Custom Chips AI News & Updates
Tesla's Dojo and Cortex: Elon Musk's Custom AI Supercomputers for Self-Driving Cars
Tesla is developing custom supercomputers Dojo and Cortex to train AI models for its Full Self-Driving technology and humanoid robots. The company aims to reduce dependency on Nvidia chips by creating its own D1 chips, with plans to scale Dojo to 100 exaflops by October 2024, though recent communications suggest a pivot toward Cortex as the primary training infrastructure.
Skynet Chance (+0.08%): Tesla's development of immense AI training capabilities aimed at creating "synthetic animals" with human-like perception increases the risk of advanced autonomous systems that could eventually operate beyond human comprehension or control. Tesla's emphasis on proprietary AI hardware-software integration creates potential for uniquely capable systems with limited external oversight.
Skynet Date (-3 days): The massive investment in proprietary AI compute infrastructure specifically designed for training autonomous systems suggests an acceleration in the development timeline for human-level AI perception and decision-making in physical environments. Tesla's commitment to deploy robotaxis by mid-2025 puts pressure on rapidly advancing these capabilities.
AGI Progress (+0.1%): Tesla's development of custom AI hardware optimized for neural network training represents significant progress in scaling AI computing infrastructure toward AGI-necessary levels. The company's integrated approach to hardware and software, combined with real-world data collection from millions of vehicles, creates a uniquely powerful capability focused on perception and decision-making.
AGI Date (-3 days): Tesla's massive investment in custom AI compute infrastructure (targeting 100 exaflops) and its aggressive timeline for unsupervised FSD by 2025 suggests an acceleration in the development of AI systems capable of human-level visual perception and decision-making in complex environments.
OpenAI Trademark Filing Reveals Plans for Humanoid Robots and AI Hardware
OpenAI has filed a new trademark application with the USPTO that hints at ambitious future product lines including AI-powered hardware and humanoid robots. The filing mentions headphones, smart glasses, jewelry, humanoid robots with communication capabilities, custom AI chips, and quantum computing services, though the company's timeline for bringing these products to market remains unclear.
Skynet Chance (+0.06%): OpenAI's intent to develop humanoid robots with 'communication and learning functions' signals a significant step toward embodied AI that can physically interact with the world, increasing autonomous capabilities that could eventually lead to control issues if alignment isn't prioritized alongside capabilities.
Skynet Date (-2 days): The parallel development of hardware (including humanoid robots), custom AI chips, and quantum computing resources suggests OpenAI is building comprehensive infrastructure to accelerate AI embodiment and processing capabilities, potentially shortening the timeline to advanced AI systems.
AGI Progress (+0.05%): The integrated approach of combining advanced hardware, specialized chips, embodied robotics, and quantum computing optimization represents a systematic attempt to overcome current AI limitations, particularly in real-world interaction and computational efficiency.
AGI Date (-3 days): Custom AI chips targeted for 2026 release and quantum computing optimization suggest OpenAI is strategically addressing the computational barriers to AGI, potentially accelerating the timeline by enhancing both model training efficiency and real-world deployment capabilities.