Welcome, AI RESEARCHERS.

Nvidia is already dominating the chipmaking industry, but its models are also effortlessly topping LLM benchmarks and outperforming the giants.

Did the company's latest Nemotron release just crown a new champion for the open-source AI movement? Let's learn...

In today’s AI Researchs:

• Nvidia's New AI Crushes GPT-4o

• and Claude 3.5

• Major AI developments

• Chipmaker Shares Surge as AI Demand Fuels Record Profits

• 4 ChatGPT Voice Prompts to Improve Your Skills

• AI Researches & AI pointers

• Trending Ai tools of the week

• More Ai news and developments

Read time: 4 minutes

Nvidia

🚨 Nvidia's New AI Crushes GPT-4o and Claude 3.5

AI Researchs: Nvidia just released its latest AI model, Llama-3.1-Nemotron-70B-Instruct, without a flashy announcement but with impressive results. The model is already outperforming top competitors like GPT-4o and Claude 3.5 Sonnet on crucial benchmarks.

The Details:

• Based on Meta’s Llama 3.1, Nvidia fine-tuned the model using advanced techniques like Reinforcement Learning from Human Feedback (RLHF), resulting in superior language generation capabilities.

• Llama-3.1-Nemotron-70B-Instruct topped benchmarks, scoring 85.0 on Arena Hard and 57.6 on AlpacaEval 2 LC, overtaking larger models.

• It’s available as an open-source model on Hugging Face, and Nvidia offers free access at Nvidia and hugging face.

• Despite its smaller size, at 70 billion parameters, it delivers high performance, making it efficient and scalable for businesses.

• You can try it for free

Why this important: Nvidia's leap from being a GPU giant to creating top-performing AI models could reshape the landscape, showing that smaller, fine-tuned models can rival the industry's largest systems, pushing companies to rethink their AI strategies.

Writer RAG tool: build production-ready RAG apps in minutes

• Writer RAG Tool: build production-ready RAG apps in minutes with simple API calls.

• Knowledge Graph integration for intelligent data retrieval and AI-powered interactions.

• Streamlined full-stack platform eliminates complex setups for scalable, accurate AI workflows.

Learn more about our production ready RAG tooling here.

Major Quick Developments

Mistral has unveiled its new Ministral 3B and 8B models, with the smaller 3B outperforming its predecessor and the 8B leading its class, offering powerful AI capabilities that can run locally on devices.

Perplexity has unveiled Internal Knowledge Search, a highly-requested feature that allows simultaneous searches across internal knowledge bases and the public web. Additionally, they’ve introduced Spaces, a customizable collaboration hub where users can upload files, choose AI models, and work collaboratively with tailored search instructions.

AI CHIPMAKER RECORDS

🏆📉 TSMC Shares Surge as AI Demand Fuels Record Profits

AI Researchs: TSMC, the world’s largest chipmaker, exceeded market forecasts with impressive third-quarter results fueled by strong demand for AI-related chips. The company posted a 54% rise in net income and guided for a strong Q4, signaling continued momentum as AI applications drive growth.

The Details:

• TSMC’s Q3 revenue hit $23.5 billion, up 39% year-over-year, beating Bloomberg’s estimate of $22.4 billion to $23.2 billion.

• Net income surged to $10.1 billion, with advanced 3nm and 5nm chips making up 69% of total wafer revenue.

• AI and smartphone demand fueled growth, with High-Performance Computing (HPC) accounting for 51% of total revenue.

• Q4 revenue is expected to reach between $26.1 billion and $26.9 billion, exceeding the $24.86 billion consensus estimate.

Why this matters: TSMC’s strong performance underscores its pivotal role in AI innovation, as its cutting-edge chip technology continues to meet surging demand from major tech players like Nvidia and Apple. This positions the company at the forefront of the AI hardware race, with no signs of slowing down.

AI TRAINING

✅ 4 ChatGPT Voice Prompts to Improve Your Skills

AI Researchs: ChatGPT Voice Mode has unlocked so many incredible use cases. You now have an AI assistant that can role-play any situation to help you improve skills like public speaking, sales, or even learning a different language.

