PLUS: Mistral's new model Forge and an AI that solves a 100-year-old physics problem
Happy reading
NVIDIA’s latest graphics technology uses generative AI to render photorealistic visuals, but its debut has sparked a major backlash. Gamers are decrying the new feature as an “AI slop” filter that destroys a game’s original artistic vision.
The clash highlights a growing tension between AI-driven realism and the preservation of human creativity. As these powerful tools become more common, will they be embraced as enhancements or rejected for overriding artistic intent?
In today’s Next in AI:
NVIDIA’s DLSS 5 sparks gamer backlash
Mistral launches Forge for custom enterprise AI
An AI model solves a century-old physics problem
Why AI may prefer certain programming languages
NVIDIA’s AI ‘Glow-Up’
Next in AI: NVIDIA’s new DLSS 5 technology uses generative AI to render photorealistic graphics, but its debut sparked a backlash from gamers decrying it as an "AI slop" filter that ruins artistic vision.
Explained:
Unlike its predecessors that focused on upscaling, DLSS 5 is a real-time neural renderer that actively overhauls a game's lighting and textures to create hyper-realistic visuals.
The community reacted with widespread criticism, labeling the effect an uncanny "AI slop filter" that erases the original art direction in favor of a homogenized, glossy look.
NVIDIA’s CEO dismissed critics as "completely wrong," stressing that developers have full artistic control and players can disable the feature entirely.
Why It Matters: This clash highlights a growing tension between AI-driven photorealism and the preservation of human artistic intent. How developers and players adopt these powerful new tools will shape the creative landscape of gaming and beyond.
Mistral's Model Forge
Next in AI: Mistral AI is launching Forge, a new platform that allows enterprises to train their own frontier-grade models using internal, proprietary data. This moves beyond simple fine-tuning to create AI that truly understands a company's unique operational context.
Explained:
Forge enables enterprises to build models from scratch using their private data, offering a deeper level of customization than standard fine-tuning or retrieval-augmented generation (RAG) approaches.
The platform is architecturally flexible, supporting both dense models for general capability and Mixture-of-Experts (MoE) models for greater efficiency, allowing companies to balance performance and cost.
It's designed to be agent-first, meaning autonomous systems can use Forge to fine-tune models, find optimal settings, and even generate synthetic data to continuously improve performance.
Why It Matters: By letting organizations encode their institutional knowledge directly into models, Forge creates a path for AI to become a core strategic asset, not just a third-party tool. This enables the development of highly reliable agents capable of navigating complex internal systems and workflows with greater accuracy.
AI Cracks Physics Problem
Next in AI: A new AI framework called THOR solves a century-old physics problem in seconds, a task that previously took weeks on supercomputers. This breakthrough dramatically accelerates the process of calculating the complex behavior of atoms within materials.
Explained:
THOR directly computes notoriously difficult calculations known as configurational integrals, which are essential for predicting a material's properties. It uses tensor networks and machine learning to overcome the computational limits that have challenged scientists for decades.
In tests on materials like copper and tin, the framework produced results over 400 times faster than traditional simulation methods without sacrificing accuracy.
Developed by researchers at Los Alamos National Laboratory and The University of New Mexico, the project is open-source and available on GitHub for others to use and build upon.
Why It Matters: This advancement enables scientists to simulate and understand material behaviors at an unprecedented speed. It opens the door to faster discoveries in metallurgy, chemistry, and physics by making extremely complex calculations practical for the first time.
Grace Hopper's Revenge
Next in AI: A new analysis flips the script on AI code generation, suggesting a programming language's design matters more than the sheer volume of its training data. Benchmarks show that functional languages like Elixir are outperforming data-heavyweights like Python, hinting at a new direction for AI-assisted development.
Explained:
While major benchmarks like SWEBench focus heavily on Python, a broader analysis shows models excel with functional languages like Elixir, suggesting a language's inherent structure trumps data volume.
Functional languages favor immutable data and pure functions, which means the AI doesn't need to track complex changes in state over time—a task LLMs struggle with.
This paradigm shift optimizes code for human verification instead of human writing, a concept now embodied by the very AI chip architecture that powers this progress.
Why It Matters: This suggests the future of software development isn't just about bigger models, but about choosing tools that align with how AI "thinks." Developers using languages with functional principles may find they get significantly more leverage from AI coding assistants.
The Shortlist
Britannica sued OpenAI alongside Merriam-Webster, alleging the company used nearly 100,000 of their copyrighted articles to train its models without permission.
Microsoft reorganized its AI leadership, centralizing its Copilot teams to allow AI CEO Mustafa Suleyman to focus entirely on its superintelligence and model-building efforts.
Anthropic posted a job opening for a chemical weapons and explosives expert to help create guardrails against "catastrophic misuse" of its AI systems.
Senators urged ByteDance to shut down its Seedance 2.0 AI video generator, calling the app a "glaring example of copyright infringement" in a letter to the company's CEO.