Researchers built a computer with human brain cells and taught it to play Pong

Sounds like a headline from The Onion. It's not. Welcome to 'Organoid Intelligence'.

Why is this man always taking off his shirt? I've seen his nipples more often than my own. Source: StarTrek.com.

This just in: Researchers in Australia have merged human and mouse brain cells with silicon and taught it how to play Pong.

No, this is not an episode of Black Mirror (though it probably will be).

The title of the scientific paper in which this news was announced is somehow even more chilling: In vitro neurons learn and exhibit sentience when embodied in a simulated game world.

I shall pause briefly while you hide under your bed for a few minutes.

The project is called DishBrain because it consists of 800,000 brain cells living in a dish (duh), on top of an array of electrodes that measure their activity and stimulate them with electrical signals. The experiment is being conducted by a company called Cortical Labs, which just got a $600,000 grant from Australia's Office of National Intelligence. Apparently the ability to bat a pixelated ball between two imaginary paddles is vital to national security in the land down under. Fair dinkum, mate.

Apologies for the soundtrack. Source: The Guardian.

They taught the cells to play pong by rewarding them with predictable electrical signals when the paddle returned the ball correctly, and feeding them unpredictable/chaotic stimulus when they whiffed. Apparently predictable stimulus is good, unpredictable is bad, and our brains know this on an extremely granular level.

The next thing they plan to do is get it drunk. Nope, not kidding. Per New Atlas:

By building a living model brain from basic structures in this way, scientists will be able to experiment using real brain function rather than flawed analogous models like a computer.

[Lead author Dr. Brett ] Kagan and his team, for example, will next experiment to see what effect alcohol has when introduced to DishBrain.

“We’re trying to create a dose response curve with ethanol – basically get them ‘drunk’ and see if they play the game more poorly, just as when people drink,” says Kagan. [1]

The idea is that by growing actual neurons in a Petri dish, researchers can better understand how real not-at-all-creepy-and-terrifying brains actually work. But it's also about building much smarter AI systems to embed into self driving cars, drones, delivery robots, and hyper-intelligent machines that will enslave and eventually eliminate humankind. [2]

OI vey!

The researchers call these half human/half electronic computers "organoid intelligence" (OI), which doesn't make me feel any better at all. [3] They're hoping to eventually develop an OI device containing 10 million neurons, or about the size of a tortoise brain. [4]

Allow me to play some peppy music while we await our inexorable extinction. Source: Science in Public.

This feels like some kind of tipping point to me. I mean, people have been talking about The Singularity for friggin’ ever (or since Ray Kurzweil published the book in 2005). There's a growing movement of whacko highly intelligent, perfectly reasonable cult members humans who believe that the merger of human and machine is not only inevitable, it is our destiny. Kurzweil himself predicted we'd hit that point roughly 20 years from now.

"I have set the date 2045 for the ‘Singularity’ which is when we will multiply our effective intelligence a billion fold by merging with the intelligence we have created," wrote Ray Kurzweil. [5]

He also predicted that AI wouldn't pass a Turing Test [6] until 2029, either. So we're about six years ahead of schedule.

The thing is, they're not wrong. The next generation of medical devices will be nanobots that live inside our bodies and fix things when they break, like tiny robot mechanics. AI will grow increasingly more 'humanish.' It will become harder to tell where the machines stop and the humans begin.

But I think their timing may be a little off. Frankly, I'm not sure the planet will remain habitable long enough for us to find out.

Robot Pong

A small production note: I use Midjourney to generate most of the art for this blog, because a) it's wicked good at it, and b) I give them 10 simeolas a month and want to get my money's worth. So today I asked it to "draw two robots playing ping pong" to use as my lead art for this post.

This is what it came up with.

So the AI knows exactly what a ping pong table is, and what a friendly robot looks like, but has no friggin’ idea of how people (or robots) would play the game. (Like, not using chopsticks, for example.) I find that stupidly reassuring.

Maybe our benevolent digital tortoise overlords won't be that smart after all.

Wouldn't you really rather be a robot? Share your thoughts in the comments below. Also, if you like this kind of shite, please subscribe and tell two friends (and they'll tell two friends, and they'll tell two friends, and so on, and so on).

[1] I, on the other hand, get much better at playing games when I'm drunk. Also, better looking.

[2] Ha ha. Just kidding. I think.

[3] These guys really could use some help with marketing.

[4] These systems would require a fraction of the energy required to run a data center — a few watts vs 15,000 megawatts. So serving our computer-turtle masters will be better for the environment, at least.

[5] He also predicted that "by the 2030s, we will connect our neocortex, the part of our brain where we do our thinking, to the cloud.” Amazon Brain Services? Thanks, but... no.

[6] Not to be confused with CAPTCHA tests, where humans are clearly dumber than machines. (Is that a lower case h or an n? Stoplight or streetlight? What the hell is that squiggly thing?)