Artificial Intelligence & Robotics

[Eco]

I have been doing some reading, and listening to different podcasts on this topic, and the invitable future that we face where certain jobs will become obsolete for humans, as machines with A.I. will be safer, quicker, and just more efficient (and cheaper!).

Of course, more jobs will likely open up, at least that's the assumption, but to this day I can't think of a lot areas in which new jobs will be created if/when this really starts become the norm.

Does any of this ever freak you out? With having a baby on the way, I have caught myself numerous times thinking what kind of world my daughter will have.

So, discuss all things A.I., pros and cons.

[Bluebird Hewitt]

Ironically, there was something I heard on the radio the other day that a child born this year may not need a driving test when they reach 17 as driverless cars will be commonplace on the roads. 

That would freak me out and provide concerns as well, simply because AI isn't perfect (ie. AI would need to be updated, upgraded, repaired etc).

Just my small thought.

[nudge]

It's a topic that I've always been extremely interested in and I try to follow the latest developments as best as I can. I find the idea of designing machines that are able to think, learn and perceive extremely fascinating, not to mention the somewhat sci-fi idea of self-awareness.

 I think AI has a lot of potential to improve our lives, especially when it comes to health and medicine, space exploration, science and industry, as they are cost-effective, can operate in pretty much any conditions and achieve greater precision and accuracy with low chances of error compared to humans. At the same time, it obviously has downsides as if machines are preferred over actual humans by employees, we will have a bigger problem of unemployment  and job security than any time in history, so there's definitely a risk of social problems that would arise as a result. My biggest worry is probably the growing dependency on machines in general, resulting in diminishing mental and cognitive abilities of humans in a long-term. There's also huge concerns of misuse of AI in general.

[Eco]

6 minutes ago, nudge said:

It's a topic that I've always been extremely interested in and I try to follow the latest developments as best as I can. I find the idea of designing machines that are able to think, learn and perceive extremely fascinating, not to mention the somewhat sci-fi idea of self-awareness.

 I think AI has a lot of potential to improve our lives, especially when it comes to health and medicine, space exploration, science and industry, as they are cost-effective, can operate in pretty much any conditions and achieve greater precision and accuracy with low chances of error compared to humans. At the same time, it obviously has downsides as if machines are preferred over actual humans by employees, we will have a bigger problem of unemployment  and job security than any time in history, so there's definitely a risk of social problems that would arise as a result. My biggest worry is probably the growing dependency on machines in general, resulting in diminishing mental and cognitive abilities of humans in a long-term. There's also huge concerns of misuse of AI in general.

Exactly. I love technology and always enjoy reading about the new endeavors and things we are able to do now, that we weren't before, thanks to AI and machines.

However, soon enough ours wars will be started with machine vs machine, we will have a total disconnect of what it is to be a free thinking individual. Or at least that's the fear.

I find myself curious about war. I think we are a species are quicker to go to war with someone because we have upgraded form the swords and bows and arrows, into long range rifles, missiles, and long range technology. We forget what it's like to kill someone in war, and the seriousness of it since nowadays you can do it from 1,000 miles away without dealing the the visual, mental, or other stress related to it. So, while it's fascinating for me, it's also scary as I would suggest we are quickly becoming more reactionary, and making quick, deadly decisions, without talking and trying to solve issues without blood shed.

[Mel81x+]

Terribly vast topic but also equally fascinating. Not so arsed about the interactivity portion of it more so the actual patterns and code that go behind making these things happen. I remember reading months ago about a system that learned to laugh at jokes and remembered my time in college where we had a professor who specialized in the topic talking about machines learning humor and deciphering enough to understand the even bigger problem with AI which is context. The machine can learn numerous ways of parsing a language based on decisions trees and language maps but what it really has to learn through trial and error with a lot more decisions is how to use the maps in the context they were meant for. 

Its also nice timing since this just came out.

https://techcrunch.com/2017/05/23/googles-alphago-ai-beats-the-worlds-best-human-go-player/

That AI is learning very very fast. Its one of the most complex games to map and I sometimes wonder if Deep Blue had that kind of backing whether Kasprov would have even stood a chance.

