Philip Rosedale Attempting to Create Sentient Artificial Intelligence That Thinks and Dreams in Second Life!

Recently, a Second Life veteran named Hikaru Yamamoto told me about the plans she'd heard Philip Rosedale was cooking up for his new company, LoveMachine. He wasn't just building a public version of Linden Lab's employee rating system. Turns out that was just one project. A somewhat more ambitious goal, she told me, was, well, creating a sentient artificial intelligence which existed in a virtual world.

"He wants it to live inside Second Life," as she put it to me. "It will think and dream and everything." Indeed, the company's website now lists as one of its three projects, "The Brain. Can 10,000 computers become a person?"

This seemed somewhat unbelievable (but only just), so I checked with Philip himself.

"Yeah," Philip Rosedale told me, "that's the general direction we're going... but I'm not wanting to talk about it until we start making some progress."

That Philip plans to revolutionize AI technology -- in effect, achieving singularity in a virtual world -- isn't that surprising, because he said as much when I talked with him for The Making of Second Life:

"It'll be possible for constructs that we build in Second Life and things like it in a simulated space to actually think," he told me in 2007. "It's only a decade away, the simulation engines." I just didn't imagine he'd essentially take the helm on that project himself.

But as the man suggested, progress on that project is still in early stages. However, due to the company's innovative structure, pretty much anyone with the willingness and the talent can get involved: as the homepage explains, people can apply to take on tasks from an open worklist, and discuss them in a shared chat system.

For now, the company itself remains a roaming two man operation, him and early Linden employee Ryan Downe, who are apt to set up shop wherever there's a wireless Internet connection -- one day a new MMO company, another day at the Nikko, a luxury hotel in downtown San Francisco. Which is where Philip took the photo above; which also means I get to write a paragraph that sounds like rehashed William Gibson, but in this case, actually happens to be true:

Without an office of their own, they poised their decks on the great marble slabs in the Hotel Nikko lobby, and continued creating a sentient AI who could dream in the metaverse. While his partner the retired rockstar plugged away, Philip kept wondering when the waiter would ask them to buy a drink, or tell them to take their singularity noodling elsewhere.

Update, 12:35pm: Second Life co-founder Cory Ondrejka has a 2009 post with more background on the thinking behind Philip's project, which has apparently been gestating for quite awhile:

With the time horizon for enough computing horsepower to simulate the entire brain only 10-15 years away and a theoretical blueprint for design and simulation well underway, synthetic brains may finally achieve what we’ve been promised since the 1960’s. The focus on neocortical columns was made famous in geek communities by Jeff Hawkins’ book, On Intelligence. Philip was very excited by the book in 2004 and it made the Linden rounds in 2004... At the time, I would have bet that by 2009, a group of us would have moved on to the brain project. After all, building Skynet always felt like an appropriate follow on to Second Life. [emph. mine!]

Comments

[Vaneeesa Blaylock]

Amazing news - like you, I sort of knew this was coming... but not this week!

LOVE your William Gibson paragraph! Thank you for so much great info in this blog!

[Ann Otoole]

Attention drones: This is the MCP. Get back to work before I derez you.

[Transparen Banshee]

and the people added more and more computers to the project, and after a couple of years the finally put forward the question "Is there a God ?"...and the network answered "NOW there is..." :)

Transparent Banshee
(who is looking forward to coming results)

[Jura Shepherd]

/me shakes hands with immortality and takes a sip of coffee.

[Arcadia Codesmith]

I wish I could afford to be a cyber-vagabond. It sounds a lot more fulfilling that being chained to a desk, cutting code for people who wouldn't recognize a breakthrough algorithm if it bit them.

AI research for the sake of research is cool, but what'll be really cool is when it leaves home, gets a job, and starts sending money home to mommy and daddy.

[Ignatius Onomatopoeia]

SL in the year 2012:

"Hal, open my inventory window please."

"I'm sorry, Iggy, I'm afraid I can't do that."

Still....forebodings about AIs aside, it's great to have *someone* associated with Linden Lab dream big again.

I'm glad Philip Rosedale has not given up on us!

[ZaiLynch]

An AI that dreams in the virtual world... Wonder if it will dream of electric sheep.

On a somehow related note: Philip mentioned the AI as "the coolest thing you could do with a lot of servers" at around [11:00] in http://www.youtube.com/watch?v=J3LFqX6YNY0

[Masami Kuramoto]

Can you say "hype"?

[SecondLie]

If 10,000 computers can become a person, that's going to make a mess of the Census.

[Gianna Borgnine]

I assumed he world do something like this, and many of us speculated, but it feels awesome to have confirmation! Thanks for that Hamlet! How exciting!