4 Powerful Prompts:

1. Public Speaking Practice

   Prompt: "You are a live audience, and I'm practicing my speech on [your topic]. I will deliver my speech, and after each section, you will ask me a follow-up question, as if you were an audience member seeking clarification or additional information. Provide feedback on my delivery, including my pacing, tone, and whether I'm keeping the audience engaged. Help me adjust my speech based on your feedback."

2. Language Tutoring

   Prompt: "You are a native speaker of [target language], and we are roleplaying a conversation where I am asking for directions to a popular landmark in [city]. I will ask you questions about how to get there, and you will respond as a local. Correct any mistakes I make and provide suggestions for how I could sound more natural. If I struggle, offer slower explanations or alternative phrases."

3. Cold Calling

   Prompt: "You are a busy professional, and I am cold-calling to offer [product/service]. We will simulate a live sales call where I need to quickly capture your attention and interest before you hang up. Provide feedback on my initial pitch and ask questions or raise objections to see how I handle them in real-time. Give me pointers on how I can improve my pitch to close more cold sales."

4. Interview Preparation

   Prompt: "You are the hiring manager for a [job title] position at [company]. We are going to simulate a job interview. Start by asking me to introduce myself and explain why I am a good fit for the role. Ask at least three situational questions, and provide feedback on my responses, including tone, pacing, and content. If my answers are unclear or off-topic, guide me to give better responses and improve my confidence during the interview."

AI Research Roundup

Researchers have utilized AI to uncover 70,500 new RNA viruses from environmental genomic data, revealing a vast array of previously unknown species.

Meta has introduced CoTracker3, an AI model that uses pseudo-labeling to track points in real videos, achieving superior performance with a simpler architecture.

AI Pointers

Sam Altman says the most important piece of knowledge discovered in his lifetime was that scaling AI models leads to unbelievable, predictable improvements in intelligence and he wondered if he was crazy or in a cult when he tried to explain it to others and they didn't understand

Ex-Google CEO Eric Schmidt says the future of war is electronic warfare and automated drones powered by artificial intelligence fighting other drones, where it is too lethal for humans to enter the front line, "if you look up and you see a drone, within 2-3 minutes you're dead"

• 🌍 Subtitly - Real-time subtitles and translations for any application.

• 🚀 Focus Buddy - Your AI-powered productivity partner to help you achieve more.

• 🤝 Claap 3.0 - An AI sales coach that manages meeting notes, enhances CRM data, reviews deals, and more.

• 🔎 Strella - Comprehensive customer research platform with AI-led interviews and instant analysis.

• 📂 Dropbox Dash for Business - Easily locate, secure, organize, and share your company's apps, files, and browser tabs.

Apple has released Depth Pro, an AI model that generates detailed 3D depth maps from single 2D images without relying on traditional camera data.

TSMC reported better-than-expected quarterly earnings and increased its revenue outlook for 2024, alleviating concerns about global chip demand and the sustainability of the AI market.

The New York Times has issued a cease and desist notice to AI search startup Perplexity, alleging unauthorized use of its content for AI-generated summaries.

In a new video, the Unitree G1 robot demonstrates its physical agility by performing a standing long jump of 1.4 meters. This achievement is especially notable considering the robot’s height of only 1.3 meters

X (Twitter) has updated its Terms of Service, introducing language that permits the use of user-generated content for AI model training, including generative AI systems.

Anthropic has unveiled an updated Responsible Scaling Policy, which introduces key safety thresholds for managing risks in advanced AI systems, focusing on bioweapons and autonomous AI research, with regular evaluations and increased transparency.

Meta has unveiled Thought Preference Optimization (TPO), a technique that prompts AI models to generate internal thoughts before responding, boosting performance in creative and non-reasoning tasks like marketing.

ChatGPT's web traffic reached a record high of 3.1 billion visits in September 2024, according to Similarweb. This represents a 112% increase year-over-year, making it the 11th most visited website globally.

Suno launched Suno Scenes, a new feature that allows users to generate songs using images or videos as prompts, not just text.

Google Public Sector announced $15 million in grants to upskill U.S. government workers in responsible AI. The initiative aims to train over 100,000 public sector employees across federal, state, and local levels.

OpenAI published research examining how ChatGPT responds to usernames with various genders, racial, and cultural backgrounds. The study found minimal bias but noted some stereotypical responses in open-ended tasks like creative writing.

Share Your Opinion