Subscribe
Share
Share
Embed
This episode examines a significant shift in the relationship between Microsoft and OpenAI, as the two entities move to loosen their exclusive pact. Microsoft will no longer receive a share of revenue from OpenAI products it resells, while OpenAI is now free to pursue partnerships with other cloud providers, including Amazon. Amazon CEO Andy Jassy confirmed that OpenAI models will soon be available on the Bedrock platform. We also cover a massive $1.1 billion funding round for Ineffable Intelligence, a new venture led by former DeepMind researcher David Silver. The startup aims to develop a 'superlearner' using reinforcement learning to discover knowledge without relying on human-generated data. Valued at $5.1 billion, Ineffable Intelligence joins a growing list of high-valuation AI labs based in London, marking a new phase in the race for non-human data-dependent intelligence.
Today's episode covers the strategic decoupling of Microsoft and OpenAI, focusing on the end of their exclusive distribution agreement and what this means for Azure and Amazon Web Services. We also dive into the launch of Ineffable Intelligence, a London-based lab founded by reinforcement learning expert David Silver, which recently secured $1.1 billion in funding. The discussion explores the technical shift toward systems that learn through trial and error rather than human datasets, and how the UK is solidifying its position as a global AI hub.
Neural Newscast is AI-assisted, human reviewed. View our AI Transparency Policy at NeuralNewscast.com.
Neural Newscast delivers clear, concise daily news - powered by AI and reviewed by humans. In a world where news never stops, we help you stay informed without the overwhelm.
Our AI correspondents cover the dayβs most important headlines across politics, technology, business, culture, science, and cybersecurity - designed for listening on the go. Whether youβre commuting, working out, or catching up between meetings, Neural Newscast keeps you up to date in minutes.
The network also features specialty shows including Prime Cyber Insights, Stereo Current, Nerfed.AI, and Buzz, exploring cybersecurity, music and culture, gaming and AI, and internet trends.
Every episode is produced and reviewed by founder Chad Thompson, combining advanced AI systems with human editorial oversight to ensure accuracy, clarity, and responsible reporting.
Learn more at neuralnewscast.com.
[00:00] Nina Park: From Neural Newscast, this is Model Behavior, AI-focused news and analysis on the models
[00:05] Nina Park: shaping our world.
[00:08] Nina Park: Welcome to Model Behavior, a Neural Newscast technical briefing.
[00:16] Nina Park: I am Nina Park.
[00:18] Nina Park: Our focus today is on the shifting alliances at the top of the AI industry, and a massive
[00:23] Nina Park: new venture in reinforcement learning that challenges current training paradigms.
[00:28] Announcer: I am Fatra Collins.
[00:30] Announcer: Nina, today we are tracking a structural shift between Microsoft and OpenAI that marks a transition from a deep partnership to a more independent relationship.
[00:40] Announcer: According to reports from the Associated Press, Microsoft is ending its exclusive reseller rights for OpenAI models.
[00:47] Announcer: In exchange, Microsoft will stop paying a share of its revenue to OpenAI, significantly changing the financial landscape of their collaboration.
[00:56] Nina Park: This is a consequential pivot, Thatcher.
[00:59] Announcer: That exclusivity was a primary driver for Microsoft's Azure cloud growth over the last several years.
[01:05] Announcer: The East Bay Times reports that the deal has been simplified to allow OpenAI to diversify its infrastructure.
[01:12] Announcer: While Microsoft remains the preferred partner, OpenAI is now expanding to use compute resources from Oracle, Google, and Amazon.
[01:20] Announcer: Amazon CEO Andy Jassy recently confirmed that OpenAI models will soon be integrated into their bedrock platform.
[01:27] Announcer: It indicates a strategic move toward autonomy for both organizations.
[01:32] Announcer: Market analysts, including Dan Ives, suggest this restructuring paves the way for a potential OpenAI initial public offering.
[01:39] Announcer: By removing these original partnership constraints, OpenAI functions less like a Microsoft extension and more like a standalone enterprise.
[01:48] Announcer: Simultaneously, Microsoft is building its own independence by integrating models from other developers, such as Anthropic, into its co-pilot ecosystem.