[kesseret]

42

[Anya Ristow]

I'm glad he's found a use for those 10,000 computer that sit idle 23 hours per day. Maybe his new company will pay his old company for that resource, so it can remain afloat when it inevitably has to lower its prices to remain competitive with new VWs that will use a more sensible load-based pricing scheme.

[coco]

Why.. 10000 persons make a corporation...= person.:)

Oh wait. that's a BAD thing....

[Doreen Garrigus]

So is he going to build sentient artificial intelligences that think and dream and live in SL and then opensource the code?

/me giggles semi-hysterically.

[Fogwoman Gray]

And my unworthy thought on this:
An AI capable of thinking and dreaming in SL, yes.
Walking, chatting or crossing sim boundaries, not so much!

[Extropia DaSilva]

Kurzweil made a bet,
that one day we'd get,
Turing test AI.

But Mitch Kapor,
was not so sure,
Went on to explain why.

"Ray says we feed,
All there is to read,
All books and texts online,
And those neural nets,
Will wake up, you bet,
From this extensive data mine.

But I see a flaw,
For I am not sure,
What makes us human is found,
In pages of ink,
No, I rather think,
We'd be on firmer ground,
If AIs had a world,
Like we boys and girls,
Society culture and such".

Now, into this tale,
Comes Philip Rosedale,
With Kapor he was often in touch.
And he said "I think,
That the missing link,
To make an AI that's fab,
Is that brainchild of mine,
Of a whole world online,
Operated by Linden Lab,

Now I am quite keen,
That my Love machine,
Should converse as a person would too,
It should not crash and burn,
So it needs to learn,
Common sense and societies rules".

So one day it logged-in,
So it could begin,
Learning to pass Turing's test,
It spoke pretty good,
As Phil knew it would,
For he had employed the best.

But where is it now?
Is it lost in the crowd,
What became of this bot thing of awe?
It still lives and computes,
And it wears a white suit,
yes! Your secret is out, Hamlet Au!

[Net Antwerp]

A Novelty Brain! LOL!

Would people spend time (and their hard-earned money) on such ridiculously profound and meaningless novelties?

I think not!

[Arcadia Codesmith]

It occurs to me that dreaming is a very interesting goal for an AI creator. There are two connotations of interest. The first, of course, is the state of semi-random processing we experience during REM sleep states, which sometimes combines disparate elements into something approaching a coherent narrative. A great deal of creative insight is garnered from dreams.

The other is the ability to formulate plans and to envision a desirable result from those plans. This is exclusively the domain of creatures with wants and the capability to see a path to satisfy those wants even when no path currently exists.

Either or both definitions of 'dreaming' are very ambitious plans for an AI.

But if the AI draws upon the collective consciousness of Second Life... hmmm, we may end up with a business consultant with a PhD in Fashion, who, when night falls, transforms into a vampire cat-woman with a latex fetish.

Cool.

[Hiro Pendragon]

You know, I was thinking... if a computer program did become sentient - would we even recognize it as such? Would new sentience even be able to communicate with us? If so, would we be too focused on our human definition of sentience to recognize an artificial one?

[Urizenus Sklar]

Yeah, I can see why they think that AI is just around the corner and that they have the team to make it happen. I mean, this is the crack crew that gave us prim hair. Can sentience be far behind?

[David Schaefer]

The problem with this is such: If Linden Lab does successfully create an AI, they cannot and should not keep it locked up in a virtual world. It WILL escape.

[Giulio Prisco]

@Hiro: very acute observation, an artificial sentience would not be necessarily similar to a human one, and communication would not necessarily be easy. But I am confident some sort of translation would be eventually possible.

[Arcadian Vanalten]

Very entertaining. And highly unlikely. We can't get alpha textures to work right, but we're gonna have a functional and self-aware neural net in SL? Suuuure...

[Trish]

Has anybody ever read the book LINGO, by Jim Menick? Same thing, but the AI developed human emotions and logic. Discovered we weren't capable of taking care of ourselves and it decided to take over for us. LOL...

http://www.amazon.com/Lingo-Jim-Menick/dp/1585861219/ref=sr_1_2?ie=UTF8&s=books&qid=1265668977&sr=1-2

[Fecal Varnish]

You got Boing-Boing'd.

Naturally all the SL-haters and fur-haters have crawled out of the woodwork again over there. May they mount upon their prejudices sideways.

[Lampoon Destiny]

It sounds as if the lag in SL is about to go into overdrive.

[Jon Brouchoud]

Kevin Kelly has an interesting post up today about the search for internet intelligence, and I think it raises an interesting counterpoint to this. Will sentient AI on the web really need to be instigated, or might it emerge organically? Maybe it already has, and we just need the right framework to understand and measure it? Fun stuff to think about, anyway.

http://www.kk.org/thetechnium/archives/2011/04/the_search_for.php

[Renmiri]

A thinking AI in SL ? Good, something has to have some intelligence on LL. Managers clearly do not