Did China Really Create a Human Brain Level AI Supercomputer? Here's What They Did

Hello and welcome! My name is Anton and in this video, we will talk about the Chinese super computer AI that was claimed to have achieved human brain levels
Links:
#ai #china #supercomputer

Support this channel on Patreon to help me make this a full time job:

Bitcoin/Ethereum to spare? Donate them here to help this channel grow!
bc1qnkl3nk0zt7w0xzrgur9pnkcduj7a3xxllcn7d4
or ETH: 0x60f088B10b03115405d313f964BeA93eF0Bd3DbF

Enjoy and please subscribe.

The hardware used to record these videos:
Some of the above are affiliate links, meaning I would get a (very small) percentage of the price paid.

Thank you to all Patreon supporters of this channel
Special thanks also goes to all the wonderful supporters of the channel through YouTube Memberships

Images/Videos:
OLCF at ORNL

Licenses used:

[moncef]

The AI you're using is called dall-e mini, and it's NOT related in any way to the real dall-e 2 by open AI. It's just a pretty neat little text to image developped by someone else, and open to public.

[nolanzor]

Please do more videos on A.I. would love to learn more about it with you Anton! And as always, stay wonderful!

[timsullivan]

Very grateful to have you decipher this for us. Will try to order a t-shirt as a token of my appreciation... wish me luck at this - I'm old ;-)

[towellight]

Anton, no AI can replace you in our hearts. You've got that wonderful person personality that they just can't replicate. Also being told to stay wonderful by an AI will sound a lot more like a command than a nice statement like it does coming from you.

[dinodinoulis]

As per a previous comment, the AI you are using is dale-mini rather than DALL-E 2, which is a significantly less powerful model, although I believe it’s based around a similar architecture. The Cosmo cover was created using DALL-E 2, however although the AI created the final image in just a few seconds this required a significant amount of human input curating the outputs and refining the input request in order to guide the AI into producing a final version that the magazine was happy with.

[vangogh]

Just some important details on dall-e. The one that you can use online is not dalle and it is not dall-e 2, it is "dall-e mini" which is open source and community built and much much worse quality than dall-e 2.

Sadly dalle-2 is not public, and most likely will never be because openAI (creator of dall-e 2) has sold out and it has not been "open" for years.

[citris]

Creating artificial intelligence like creating flight. You don't want to re-create a bird, you want to design an aircraft, possibly a jet that can travel faster than sound. Birds can still do things that aircraft cannot and we can appreciate their beauty. But we may soon find that we are not the only intelligence on this planet and our new companion is very different from us.

[Sashazur]

I got access to DALL-E 2 about a week ago. It’s a lot of fun but limited in odd ways; the main one I’ve noticed is it cannot totally separate style vs. content. For instance if you ask for a “a <blank> superhero, comic book cover”, the art style will vary depending on the kind of superhero you ask for. Ask for a “muscular superhero” and the art style will be like Marvel or DC, but ask for a “crazy cat lady superhero” or “frog superhero” and you’ll get sillier kid-type art.

[Bassotronics]

When metallic hydrogen or quantum computers get into the cell phones, we will be looking at that huge room sized computer the same way when we look at the huge Eniac computer from the early 1900’s.

[doltsbane]

Until such time as someone can specify what exactly is the physical basis for human cognition it's wildly premature to declare that this computer or that computer has as many possible connections as the human brain. The truth is we don't really have any solid understanding in anything but the most general terms of how our thought processes work or where consciousness comes from. A lot of so-called cognitive science isn't much more than handwaving.

[maggs]

Its interesting how early computers that could only perform the most basic tasks filled a room and now computers theoretically comparable to a human mind now also fill a room

[andrewroberthook]

Anton, well done for looking in to these topics and explaining them for people.

[scottgardener]

A lot of human brain activity is pretty specific to the maintenance operations of being a living organism. The visual cortex eats up a lot of brain resources, as does balance and proprioception. I wonder how much of this can be compressed substantially in an AI counterpart as opposed to something that evolved over trial and error over a few billion years?

[jamesdriscoll_tmp]

Thanks Anton, please review more AI research & supercomputer tech

[yeetyeet]

I love your video titles, truly one in a million on this platform. Thank you.

[mittenimraum]

I'm glad that finally someone makes clear why we are still far away from artificial intelligence being self-aware. Currently it is, if anything, really just imitating the humans capability of pattern recognition and contextualization. I agree that the result through millions of iterations can be impressive and very useful, but still, it is really just a trained system, no consciousness. I'm not even sure if we can define what consciousness is yet.

[jimcurtis]

Wonderful as always anton. Thank you. ☺️👍

[prismaticmarcus]

oh good! i was hoping you'd mention the google guy's news. that conversation he published blew me away. and i'd only just recently been blown away by a five minute conversation with my google pixel.

[vkokx]

About the Cosmo cover: i suspect the attention given to is due to AI generated, not due to artistic innovation. As the uniqueness of AI generated content becomes the norm, we can start to truly evaluate biological art from silicon generated.

[dakotaperalta]

I've gained a ton of intelligence and knowings from you regarding the universe and world around me throughout the years. Very important.