NVIDIA AI Milestones

PLUS: Stability AI Faces Key Departures, GitHub Launches Auto-Fix Tool and more.


  • NVIDIA AI Milestones

  • Stability AI Faces Key Departures

  • GitHub Launches Auto-Fix Tool

  • OpenAI Prepares GPT-5 Launch

  • Google Fined $270M in France

  • ServiceNow Integrates Gen AI Features

  • AWS, Accenture, and Anthropic Partnership

NVIDIA Reveals STUNNING Breakthroughs: Blackwell, Intelligence Factory, Foundation Agents [SUPERCUT]

The AI industry benefits greatly from scaling, particularly evident in large language models like the Transformer. With models doubling every six months, computational demands surge, now reaching trillions of parameters and tokens. To meet these needs, larger models must be trained on multimodal data, like text and images, demanding substantial computational power. 

Nvidia introduces Blackwell, a massive chip platform designed for generative AI, boasting superior inference capabilities compared to Hopper. The vision extends to AI factories, where intelligence generation is prioritized, culminating in projects like Groot, a humanoid robot powered by advanced AI technologies.

Key Stable Diffusion Researchers Leave Stability AI As Company Flounders

Key researchers behind the Stable Diffusion text-to-image model have left the struggling AI startup Stability AI. The departures of Robin Rombach, Andreas Blattmann, and Dominik Lorenz mark another setback for the company, which has faced financial challenges and executive turnover. 

Stability's success was tied to Stable Diffusion, but recent cash flow issues and copyright lawsuits have led to a decline in its fortunes. Amidst a mass exodus of executives and financial woes, including struggles to pay wages and payroll taxes, Stability is grappling with its future. Despite efforts to raise funds and divest assets, the company's prospects remain uncertain.

GitHub’s latest AI tool can automatically fix code vulnerabilities

GitHub introduces a new tool to automatically fix code vulnerabilities, complementing its existing security features. This feature, part of GitHub Advanced Security, utilizes real-time analysis and AI to identify and repair security flaws while coding. By integrating Copilot and CodeQL, it promises to resolve over two-thirds of vulnerabilities without manual intervention. 

GitHub aims to streamline development processes by reducing time spent on remediation, allowing teams to focus on strategic security measures. Leveraging AI models like GPT-4, the tool offers accurate fixes, although some may require manual verification. This initiative aligns with GitHub's commitment to enhancing code quality and security.

OpenAI is expected to release a 'materially better' GPT-5 for its chatbot mid-year, sources say

OpenAI is gearing up to unveil GPT-5, a significant upgrade to its ChatGPT tool, expected around mid-year. The new model, described as "materially better," is undergoing training and internal safety testing. OpenAI aims to address performance issues encountered with GPT-4, which faced criticism for degraded responses. 

Enterprise customers have received demos showcasing GPT-5's capabilities, hinting at enhanced features like autonomous AI agents. The company's revenue largely depends on sales to enterprises, emphasizing the significance of this release for its business growth. OpenAI's continued development seeks to maintain its position as a leader in generative AI technology.

AWS, Accenture and Anthropic partner to accelerate enterprise AI adoption

AWS, Accenture, and Anthropic are teaming up to speed up AI adoption in regulated sectors like healthcare and banking. Through a unique partnership, enterprises can access Anthropic's advanced AI models via Amazon's Bedrock platform, tailored by Accenture for industry-specific use. The alliance aims to simplify AI integration and address concerns about responsible AI deployment. 

The trio has developed a chatbot for the DC Department of Health, showcasing practical benefits. This collaboration reflects a push for customizable, scalable, and responsible AI solutions. With AWS's infrastructure, Anthropic's models, and Accenture's expertise, the partnership sets a high bar for enterprise AI adoption.

Google hit with $270M fine in France as authority finds news publishers’ data was used for Gemini

France fined Google $270 million for breaching agreements with news publishers and misusing their data to train its AI model, Bard/Gemini. The saga began with EU copyright reforms in 2019, extending protections to news snippets. Google's initial response was to shut down Google News in France, but this was deemed anti-competitive. 

Despite settling previous disputes, Google now faces fines for using publishers' content without proper notification. The French competition authority highlighted Google's failure to provide opt-out solutions for publishers and its opaque payment methodologies. Google accepted the fine but criticized its proportionality, aiming to move forward with fair content use agreements.

ServiceNow’s Washington, DC update empowers IT operations and virtual agents with Gen AI

ServiceNow unveiled its latest "Now Platform Washington, D.C." update, integrating new generative AI features and productivity tools. Leveraging large language models, the update aims to streamline IT operations and enhance virtual agent creation. CEO Jon Sigler emphasizes the importance of responsible AI adoption for business transformation. Notable additions include AI-powered ITOM assistance for analyzing complex alerts and a Virtual Agent Designer utilizing generative AI for chat interactions. 

ServiceNow also introduced AI accelerators and general platform enhancements like Sales and Order Management and Workflow Studio. The update is available now, aiming to boost efficiency and productivity across various workflows.


MindEye2, a breakthrough method in fMRI-to-image reconstructions, cutting down training time from 40 to just 1 hour. Traditional methods require extensive fMRI data per subject, limiting practicality. MindEye2 pretrains across 7 subjects, then fine-tunes on minimal data from a new subject, improving generalization. It maps brain data to a shared-subject space, then to CLIP image space, and finally to pixel space using Stable Diffusion XL. By fine-tuning an unCLIP model, it achieves high fidelity reconstructions. MindEye2 outperforms other methods, enabling accurate reconstructions from a single MRI visit. Notably, it achieves state-of-the-art performance with minimal training data, revolutionizing fMRI-based image reconstructions.

