# Is Grok 4 Showing Signs of "FLUID INTELLIGENCE"? ARK AGI 2 Results | SVIC Podcast & Wes Roth

Source: https://www.youtube.com/watch?v=s7hEop3fbNo
Recap page: https://rapidrecap.app/video/s7hEop3fbNo
Generated: 2025-07-20T19:01:58.643+00:00

---
## Quick Overview

Grok 4 demonstrates non-zero levels of fluid intelligence, a significant advancement attributed to extensive reinforcement learning (RL) training, potentially making it the first AI model to exhibit this human-like cognitive ability. While its performance on traditional benchmarks varies, its rapid development by XAI's lean team, despite high reasoning costs, indicates a new phase in AI innovation where RL plays a dominant role in achieving novel problem-solving capabilities.

**Key Points:**
- Grok 4 exhibits non-zero levels of fluid intelligence, a significant cognitive leap for an AI model, attributed to a 10x increase in reinforcement learning (RL) compute during its training.
- Fluid intelligence, characterized by novel problem-solving and quick learning in new situations, contrasts with crystallized intelligence, which relies on accumulated knowledge, a traditional strength of LLMs.
- While some metrics rank Grok 4 as a top-performing model, others place it significantly lower, highlighting a debate over its true capabilities and the relevance of different benchmarks.
- The rapid development and deployment of Grok 4 by xAI's relatively small team, including building out extensive GPU infrastructure, is considered a remarkable organizational and technical achievement.
- Experts predict a future shift where reinforcement learning will constitute a much larger proportion of AI training compute, potentially leading to successive 'S-curves' of technological advancement.
- The high reasoning costs associated with Grok 4, as shown by the Artificial Analysis Intelligence Index, indicate the computational intensity of its advanced capabilities.
- The strong, even litigious, reactions from other major AI companies and researchers are seen as a 'vibe test' confirming the disruptive potential and genuine innovation of Grok 4.

![Screenshot at 07:27: A detailed chart from the 'Artificial Analysis Intelligence Index' comparing the cost of running various AI models, including Grok 4, across input, output, and reasoning tokens.](https://ss.rapidrecap.app/screens/s7hEop3fbNo/00-07-27.png)

**Context:** Elon Musk's xAI recently released Grok 4, a new AI model that has generated considerable discussion and controversy within the artificial intelligence community. This podcast episode features a panel of experts discussing Grok 4's performance, its underlying training methodologies, and its implications for the future of AI, particularly in the context of 'fluid intelligence' and the broader landscape of large language models (LLMs).

## Detailed Analysis

Elon Musk's Grok 4 model has sparked controversy due to its varied performance metrics, ranking both as a top-tier model and significantly lower on different benchmarks. The discussion highlights that Grok 4 achieved remarkable improvements by applying 10 times more reinforcement learning (RL) compute to a weaker base model (Grok 3). This intensive RL training appears to have unlocked 'fluid intelligence' in Grok 4, a cognitive ability distinct from 'crystallized intelligence' (accumulated knowledge). While large language models (LLMs) typically excel in crystallized intelligence, Grok 4's emergence of fluid intelligence—the capacity for novel problem-solving and quick learning in new situations—is a groundbreaking development, potentially making it the first model to exhibit this trait. Experts suggest that the trend in AI development will see an increasing proportion of compute dedicated to RL training, leading to successive 'S-curves' of technological advancement. XAI's rapid progress in building out the necessary GPU infrastructure and coordinating their team is lauded as a spectacular achievement, especially compared to established tech giants like Google. Although Grok 4's reasoning costs are currently high, the ability of the XAI team to quickly bring such a complex model to market, and the strong reactions from competitors, signal a significant shift in the AI landscape towards models capable of more human-like, adaptive intelligence.

### Grok 4's Performance & Controversy

- Grok 4 exhibits controversial performance, ranking both as a top model and significantly lower on different benchmarks like Yep.ai
- Its development involved applying 10x more reinforcement learning (RL) compute to a weaker base model, Grok 3
- This led to significant improvements in some areas but not others, creating a mixed perception of its capabilities.

### Emergence of Fluid Intelligence

- Grok 4 shows non-zero levels of fluid intelligence, a cognitive ability for novel problem-solving and quick learning in new situations
- This contrasts with crystallized intelligence, which relies on accumulated knowledge and experience, a strength of traditional LLMs
- A founder of ARK AGI 2 noted Grok 4 might be the first model to demonstrate this human-like fluid intelligence.

### Role of Reinforcement Learning

- Extensive RL training is believed to be responsible for Grok 4's fluid intelligence capabilities
- The trend suggests future AI development will increasingly prioritize RL compute, potentially shifting from 90% pre-training to 90% RL training
- This shift is expected to drive further breakthroughs in AI capabilities.

### Technological & Financial S-Curves

- AI advancement follows S-curves, with initial rapid technological innovation followed by a leveling off and then financial innovation
- The current phase of AI, particularly with RL, is in a steep ramp-up, leading to significant performance gains
- This period of 'mania' is characterized by rapid progress and high valuations, before eventually leveling off.

### XAI's Rapid Development & Market Impact

- XAI's ability to quickly build out massive GPU infrastructure and coordinate a lean team is a spectacular achievement
- This rapid scaling allowed them to enter the competitive AI landscape quickly, despite starting with nothing
- The strong, even extreme, reactions from other AGI developers and the market indicate Grok 4's significant impact and potential to disrupt existing models.

![Screenshot at 00:00: Four podcast hosts in a grid layout, with the bottom-left host speaking into a microphone.](https://ss.rapidrecap.app/screens/s7hEop3fbNo/00-00-00.png)
![Screenshot at 00:01: Elon Musk and another individual seated, with Musk speaking and gesturing.](https://ss.rapidrecap.app/screens/s7hEop3fbNo/00-00-01.png)
![Screenshot at 00:03: A graph titled 'Scaling Training' and 'Scaling Test Time' with data points showing performance trends.](https://ss.rapidrecap.app/screens/s7hEop3fbNo/00-00-03.png)
![Screenshot at 01:38: The bottom-left host explaining concepts with hand gestures, emphasizing different types of intelligence.](https://ss.rapidrecap.app/screens/s7hEop3fbNo/00-01-38.png)
![Screenshot at 03:44: The top-right host gesturing with his hand to illustrate the concept of an 'S-curve' in technological advancement.](https://ss.rapidrecap.app/screens/s7hEop3fbNo/00-03-44.png)
![Screenshot at 07:27: A data visualization showing the 'Artificial Analysis Intelligence Index' with various AI models and their associated costs for input, output, and reasoning tokens.](https://ss.rapidrecap.app/screens/s7hEop3fbNo/00-07-27.png)
![Screenshot at 08:50: The bottom-right host making a point with a finger gesture, discussing the 'vibe test' for real technological breakthroughs.](https://ss.rapidrecap.app/screens/s7hEop3fbNo/00-08-50.png)