Today in AI

• Today’s big story

• News roundup

• AI research changing the world

• The tool we’re loving

• This week on the podcast

What’s happening in AI right now

AI Models evolve: supercomputers and synthetic reality

The AI landscape is rapidly transforming, with new models pushing the boundaries of what's possible in natural language processing, content generation, and synthetic media. Recent developments highlight both the remarkable progress and persistent challenges in creating AI systems that are not only powerful, but also efficient, robust, and ethically sound.

Superclusters and the pursuit of scale

Elon Musk's xAI made waves with its Memphis Supercluster announcement, claiming to have built the "most powerful AI training cluster in the world" with 100,000 liquid-cooled Nvidia H100 GPUs. This massive infrastructure investment underscores the ongoing trend towards larger language models trained on ever-expanding datasets.

While such computational behemoths promise unprecedented capabilities, they also raise questions about the sustainability and practicality of this approach. As we'll see, raw power alone doesn't guarantee optimal performance in real-world applications.

The achilles heel of complex models

A recent experiment exposed surprising vulnerabilities in state-of-the-art language models. Researchers pitted an advanced LLM chatbot against simple bots, revealing weaknesses in the AI's ability to manage certain interactions.

The study found that the LLM continued engaging with nonsensical or repetitive prompts far beyond what a human would tolerate. More tellingly, the simple bots proved to be 50,000 to 6 million times more computationally efficient than the LLM. This stark disparity highlights potential vulnerabilities in high-volume interaction scenarios, especially as AI chatbots become more prevalent in customer-facing applications.

These findings underscore a critical challenge for model developers: creating systems that not only process information but also manage conversations with human-like discernment and efficiency.

AI's creative leap

While some researchers probe AI's limitations, others are expanding its creative horizons. The MarioVGG project by Virtuals Protocol demonstrates both the potential and current constraints of AI in recreating interactive content like video games.

Trained on over 730,000 frames of Super Mario Bros. gameplay, the model can generate limited gameplay based on simple prompts. However, MarioVGG faces challenges such as glitches and restricted actions, highlighting the complexity of AI game recreation.

This project offers valuable insights into the future of AI-assisted content creation, suggesting a potential democratization of game development and other interactive media. However, the current limitations of MarioVGG also serve as a reminder that we're still in the early stages of this technology.

Synthetic humans: the next frontier

Perhaps the most striking model development comes from Infinity AI, which has created a foundation video model capable of generating lifelike AI characters that can speak and move naturally. This technology can produce videos up to 30 seconds long using a single image input, along with audio and other conditioning signals.

Trained for 11 GPU years at a cost of $500,000, this model represents a significant leap forward in synthetic media creation. Its potential applications span entertainment, education, and marketing. However, it also raises ethical concerns regarding deepfakes and identity rights.

The success of this technology will depend not only on technical advancements but also on its integration into existing workflows and how society addresses the ethical implications of increasingly realistic synthetic media.

Optimizing model deployment

As models grow more complex, efficient deployment becomes crucial. Hugging Face's Text Generation Inference (TGI) is making strides in this area, offering significant advantages in cost, privacy, and customization for Large Language Model deployment in production environments.

TGI's architecture, consisting of a Router and Inference Engine, optimizes LLM performance and resource utilization. The system focuses on key metrics and employs advanced techniques to improve inference performance, addressing both compute and memory-bound challenges.

TGI's successful implementation at Adyen for their internal Generative AI platform demonstrates its potential to make advanced AI technologies more accessible to enterprises.

News roundup

The top stories in AI today.

FUTURE OF WORK

Jobs are changing — are you?

• How Power BI and Tableau Empower Marketers in the AI-Driven Economy

• Human Cognition and the Evolutionary Leap Being Driven by AI

• How AI Assistants Will Revolutionize IT Operations with Smart Data Tools

• AI Boosts Developer Productivity by 26% in Landmark Study

• Why Generalists Will Thrive In The AI-Driven Economy

NEW LAUNCHES

The latest features & products in AI innovation.

