Is anyone else confused by who’s making what at Tesla/XAI ? Tesla to compete with Sora by the end of 2024.

    OpenAI shocked the world this week with a new text-to-video model called Sora. In response to a video about Sora, Elon Musk says that Tesla has been doing real-world (read photorealistic) video generation with accurate physics for about a year, however, this has only been used internally for the development of FSD.

    Musk goes on to explain that the use of this AI-power generative video was only trained on video from the fleet to create synthetic video of edge cases to further train the FSD model. The videos created, look similar to what we see from Tesla’s dashcam video and therefore certainly not as impressive as what Sora offers, but Musk suggests that could be possible later this year.

    Tesla is currently compete-constrained (which may explain why FSD Beta V12, based on end-to-end video training, is not rolling out as rapidly as expected, available just ~100 people outside the company, some 6 months after we saw Musk livestream the first drive of an alpha build.

    Musk suggests they ‘certainly could’ build something similar to Sora, later this year when we (Tesla) have spare capacity. Tesla has announced previously that they were not only investing in their own hardware with Dojo, but also buying as many H100s from Nvidia as they can get their hands on. The total investment is measured in billions of dollars annually.

    So here’s where things get confusing.

    TextOpenAIChatGPTX.AIGrok (via X)

    Musk started X.AI last year and by November, they released their first product, Grok – large language model (LLM) that would compete with OpenAI’s ChatGPT. Grok was integrated into Musk’s other company, X (formerly Twitter) as a service users could interact with with text prompts. This requires a Premium+ subscription on X to access.

    If Tesla is now going create a service for AI-powered generative video to compete with OpenAIs Sora (currently unreleased to the public), then this is likely to create confusion as to why one service is from XAI and the other is from Tesla. Will this also be available via X? Will it be an expansion of Grok’s capabilities, or a whole new service, with a new name?

    Ultimately this feels like a battle of who owns the compute required to acheive this. While the hardware and infrastructure required to create Grok is not small, it is owned by X.AI. The hardware and infrastructure required for FSD development is owned by Tesla. If Tesla is to overbuy to facilitate AI side projects, this is likely to show up in Tesla services revenue and we should expect questions from Shareholders around the scope of new endeavours outside of the FSD focus.

    I think the best possible outcome here is to formalise this agreement between the 3 Musk-owned companies and allow them to leverage hardware/software and services from each other.

    There are certainly additional AI use cases at Tesla, for example, improving voice recognition (this needs a lot of work), and the company leverages AI in their energy division for their Autobidder service.

    AI is clearly going to be an increasing part of our lives, our cars, and our social media.. but this structure certainly needs so cleanup, as the potential revenues that could be derived from success here, won’t be small.

    Jason Cartwright
    Jason Cartwright
    Creator of techAU, Jason has spent the dozen+ years covering technology in Australia and around the world. Bringing a background in multimedia and passion for technology to the job, Cartwright delivers detailed product reviews, event coverage and industry news on a daily basis. Disclaimer: Tesla Shareholder from 20/01/2021

    Leave a Reply


    Latest posts


    Related articles