Biggest data moments past 2 weeks: OpenAI’s next model, Microsoft’s AI expansion, and the rise of LLLMs

?? Top stories of weeks 45 - 46

ChatGPT search is here

OpenAI just launched ChatGPT search, a built-in search engine that finally lets ChatGPT access live internet info directly, moving away from the dependency on plugins like Bing. Developers can now quickly get real-time data, making it easier to find current facts without sifting through links. Think of it as a streamlined, AI-powered Google alternative, tailored for fast answers and digging deeper with ease.

Microsoft expands Azure AI infrastructure

Microsoft has launched new AI capabilities in Azure, including expanded Event Hubs features and enhanced data analytics infrastructure for AI workloads

NVIDIA unprecedented growth

NVIDIA reported unprecedented growth in AI infrastructure, with Q1 fiscal 2025 revenue reaching $26.0 billion, marking a 262% increase driven by accelerated computing and generative AI adoption.

?? Breaking: OpenAI's next frontier model "Orion" expected soon

OpenAI is reportedly preparing to launch its next major AI model, codenamed "Orion," marking a significant milestone in AI development. Unlike previous releases such as GPT-4 and Q*, this launch represents a strategic shift in deployment strategy. Initially, access will be restricted to select partner companies for product development, rather than immediate public release through ChatGPT. Key developments:

• Microsoft engineers are reportedly preparing Azure infrastructure for Orion deployment as early as November
• The model is considered a successor to GPT-4, though the final branding remains unclear
• OpenAI executives have suggested it could be up to 100 times more powerful than GPT-4
• OpenAI researchers celebrated training completion in September
• The release coincides with OpenAI's recent $6.6 billion funding round and corporate restructuring

However, there's some controversy around the release timeline. After initial reports, OpenAI CEO Sam Altman labeled the December release news as "fake news. " OpenAI's spokesperson later clarified that they "don't have plans to release a model code-named Orion this year" but will be releasing "other great technology." This development coincides with significant changes at OpenAI, including the departure of key executives like CTO Mira Murati, Chief Research Officer Bob McGrew, and VP of Post Training Barret Zoph. The timing and nature of this release could significantly impact the AI industry landscape and set new benchmarks for large language model capabilities.

?? Industry insights

What the next U.S. President likely means for AI regulations

Donald Trump’s approach to AI is poised to stir things up. His plan includes dismantling Biden’s AI regulations, cutting red tape for startups, and ramping up competition with China. While pushing for fewer safety regulations, he’s also eyeing military AI projects. His "light-touch" regulatory approach could boost innovation but may shift control to the states. Though he acknowledges AI’s dangers, Trump is focused on accelerating U.S. dominance—potentially at the cost of global governance and safety standards. The stakes for AI’s future are high. To know more, check his convention speech, where he made one comment about it.

Microsoft's framework for LLM chatbot evaluation:

Microsoft's data science team has developed an evaluation framework for chatbots using large language models (LLMs). It focuses on seven key areas:

• Search performance in retrieval-augmented generation (RAG) systems
• Task completion and false information identification
• Response metrics
• Responsible AI usage
• Resource efficiency
• User sentiment analysis
• Understanding user intent

The framework includes measures like Precision@K and A/B testing, emphasizing the value of user feedback. This approach aims to standardize chatbot evaluations to enhance performance and ensure ethical practices.

Palantir: Leading the charge in enterprise AI

Palantir, founded by Alex Karp and Peter Thiel in 2003, is uniquely leveraging the AI boom. While many AI companies focus on consumer-facing products, Palantir specializes in helping organizations make sense of huge amounts of data. Its platforms, Gotham and Foundry, support everything from intelligence agencies to businesses in connecting critical data points. Palantir sets itself apart by embedding engineers with clients, creating customized solutions to fit each need. It's reshaping the enterprise AI space. Learn more here.

The rise of LLLMs

Local Large Language Models (LLLMs) are changing the AI landscape by letting companies and developers work with powerful AI tools entirely on their own devices. This approach not only protects sensitive data but also reduces dependence on cloud services, which can be costly and potentially less secure. Designed to run efficiently on smaller hardware, LLLMs open new doors for businesses to integrate AI without compromising privacy.

Setting up LLLMs requires tuning them to fit specific hardware and business needs, making these models highly customizable. This flexibility has broad appeal, especially in sectors where data security is paramount, such as healthcare, finance, and research. With LLLMs, businesses gain control over AI workflows and can build unique, secure solutions that align with their specific goals.

?? Tool of the fortnight

QuillBot is an AI-powered writing assistant that helps you take your writing to the next level, with features like paraphrasing, summarizing, and grammar checking to make your content clearer and more polished.

?? Pro tip: Deepnote now lets button blocks set a variable when clicked, making it easy for creators to respond to specific button actions right in their Python code. Curious how it works? Take a look at this Loom video for a quick demo! With this feature, users can build even more dynamic apps with multiple buttons, each designed to trigger its own action.