AI: Weekly Summary. RTZ #541
...week ending November 15, 2024
Industry concerns on Slower AI Scaling: Media reports are focused on both OpenAI and Google potentially showing signs of slower benefits of ‘AI Scaling’, with their next generation model training efforts. For OpenAI the focus seems to be on “slowing rates of improvement” for their next generation GPT-5 ‘Orion’ models, which at this point in its training cycle is “seeing increases in quality smaller compared with the jump between its prior GPT-3 and GPT-4 models”. For Google, the reports stress a “struggle to achieve performance gains in its Gemini” models “at the same rate it did last year”. Both companies and others are also working hard on new applications of their next gen model efforts to perform better ‘AI Reasoning and Agentic’ tasks. This indicates a shifting of the definitions and core applications of ‘AI Scaling’ vs prior approaches. Particularly for new applications like ‘Synthetic Data’ and ‘Synthetic Content’. These should continue to justify the industry’s AI Compute Capex ramp for datacenters and power. More here and here.
Google releases Gemini AI app for iOS: Google now joins OpenAI in a standalone AI app for consumers, with a Gemini multimodal AI app for iOS (Android presumably to come). The launch is expected to be a wider rollout globally, leveraging Google Gemini’s leadership in its own AI compute infrastructure. The app can be queried by both text and voice, and is generally getting a positive early reception. It will also of course be an onramp for Google’s next iteration of Gemini, expected soon, likely concurrent with OpenAI’s release of GPT-5 (Orion) early next year. The Google App will also be able to uniquely leverage Google’s other apps available globally. Currently the app can be used to play music off YouTube, with other cross-app capabilities to come. More here.
Enterprise AI Search sees growing competition: The Ai Search competition on the Consumer side between OpenAI, Google, Perplexity and others, is seeing a concurrent focus on AI Search in Enterprises. Glean is a fast growing startup that is getting a response from the bigger LLM AI companies, as well as database companies like Snowflake and others. Ironically, the two leading startups in AI Search for Consumers (Perplexity) and the Enterprise (Glean), are founded and led by ex-Googlers Aravind Srinivas and Arvind Jain. The enterprise market for Search has its own specific requirements. Especially in how businesses leverage their own Data for AI applications and Search across their internal and partner domains. More here.
Small AI Models get even smaller: Small Language Models (SLMs) get more capable with ongoing next gen models by Microsoft, Google, Meta and others. New research is suggesting even smaller models, dubbed Tiny Language Models (TLMs) and Micro Language Models (MLMs), are showing promise in applications ranging from local device inference on low power, to synthetic data and content applications. These models are even smaller than a billion parameters, and are showing efficacy in tightly defined, vertical applications. This is a concurrent development with the traditional LLM AI models continuing to scale to trillions of parameters and beyond, in ever larger AI data centers. More here.
Mideast Sovereign Funds ramp up on AI Investments: Mideast Investment fund MGX, a combination of UAE sovereign funds Mubadala and G42, is ramping up a $100+ billion AI Fund for investments in the Mideast, US and elsewhere. This continues similar focus by sovereign funds in Saudi Arabia and other Gulf countries for AI investments, particularly centered around AI Data Center and Power investments. MGX is reportedly also in discussions with investing in Elon Musk’s xAI, after investing $500m+ in OpenAI’s recent $6.6 billion round. xAI reportedly raised $5 billion at a $45+ billion valuation, double the value a few months ago. These activities poing to a ramp in multiple entities in the region focused on AI Infrastructure investments led by US tech companies and VCs in the near term. More here and here.
Other AI Readings for weekend:
New study on growth of AI Adoption slowing among US workers, led partly by how AI is viewed in the workplace. More here.
Notable piece on how ChatGPT AI impacted the leading online education company Chegg. More here.
Up next, the Sunday ‘The Bigger Picture’ tomorrow. Stay tuned.
(NOTE: The discussions here are for information purposes only, and not meant as investment advice at any time. Thanks for joining us here)