Weekly AI News Roundup

By Brian Wang

1. SanctuaryAI unveiled the seventh generation of its general-purpose robot Phoenix. The latest generation robot and its AI control system, Carbon™, draw even closer to that of a person, with wide-ranging improvements to both the hardware and AI software. The announcement comes less than 12 months after Sanctuary AI’s sixth-generation robot.

The changes include:

Increased uptime, making robots more available for training and data capture

Increased build and commissioning speed, allowing us to bring more units online faster

Reduced bill of materials, lowering the cost of manufacture

Significant hardware improvements, increasing range of motion in the wrists, hands, and elbows, and increasing hand durability

Further miniaturized hydraulics, reducing the weight, power consumption, and complexity with increased hardware and software measures that exceed specified safety standards

Improved visual acuity and tactile sensing, leading to higher-fidelity data to train the AI control system

The time it takes for new tasks to be automated has gone from weeks to less than 24 hours, marking a major inflection point in task automation speed and autonomous system capability

2. XAI is close to raising $6 billion at a valuation of $18 billion. Venture Capital Firms like Sequoia are among the investors. XAI will be able to buy 100, 000 Nvidia H100 chips to train the Grok 3 model which will be competitive with GPT5.

OpenAI is worth $80 billion.

3. Chinese firm SenseTime launched a new LLM. It can beat GPT-4 Turbo across nearly all key benchmarks. The ~600 billion parameter model features a 200k context window and was trained on over 10TB of largely synthetic data.

SCMP - SenseTime’s shares gained as much as 36 percent after the company launched the latest iteration of its SenseNova large language model.

There are currently more than 200 domestic-developed AI models in China.