• Infinity AI Creates Lifelike Video Characters from Single Images

• MSI Unveils 8-Inch Lunar Lake AI-Powered Gaming Handheld for 2025

• New Gemini Advanced Referral Program Lets Subscribers Give 4 Free Months to Friends

• Roblox to Unveil AI Tool that Generates 3D Game Environments

• AI-Powered Personalization Transforms Qatar Airways’ Marketing Strategy

GADGETS

Computers, phones, wearables & other AI gizmos.

• AI-Powered Samsung Galaxy Laptops Push Computing Boundaries

• Qualcomm, Samsung, Google to Partner on AI-Powered Mixed Reality Glasses

• AI-Powered Samsung Galaxy Book4 Edge Now $450 Off at Best Buy

• Samsung Releases AI-Powered One UI 6.1.1 Update for Galaxy Devices

• AI-Powered Lenovo Laptop Tracks Users and Responds to Voice

• How to Watch Apple’s ‘Glowtime’ iPhone 16 Event — And What to Expect

MODELS

Deployment, research, training, infrastructure

• Novel Experiment Demonstrates That Advanced Doesn’t Always Mean Better AI

• New AI Video Tool Recreates (Glitchy Version) of Super Mario Bros

• Try The ‘Self-Ask’ Technique Next Time You Have a Complicated Task for AI Chatbots

• How to Use HuggingFace’s ‘TGI’ to Deploy LLMs at Scale

• Capx AI Launches 8B-Parameter Multimodal Vision Model

• Salesforce Unveils AI Models to Change the Game for Sales Organizations

GOVERNMENT

Press releases, regulation, defense & politics.

• Welcome to the AI Election: Deepfakes, Misinformation and Personalized Propaganda

• What the Latest International AI Treaty Means for US Tech Giants

• 4 New Experts Join U.S. National AI Advisory Committee

• AI Training Violates Copyright Law, New Study Finds

• AI’s Influence on the 2024 US Election is Raising Concern

• Xockets Sues Nvidia and Microsoft Over DPU Market Practices

FINANCE & ECONOMICS

Economics + the movers & shakers on Wall Street.

• How the AI Funding Boom is Reshaping the Venture Capital Landscape

• 3 Industrial Stocks Poised to Benefit from Soaring AI Demand

• Broadcom Projects $12B in AI Sales for 2024, Beating Expectations

• Publishing Giant Wiley is Reorienting Itself as an AI Licensing Company

AI research changing the world

The latest breakthroughs and most pivotal papers — broken down in language anyone can understand.

Advertising Association Report: AI Revolutionizes Advertising, Promising 58% Time Savings

Imagine a world where AI doesn't just create ads, but revolutionizes the entire advertising ecosystem - from strategy to compliance, while simultaneously addressing pressing ethical concerns. This future isn't just a possibility; it's unfolding now, according to a groundbreaking report from the Advertising Association's AI Taskforce. The study reveals that early AI adopters in advertising are already seeing up to 58% time savings and 30% quality improvements, fundamentally altering the industry's landscape. But the implications stretch far beyond efficiency gains. The report showcases AI's potential to enable richer creative exploration, from AI-generated brand mascots to personalized audio content, suggesting a future where AI-human hybrid creative teams become the norm. However, this transformative power comes with a responsibility.

The report emphasizes the critical need for ethical AI adoption, addressing issues like privacy, fairness, and sustainability. In fact, it suggests that AI could reduce carbon emissions in ad creation by a staggering 99%, potentially catalyzing sustainability efforts across industries. As AI reshapes advertising, it raises profound questions about the future of creativity, data-driven decision making, and even consumer behavior. Could AI-driven advertising inadvertently influence societal norms?

The AI tool we’re loving right now

The best way to get AI literate? Try the tools!

Other tools worth trying

• Apiway.ai - A no code integration platform for Facebook and TikTok ads experts

• Apollo.io - Find, contact, and close your ideal buyers with streamlined engagement workflows powered by AI.

• ChatPDF - Chat with any PDF, and navigate legal contracts, financial reports, manuals, and training material. Ask questions to any PDF for fast insights.

This week on the podcast

Can’t get enough of our newsletter? Check out our podcast Future-Proof.

Get more with a Pro account

Paid members get access to discounts on AI tools, expert-written tutorials and deep industry data and leaderboards.