"mPLUG-DocOwl 1.5" enhances understanding of text-heavy images like documents, tables, and charts. It introduces Unified Structure Learning, improving Multimodal Large Language Models (MLLMs). H-Reducer, a vision-to-text module, enhances model efficiency by merging adjacent patches. DocStruct4M and DocReason25K datasets support training, resulting in state-of-the-art performance on visual document benchmarks.

"TnT-LLM" revolutionizes text mining with Large Language Models (LLMs). It automates label taxonomy generation and assignment, reducing reliance on manual curation and domain expertise. The framework iteratively refines label taxonomies and generates training samples for supervised classifiers, enhancing accuracy and efficiency in large-scale text analysis. Applied to Bing Copilot, it outperforms baselines, offering valuable insights for real-world applications.

"AnimateDiff-Lightning" introduces rapid video generation using progressive adversarial diffusion distillation. It sets a new benchmark for few-step video generation, enhanced through modifications tailored for the video modality. Additionally, it distills multiple base diffusion models into a single motion module, expanding style compatibility. The distilled model is publicly available for community use.

"LLMLingua-2" revolutionizes task-agnostic prompt compression for improved efficiency and generalizability. Unlike prior methods, it leverages bidirectional context and introduces a data distillation procedure to compress prompts faithfully. Using token classification and Transformer encoder, it achieves significant performance gains and faster processing, making it versatile across various language models with smaller models like XLM-RoBERTa-large and mBERT.


HubSpot X Jasper - Using Generative AI to Scale Your Content Operations

Lemlist - Build your lead list with verified emails, write and personalize at scale, and send cold emails that actually get customers.

Flashcasts - Flashcards in a private podcast feed 

Dataaku - Extract valuable insights from documents and texts.

Anytalk - Translate audio and video content into your preferred language in seconds

Focal - Convert books and screenplays into movies.



OpenAI’s chatbot store is filling up with spam

OpenAI's GPT Store is overrun with spam and potential copyright violations. Despite moderation efforts, GPTs promoting academic dishonesty and impersonating public figures flood the platform. The rapid growth sacrifices quality, posing legal and ethical dilemmas. OpenAI's monetization plans may exacerbate these issues, mirroring challenges faced by other digital marketplaces. TECHCRUNCH

New Open License Generator helps ensure AI is used responsibly

The new Rail License Generator aids in ensuring responsible AI usage. With customizable licenses, developers can restrict AI model usage, addressing ethical concerns. Adoption of Responsible AI Licenses grows, promoting standardization and flexibility. However, challenges persist in enforcing restrictions and navigating open-source complexities. Standardization and effective communication are crucial for ethical AI deployment. VENTUREBEAT

Using AI to expand global access to reliable flood forecasts

Google Research developed AI-powered flood forecasting to mitigate the increasing impact of floods globally, exacerbated by climate change. Their models, utilizing machine learning, provide accurate real-time predictions, particularly in data-scarce regions like Africa and Asia. Collaborations and open science efforts aim to enhance resilience against flood risks worldwide. GOOGLE

The UAE Is on a Mission to Become an AI Power

The UAE aims to excel in AI, exemplified by its Falcon AI project, making waves globally. With substantial investments and a supportive environment, UAE attracts top talent and partners with tech giants like OpenAI. Despite challenges, the UAE government's ambition drives its quest to shape the future of AI. TIME

An AI-driven “factory of drugs” claims to have hit a big milestone

Insilico Medicine, led by Alex Zhavoronkov, claims a breakthrough in drug discovery using AI, heralding it as the "next amazing revolution" in biology. Their AI-designed drug for lung disease marks a milestone, aiming to accelerate drug development. Despite skepticism, AI's potential in biotech sparks hope for faster, cheaper cures. MIT

Rival nations seek to poach top UK and European AI start-ups

Leading European and UK artificial intelligence start-ups have been targeted by rival nations, such as Canada and the United Arab Emirates, in a global competition to develop cutting-edge AI technology. These nations have offered subsidies, lenient tax regimes, and light-touch regulation to persuade these AI companies to relocate. Canada has already attracted thousands of AI start-ups and researchers with immigration policies and R&D credits, while the UAE has offered long-term visas and tax rebates to AI talent. Despite the approaches, some AI companies expressed their commitment to their current locations due to factors such as a strong talent pool, world-leading universities, and favorable funding environments. FINANCIAL TIMES

AI Breakthrough: Machines Mastering Human Tasks Through Language

Researchers at the University of Geneva have achieved a significant breakthrough in artificial intelligence (AI) by developing a model capable of learning new tasks from verbal or written instructions and then describing them to another AI, enabling it to perform the same tasks. This advancement, mimicking human-like learning and communication, holds great promise for robotics and enhances our understanding of the interaction between language and behavior. The model, simulating brain areas responsible for language processing, marks a crucial step towards machines that can communicate and learn from each other in human-like ways, potentially revolutionizing various industries. NEUROSCIENCE NEWS

8 Google Employees Invented Modern AI. Here’s the Inside Story

In 2017, eight Google researchers revolutionized AI with a paper on transformers, enabling advanced language processing. This breakthrough led to AI marvels like ChatGPT and sparked numerous startups by its authors. Despite Google's initial hesitance, the work's impact reshaped tech, highlighting the transformative power of collaborative innovation in the digital age. 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.

Join the conversation

or to participate.