Google PALM-E replaces robot programmers by combining language, vision & robotics

  • Finally no more software programmers! Does anyone know how to get access to google palm? I am dying for AI and robotics to really add value/convenience.


    With PaLM-E, Google Robotics, TU Berlin, and Google Research present a new AI model that can understand and generate language, understand images, and use both together for complex robot commands.

    PaLM-E’s largest model has 562 billion parameters and combines Google’s massive PaLM language model with ViT-22B, the largest vision transformer to date.


    Google PaLM-E combines language, vision and robotics
    With PaLM-E, Google Robotics, TU Berlin, and Google Research present a new AI model that can understand and generate language, understand images, and use both…
    the-decoder.com

  • Honestly, I'm reminded of a conversation I had with one of my college professors, when I was getting my engineering degree. He was an "old salt", and had lots of war stories about what engineering (and engineering education) was like when slide rules and constant-density graph paper were the most advanced tools most engineers or students had.


    I asked him if he thought students of my generation were learning less because we had fancy graphing calculators (TI-84 4EVAR!!!!). And he said, "Nah, we just made the tests harder."


    In his day, the final exam might be one or two heavy problems that required 3-4hours of pencil math to execute. For my classes, we got a couple dozen problems, often of greater complexity and variety, that would take 3-4hrs to execute even with the graphing calculators and other tools. And the really clever (sadistic?) professors would throw in "ringer" extra-credit questions that were not anything ever encountered in the book or on previous tests, but you could figure out if you really understood the material, as opposed to simply memorizing the formulae.


    No matter how fancy the tools get, all it really does is raise requirements for what has to get done (and how fast and cheap). I don't expect this to be any different.

  • This is cool research and could lead to some general commercial use cases. But I have difficulty believing it will make an impact in manufacturing. Manufacturing must follow an exact set of rules with precision. In order to meet manufacturing goals such as <3.4 defects per million parts, you can't rely on AI making imperfect decisions.

  • i think there will be also a lot of resistance from those who do not want to be at mercy of hackers of system glitches. just like cloud computing, some people will always want to keep their most valuable data and processes safe and not connected to the internet.

    1) read pinned topic: READ FIRST...

    2) if you have an issue with robot, post question in the correct forum section... do NOT contact me directly

    3) read 1 and 2

  • - Offline (Intranet)

    - Powerful edge compute like FPGA's which cost a few thousand quid placed within the Robotic Arm

    - Ai meaning LLM's (Large Language Models) on the Intranet controlling the whole process.


    See this as the future.

Advertising from our partners