[Eco]

3 minutes ago, Mel81x said:

Terribly vast topic but also equally fascinating. Not so arsed about the interactivity portion of it more so the actual patterns and code that go behind making these things happen. I remember reading months ago about a system that learned to laugh at jokes and remembered my time in college where we had a professor who specialized in the topic talking about machines learning humor and deciphering enough to understand the even bigger problem with AI which is context. The machine can learn numerous ways of parsing a language based on decisions trees and language maps but what it really has to learn through trial and error with a lot more decisions is how to use the maps in the context they were meant for. 

Its also nice timing since this just came out.

https://techcrunch.com/2017/05/23/googles-alphago-ai-beats-the-worlds-best-human-go-player/

That AI is learning very very fast. Its one of the most complex games to map and I sometimes wonder if Deep Blue had that kind of backing whether Kasprov would have even stood a chance.

Kasparov lost to deep blue, so my guess is that if they were to hold that match today, or even use Magnusson, that the human wouldn't have a chance. 

[Mel81x+]

4 minutes ago, Eco said:

Kasparov lost to deep blue, so my guess is that if they were to hold that match today, or even use Magnusson, that the human wouldn't have a chance. 

Didn't he beat the first iteration? Thats kind of how I remember it but after it learned there was really no saving grace. Also have to remember that processing power has come a long way since then.

Yep looked it up. The 1996 edition wasn't as good as it should have been.

https://en.wikipedia.org/wiki/Deep_Blue_versus_Garry_Kasparov

Edited June 7, 2017 by Mel81x

[Eco]

Just now, Mel81x said:

Didn't he beat the first iteration? Thats kind of how I remember it but after it learned there was really no saving grace. Also have to remember that processing power has come a long way since then.

It was the best of out 6, done back in the 90's. The first series Kasparov won, but he lost the second. The second set were the one televised here in America, and Kasparov had MATE IN 3, missed it, and ended up losing the match. He has sense referred to it as the biggest blunder in his career. 

To put that in perspective, if you download the latest version of Grandmaster on your photo, the processing power on that alone is enough that it would have beaten Big Blue and any human ever. Kasparov has teamed with some of the best chess players in the world for a team event where its' human vs machine, and they, as a collaborated effort, can't beat the latest chess technology. It's unbelievably fascinating. 

[Mel81x+]

https://www.theguardian.com/technology/2018/apr/05/killer-robots-south-korea-university-boycott-artifical-intelligence-hanwha

Seems South Korea is trying to build autonomous weapon systems and the fear has already sparked a shutdown of the project. Can't say I'd blame anyone for trying to shut it down as I can't imagine how they'd really be able to distinguish real targets from false positives. Then again AI has come a long way since its primary logic block inception and KAIST is making some stellar progress.

[nudge]

While I think that autonomous weapon systems is a very bad idea in general, I don't think AI is the problem - people are. Lowering of the cost associated with killing someone is one of my biggest issue with this.

It reminded me of a short video I've seen a while ago:

 

[RandoEFC+]

Stuff like that is apocalypse territory. There always needs to be some input from a human being when it comes to taking out a target or threat, even if the AI does all the aiming and calculations and just leaves the human to press the ok button.

[Mel81x+]

On 4/5/2018 at 17:38, nudge said:

While I think that autonomous weapon systems is a very bad idea in general, I don't think AI is the problem - people are. Lowering of the cost associated with killing someone is one of my biggest issue with this.

It reminded me of a short video I've seen a while ago:

 

After watching that I was half tempted to go and buy a drone to try some 'homebrew' activities out. But it will never be a reality based on the kind of AI development that we currently have. No system has that much horsepower to do what the video in the space suggested ... not yet atleast. But what I have seen is a system that can laugh at jokes and use secondary comprehension to make sense of sentences that have a totally different meaning to what is actually written (on your garden variety processor). These kinds of macro-advancements in AI lead to much better usage when it comes to searching and making recommendations on what kind of condom is best suited for bedding the partner of your dreams but crumble horribly when trying to make tactical strike decisions. There's just too much data involved to be able to provide a precise strike option and the other factor that targets in motion cause the system even more of a headache because the stabilizing component requires to be in constant calculation mode. 

As was mentioned just above its mostly hocus-pocus right now but I'd like to imagine that with the advancements in robotics and the nano-form-factor we are seeing in prosthetic science that the day isn't very far where making tactical strike-engine based AI that uses robotic mobility to deploy weapons that can neutralize threats in the war field isn't very far away. 10 years ago if you lost your leg you were granted a stump and a captain hook walking style for the rest of your life. Try telling that to the Robotics division of MIT, Boston Dynamics or even the guy in Israel who built the mine detector system today and you'd be laughed at.

