General Science/Technology/Futurist Videos Thread

[wjfox]

[Yuli Ban]

There's a lot of neat phenomena that lies just beyond our daily lives that we might rarely ever see
Here's a Kern arc by the way

[Yuli Ban]

[Yuli Ban]

[Yuli Ban]

Astrum covers something I've mentioned for about a decade now
While human labor is absolutely necessary to get off-world colonies ready and fit for human living, the vast majority of construction will need to be done by machines ahead of time. The advancement of robotics will directly contribute to our ability to actually colonize space.
I know we could have theoretically gone to Mars in the 1980s, but as I've long said, until we have even first-generation AGI-powered robots, an actual colony is unfeasible. But once we have that, the stars are the limit

For the record, my idea of what is "first generation AGI" hasn't fundamentally changed over the past 6 years, but I've been able to hone it down into something more tangible. Back in 2020, 2021 or so, I said that a first generation AGI is "Wolfram Alpha, Siri, GPT-3, DALL-E, and WaveNet all in a single AI model"

Nowadays that sounds exactly like an omnimodal model. We have AI models that are multimodal— we literally just got 4o multimodal with native image generation. However, it's still too expensive to do more than two modalities (text and image, or text and audio). Let alone putting a reasoning model on top of that, and let alone adding agents to that.
But if we can do all that, and economically, we could expand the modalities to include video, 3D, and eventually outright physical modalities like tactile, olfactory, gustatory, proprioception, etc. As well as a more robust mathematical and coding ability (we could get AI models that are mathematically capable, but LLMs aren't those). An omnimodal agentic reasoning model with an extremely long context window is, to me, a first-generation AGI. Not necessarily the full artificial human mind we want, but the first genuinely general AI. So we're not far. And once that's licked, we're likely going to see a crazy boom in space exploration as a side effect.
After all, one of the whole reasons why generative AI was pursued was because of the successes in image recognition and vision systems. And if you can get an AI to recognize and label what a cat is, you can do the inverse, giving an AI a label and telling it to reconstruct an image of a cat. That was all accomplished back in 2012, even Starspawn0 was talking about it when he was on here back then, saying clearly where ImageNet and other advances in deep learning were going to lead.
By 2032, the outright "AI art only" era will probably be over when generative AI's true purpose comes into focus: giving machines the ability to understand and model the real world. ANY world.

[wjfox]

[funkervogt]

We're in a weird moment of history where it's unclear whether robots like this are real or whether the video was produced for about $10 using an AI whose capabilities match that of a multimillion-dollar CGI studio from 20 years ago.
https://www.facebook.com/share/r/14GehiFTkRS/

[Yuli Ban]

[Yuli Ban]

Just because I was curious (this was in the video itself to be fair, but I wanted a dedicated image)

[Yuli Ban]

[wjfox]

[wjfox]

[wjfox]

[wjfox]

[wjfox]

[zzz]

https://i.imgur.com/Lp4AYZY.jpeg

[wjfox]

[Yuli Ban]

[wjfox]

[wjfox]