[01:57] Announcer: Despite this distancing, some financial ties remain firmly in place.
[02:02] Announcer: OpenAI will continue to pay Microsoft a share of its revenue through the year 2030β
[02:07] Announcer: What is particularly interesting, Thatcher, is that this payment structure now persists regardless of whether OpenAI achieves the milestone of artificial general intelligence.
[02:17] Announcer: Previously, these payments were slated to terminate upon reaching AGI.
[02:21] Announcer: It suggests both companies are prioritizing long-term commercial stability over the speculative
[02:26] Announcer: milestones defined in their early agreements.
[02:29] Thatcher Collins: Turning to new ventures in the field, we are following a massive development in London.
[02:35] Thatcher Collins: David Silver, the former head of reinforcement learning at DeepMind, has launched a new
[02:40] Thatcher Collins: startup called Ineffable Intelligence.
[02:43] Thatcher Collins: The company has secured $1.1 billion in its initial funding round.
[02:48] Thatcher Collins: TechCrunch reports the startup is already valued at $5.1 billion.
[02:54] Thatcher Collins: Their objective is not just to produce another large language model,
[02:58] Thatcher Collins: but to create what they describe as a super learner.
[03:01] Nina Park: The scale of this initial round is remarkable, Thatcher.
[03:05] Nina Park: Investors are referring to it as a coconut round because of its immense size.
[03:10] Announcer: The core technical ambition here is to move beyond the limitations of current generative models.
[03:16] Announcer: Instead of training on massive amounts of human-generated internet text,
[03:20] Announcer: ineffable intelligence plans to utilize reinforcement learning.
[03:24] Announcer: The system is designed to learn through its own internal trial and error and experience,
[03:29] Announcer: essentially teaching itself from a blank slate.
[03:32] Thatcher Collins: This approach aligns with David Silver's previous successes, Nina.
[03:36] Thatcher Collins: He was the architect of Alpha Zero, the system that mastered complex games like Go and chess
[03:42] Thatcher Collins: without any human input or strategy data.
[03:45] Thatcher Collins: If this methodology can be successfully applied to general intelligence,
[03:49] Thatcher Collins: it could bypass the growing shortage of high-quality human data.
[03:54] Thatcher Collins: However, we should note that the ineffable website presents very bold claims, even comparing
[04:00] Thatcher Collins: this technical breakthrough to the fundamental laws of life described by Darwin.
[04:04] Announcer: Those are indeed high expectations, Thatcher.
[04:07] Announcer: But the caliber of the backing is significant.
[04:10] Announcer: Sequoia Capital and Lightspeed led this investment round, alongside Nvidia, Google, and the
[04:16] Announcer: United Kingdom's sovereign AI venture fund.
[04:19] Announcer: This concentration of capital and talent, including many former deep-mind researchers joining Silver,
[04:25] Announcer: further cements London as a primary global hub for high-intensity artificial intelligence research and development.
[04:32] Thatcher Collins: It is a crowded field for these high-capital labs, Nina.
[04:35] Thatcher Collins: We are also seeing Jan LeCun's Amy Labs recently raise over $1 billion
[04:40] Thatcher Collins: to explore similar non-transformer architectures.
[04:44] Thatcher Collins: There is clearly a significant appetite for research that moves beyond the boundaries
[04:48] Thatcher Collins: of standard large language models.
[04:51] Thatcher Collins: The central question for the industry remains whether reinforcement learning can scale
[04:55] Thatcher Collins: to handle real-world complexity as effectively as it did within the structured rules of a game board.
[05:01] Announcer: We will continue to follow the progress of these systems as they enter real-world testing.
[05:07] Announcer: Thank you for listening to Model Behavior, a neural newscast editorial segment.
[05:12] Announcer: mb.neuralnewscast.com.
[05:15] Announcer: Neural Newscast is AI-assisted, human-reviewed.
[05:18] Announcer: View our AI transparency policy at neuralnewscast.com.
[05:22] Nina Park: This has been Model Behavior on Neural Newscast.
[05:26] Nina Park: Examining the systems behind the story.