Amazing new AI system - GPT-3 and generating images from text

[excreationist]

In the future I think AI would be able to generate video in real time and respond to keyboard/mouse input so you can interact with it.

Now AI can generate interactive games though at the moment it is low resolution at 1 frame per second. It generates the game logic just based on game play footage it has viewed. It can generate the game based on text, drawing or photo input.

 

[lpetrich]

Now AI can generate interactive games though at the moment it is low resolution at 1 frame per second. It generates the game logic just based on game play footage it has viewed. It can generate the game based on text, drawing or photo input.

Seems like it's reverse engineering the game logic.

 

[Jimmy Higgins]

Oh dear. I hope they don't have it watch action films and it'll reverse engineer fight scenes where 12 people face one hero... but each fight him one at a time.

 

[excreationist]

The end of this video shows Tesla's Optimus Gen 2 moving an egg from one hand to another....

So much engineering has gone into the hands. In a year, it will be able to thread a needle.

That is mind blowing....

BTW I think Telsa's bots will be able to use a language model so that they can deal with natural language when people make requests.

Well OpenAI's latest robot "Figure 01" is 100 times better... (2:34 minute video)


BTW Telsa's robot involves it initially imitating humans in motion capture type suits/gloves but OpenAI's "Figure 01" learned it itself.

The voice sounds like a pretty awkward guy rather than a professional.

 

[excreationist]

I’m not sure how it learnt its movements. Maybe partly from videos it’s watched?