Which sector will invent AI first?

[Deleted]

Porn or military.

None of the others come even close.

[breadisfunny]

Definitely sexbots, that's how it happened in Terminator.

hopefully it's succubibots.

[twiddler]

Porn or military.

None of the others come even close.

No way. Ignoring that data and patents are key, the porn industry pays garbage, and every maths/CS Ph.D flocks to a top tech/fin company, the porn industry isn't involved in the advancements of computer vision, NLP, robotics, etc., why in the world would they suddenly be an innovator in sentient AI? I can't even think of a porn site with a proper recommender while many still kill my boner with gay porn on the front page despite knowing damn well that men make an overwhelming majority of their market.

[Deleted]

No way. Ignoring that data and patents are key, the porn industry pays garbage, and every maths/CS Ph.D flocks to a top tech/fin company, the porn industry isn't involved in the advancements of computer vision, NLP, robotics, etc., why in the world would they suddenly be an innovator in sentient AI? I can't even think of a porn site with a proper recommender while many still kill my boner with gay porn on the front page despite knowing damn well that men make an overwhelming majority of their market.

The first porn AI won't be developed by some porn company trying to make a buck on it. It'll be developed by someone somewhere for their own personal pleasure. That counts as porn in my book.

[Annoying]

Interesting seeing no discussion of the academic sector. It's generally ahead in the field of machine learning and usually is the basis for any of these weak AI's that we see playing jeopardy or chess or go.

Then again, I'm probably biased, having a degree in machine learning and all.

[Dellis0991]

The military probably would....they turn everything into a weapon.

[Kujako]

I'm gonna go with Russian bot-nets.

[Deleted]

I'm gonna go with Russian bot-nets.

I don't think anyone who writes code would want to write code that could then think for itself, thus making the writer of the code obsolete. I think those who do write code are the first ones to realize how dangerous AI could actually be.

Script kiddies who use "tools" like LOIC, sure. I'm sure they'd think it'd be extremely hilarious to release an AI into the interwebz, but they typically can't write any code anyway, so that's not a problem.

[twiddler]

I don't think anyone who writes code would want to write code that could then think for itself, thus making the writer of the code obsolete. I think those who do write code are the first ones to realize how dangerous AI could actually be.

Script kiddies who use "tools" like LOIC, sure. I'm sure they'd think it'd be extremely hilarious to release an AI into the interwebz, but they typically can't write any code anyway, so that's not a problem.

I'm now reminded that you're the guy who thinks the asynchronous web paradigm is bad...

[Independent voter]

I don't think anyone who writes code would want to write code that could then think for itself, thus making the writer of the code obsolete. I think those who do write code are the first ones to realize how dangerous AI could actually be.

Script kiddies who use "tools" like LOIC, sure. I'm sure they'd think it'd be extremely hilarious to release an AI into the interwebz, but they typically can't write any code anyway, so that's not a problem.

It's like we will be creating demons.

[twiddler]

Interesting seeing no discussion of the academic sector. It's generally ahead in the field of machine learning and usually is the basis for any of these weak AI's that we see playing jeopardy or chess or go.

Then again, I'm probably biased, having a degree in machine learning and all.

All else fails, bet Stanford, right?

[willtron]

Military or vehicle manufacturers fo sho.

[unbound]

If you are looking for something closer to true AI like you see in SciFi movies, it will most likely be the military.

Most businesses are looking to produce things like Deep Blue which responds to very specific areas of focus (such as post moderation and CSR replacements; as well as simple labor replacement). The games industry is very slowly improving the scripting of PvE enemies, but much of the industry is just looking to leverage PvP because it is much cheaper.

The military, on the other hand, is increasingly looking to replace humans with robotic combatants, and that requires information processing and analysis along far more dimensions than other industries need to currently consider. How do you determine what is current threat? Future threat? Friend or foe (IFF does get damaged)? Intelligent coordination with other units? Responsiveness to current conditions (old military saying, "No military plan survives engagement with the enemy")? Responsiveness to new, novel conditions? Ability to consider / coordinate multiple types of inputs (optical, other electro-magnetic, audio, pressure sensors) in conjunction with all of the above?

[TrumpIsPresident]

Holy shit! That's a really nice hat! Can you tell me where I can get one of those?!

Cute. Except all that stuff is happening already and I'm sure these people would love a way to automate it.

[PC2]

My guess would be one of the many private sector teams and I don't think they need to have Google level computing resources. The publicly funded organizations have the biggest computers, however HPC's use highly distributed memory and only work well on highly decomposable problems. They can take up the volume of a tennis court 6 feet high. Most of what we know about intelligence from the brain suggests it requires something closer to shared memory for efficient task and data parallelism.

[twiddler]

Cute. Except all that stuff is happening already and I'm sure these people would love a way to automate it.

Cute. Except all that stuff isn't really happening already and I'm sure these people already can implement given their ocean of data, their own advancements in social graphs, and their understanding of applied statistics. You don't need sentient AI nor do you need 100% accurate semantic search to implement what you're proposing. You're like those guys on Yahoo who scream that Yahoo is censoring conservative comments despite the place being full of nothing but conservatives.

[Unholyground]

One of my close friends has a double masters in computer science and has been working on creating a learning AI for the last like 6 years, and he said it is pretty much impossible with how computers currently work. However when quantum computing is readily available he said it is going to happen there is no doubt in his mind.

[Kellhound]

We already have AI, its called a BA degree.

[Pengekaer]

The games industry is very slowly improving the scripting of PvE enemies,

Are they? The hostiles of Far Cry (2004) are by and far the smartest enemies I've encountered to date. They actually did try to flank you and flush you out of hiding spots with grenades. AI was mostly decent in Crysis too. Nobody masters gaming-programming like Crytek does, too bad they suck at most other things.

Anyways, good post. I agree it's probably the military that will do it first. Drones that make the decision to attack or not without the delay that human involvement in decision-making would cause, perhaps something like armed MULE-robots lying in ambush?
Or perhaps nuclear submarines controlled by an AI? Never having to surface, they would be much less visible. And no humans means no infirmaries, no food-storages or dormitories to take up space. And furthermore, AI-controlled nuclear subs would also not suffer from reduced crew efficiency during long-term patrols (humans trapped in cramped spaces for a long time tend to go kuck kuck). If such an AI on a nuclear-sub was also a true AI, I guess it could also do decision-making on its own without contact with the outside world, like Washington?

[TrumpIsPresident]

Cute. Except all that stuff isn't really happening already

Wait so...

...stories about refugee rapes, assaults, child molestation, riots, violence, etc, aren't being burried
...911 tapes aren't being edited
...mentions of radical Islam aren't being scrubbed from speeches, documents, transcripts, intel reports, etc
...and social media moderators (twitter, fb, reddit) aren't deleting information critical about the agenda they support

...that's what you're telling me?

Hmmm, I guess you're living proof of the exact sort of goal I spoke about when I said:

AI could be programmed to sense out anything that is critical of the agenda that whoever develops it supports.