surveillance tools to detect potential crimes and safety hazards. This trial, which ran from October 2022 to September 2023, involved combining AI algorithms with live CCTV footage at Willesden Green Tube station. The AI system aimed to identify aggressive behavior, weapons, people falling onto tracks, and fare evasion. WIRED

What'd you think of today's edition?

Login or Subscribe to participate in polls.

What are MOST interested in learning about AI?

What stories or resources will be most interesting for you to hear about?

Login or Subscribe to participate in polls.

Reply

or to participate.