The system would also need time to learn and make ALOT of mistakes but if you're interested in seeing a system gain advance in a closed-space watch this real-life video. Granted its a game but what it does with its learning is astounding. 

 

 

[Panflute]

In short robots are going to murder us all.

[Honey Honey]

We'll have sex on tap though 

[nudge]

On 4/6/2018 at 20:22, Mel81x said:

After watching that I was half tempted to go and buy a drone to try some 'homebrew' activities out. But it will never be a reality based on the kind of AI development that we currently have. No system has that much horsepower to do what the video in the space suggested ... not yet atleast. But what I have seen is a system that can laugh at jokes and use secondary comprehension to make sense of sentences that have a totally different meaning to what is actually written (on your garden variety processor). These kinds of macro-advancements in AI lead to much better usage when it comes to searching and making recommendations on what kind of condom is best suited for bedding the partner of your dreams but crumble horribly when trying to make tactical strike decisions. There's just too much data involved to be able to provide a precise strike option and the other factor that targets in motion cause the system even more of a headache because the stabilizing component requires to be in constant calculation mode. 

As was mentioned just above its mostly hocus-pocus right now but I'd like to imagine that with the advancements in robotics and the nano-form-factor we are seeing in prosthetic science that the day isn't very far where making tactical strike-engine based AI that uses robotic mobility to deploy weapons that can neutralize threats in the war field isn't very far away. 10 years ago if you lost your leg you were granted a stump and a captain hook walking style for the rest of your life. Try telling that to the Robotics division of MIT, Boston Dynamics or even the guy in Israel who built the mine detector system today and you'd be laughed at.

The system would also need time to learn and make ALOT of mistakes but if you're interested in seeing a system gain advance in a closed-space watch this real-life video. Granted its a game but what it does with its learning is astounding. 

 

 

Watched the video today, and while I tried hard to focus, I couldn't do anything else but wonder how cringey it was  Are all "e-sport" event full of such awkwardness?  The public, the host, and the "champion" were all so cringey I actually rooted for the bot to win 

Anyway, back on topic - while this OpenAI Dota bot is obviously impressive, is it any more impressive than Deepmind's AlphaGo Zero?

[Mel81x+]

14 hours ago, nudge said:

Watched the video today, and while I tried hard to focus, I couldn't do anything else but wonder how cringey it was  Are all "e-sport" event full of such awkwardness?  The public, the host, and the "champion" were all so cringey I actually rooted for the bot to win 

Anyway, back on topic - while this OpenAI Dota bot is obviously impressive, is it any more impressive than Deepmind's AlphaGo Zero?

Haha yes they are very quite cringey that way. You should see the finals its quite the spectacle.

Back on topic, I think there's very little difference between the OpenAI bot and Deepmind's AlphaGo Zero, both learn through experience and trial-and-error and that's really where the AI limitation is. The time to learn and ramp up quickly enough to beat an opponent isn't guaranteed by simply seeing the playing field because the system isn't capable of understanding all the permutations and combinations by knowing the rules and applying them as-is. This is where the next level of AI is taking precedence where it can gain the the environment faster than the two AI's we've mentioned. When that day comes the bot in the game and AlphaGo will never have to play a single game and beat its opponent purely on data about the environment. If it laughed at the opponent once its won I'd start getting worried but we're not there yet and it will probably take another 3 to 5 years atleast.

[nudge]

Found this to be a very interesting read:

https://www.theatlantic.com/international/archive/2018/07/artificial-intelligence-religion-atheism/565076/

 

@Mel81x @Eco

[Dr. Gonzo]

On ‎4‎/‎7‎/‎2018 at 00:28, Harvsky said:

We'll have sex on tap though 

I, for one, can't wait to be killed by my SexBot6000

[Mel81x+]

On 7/24/2018 at 13:11, nudge said:

Found this to be a very interesting read:

https://www.theatlantic.com/international/archive/2018/07/artificial-intelligence-religion-atheism/565076/

 

@Mel81x @Eco

I like the idea and while I can get behind the system being capable of resetting what is more interesting is what happens when you stack choices on top of each other to balance or cleanse out issues that already exist in societies or niche-communities that have already been disrupted. It's still miles away from Black Swan event forecasting that theoretical Economists use for predicting why changes happen to societies based on trigger events but its a start in the right direction. I can see this really benefiting politicians only if the system is closed and isn't really used by any entity to shift societal distress based on a model that is the current society that people live in. 

