Cool but I'm more a text-presenting guy. Only if I pre-film do I attempt to do video to present my work.
Does anyone here but me and openAI actually have any generated text or image completions? If not, nor any evaluation that measures predictions, it's not AGI. What can it do for me if not find patterns/ create new discoveries? If you have something bleeding edge then it would have as good scores/ results as the things seen on openAI.com, i.e. really matured results, little own no results :D. "These insights have been used to create Understanding Machines that learn any human language simply by reading un-annotated text, and using a million times less energy than GPT/3 or BERT." I have heard this said before, such as about Pei Wang's work. I have not ever heard of how such AGI works though (where does the Predicted Next Letter COME from, explain how each mechanism improves prediction what type of pattern it finds for each mechanism), nor ever saw any generated data results, ever. You simply can't get something that free. BTW, GPT3 is over trained, it does really good on 1GB, little own 40GB as in GPT[2]. Sure the brain needs more but that'd be the final stage and only if can't add other pattern finding mechanisms do you resort to adding more data. Not saying you can't make an AI efficient, but there isn't a thousand times faster version, not even 100 times faster, only maybe 10 times and you are openai2 and amazing. ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/Td5767526440a5310-M546a1093338feee0ee922d12 Delivery options: https://agi.topicbox.com/groups/agi/subscription