The bigger issue also lies in the fact that the entities in the system will probably never really behave like people do in the real world simply because they are singular entities with a goal-set that drives towards outcomes based on how the environment treats them. As opposed to real people who can make drastic decisions based on extraneous factors that even AI can't completely comprehend and bleed into its own logic due to logic-failure and system shutdown. I can't remember where I read this but there was talk about systems designed for failure instead of success and how they were trying to bring the two together to build autonomous systems which can use these two extremities to mimic real societies but once again far far away.

I wonder whether this system would have predicted that Trump would be president? Probably would have considering it had very little to work with.

Great read as always @nudge

[Mel81x+]

One day they'll have these things fight themselves so people can watch and enjoy the madness that ensues. What's very cool about this tech is that months ago (I am thinking 10 months) this thing couldn't even play DOTA much less make the kinds of ridiculous decisions its making now. If you have ever played DOTA watch the strategy this machine uses and the relentless pursuit it takes to fuck over its human opponents. I can't find a link to the original video suffice to say this thing didn't come out of the gate winning games like it is now, it learned that making trades on resource management versus ganking (ganging up on enemies) and spot hitting NPCs was the right thing to do, then it got smarter and decided to hit very early to see what would happen. When that failed after a few tries it then discovered the art of building up strategy points by amassing weapons on tracks and then it went to the amateur circuit to play against lower level DOTA players till it got to the video above. Someone rightly said it when this thing was showcased, "there is one thing this machine can do that a human can't. It just won't stop learning and its execution is flawless with whatever it learns".

[Mel81x+]

https://www.theguardian.com/world/2018/nov/09/worlds-first-ai-news-anchor-unveiled-in-china

[nudge]

3 hours ago, Mel81x said:

https://www.theguardian.com/world/2018/nov/09/worlds-first-ai-news-anchor-unveiled-in-china

I wonder if they have a Mandarin instead of English speaking one? I reckon it would look and sound even more believable. Overall, it does look a bit creepy haha. The tone, pronunciation and lip sync seems a bit off sometimes still, but I was mostly surprised how they got te blinking right!

[CaaC (John)+]

Just saw this thread and all I could think of was one of my favorite characters in my favorite show, Star Trek and Data, don't get me wrong as I do think sooner than later we will end up with Androids just like Data walking around planet earth and mixing in with the human race as just a normal thing.

'Greatest minds in physics playing poker' this episode is called and Data created this in the 'Holodeck''

Holodeck

The holodeck is a fictional plot device from the television series Star Trek. It is presented as a staging environment in which participants may engage with different virtual reality environments. From a storytelling point of view, it permits the introduction of a greater variety of locations and characters that might not otherwise be possible, such as events and persons in the Earth's past, and is often used as a way to pose philosophical questions.

Although the Holodeck has an advantage of being a safer alternative to reality, many Star Trek shows often feature holodeck-gone-bad plot devices in which real-world dangers (like death) become part of what is otherwise a fantasy.

https://en.wikipedia.org/wiki/Holodeck

 

 

[Mel81x+]

2 hours ago, nudge said:

I wonder if they have a Mandarin instead of English speaking one? I reckon it would look and sound even more believable. Overall, it does look a bit creepy haha. The tone, pronunciation and lip sync seems a bit off sometimes still, but I was mostly surprised how they got te blinking right!

I'd imagine digital intonation wouldn't allow for something to be created that way. It's quite hard to make those kinds of language dialect trees vocalize themselves when you've got spoken language. And of all the languages in the world I'd imagine Mandarin being the hardest because to the way context is actually calculated. Still, it is quite a feat and they've now started to use something which I think has good applications for news networks too. The real question is whats next?

 

[Mel81x+]

2 hours ago, CaaC - John said:

Just saw this thread and all I could think of was one of my favorite characters in my favorite show, Star Trek and Data, don't get me wrong as I do think sooner than later we will end up with Androids just like Data walking around planet earth and mixing in with the human race as just a normal thing.

'Greatest minds in physics playing poker' this episode is called and Data created this in the 'Holodeck''

I wouldn't say sooner rather than later but if you're interested in seeing how robotics has caught up recently take a look at these folks and some of the crazy things they are making robots do.

https://www.bostondynamics.com/