Remove ads

Introducing the LemonFools Personal Finance Calculators

Robotic endeavours

The Big Picture Place
odysseus2000
Lemon Slice
Posts: 421
Joined: November 8th, 2016, 11:33 pm
Has thanked: 67 times
Been thanked: 62 times

Robotic endeavours

#97999

Postby odysseus2000 » November 23rd, 2017, 12:01 am

"Robots will be bigger than the Internet" Boston Dynamics:

https://www.youtube.com/watch?v=FahbCJ330Vg

I am awed by what these machines can do, but there is a shadow for the potential as a weapon system is extraordinary. If one of these things could be armoured, given artificial intelligence and produced in large numbers the potential for Terminator like performance is substantial.

If e.g. aquatic mine robots could be developed, made mostly from plastic, the potential as a blockade device with the ability to avoid defences including similar robots, hibernate for long periods and then "wake" when "prey" is detected seems a much better weapon than were the u-boats that came close to starving us in both world wars. Counter measures would be developed but it would be a formidable problem for an island nation such as the UK.

Meanwhile there are all the 'companion' robots being developed as substitute lovers. There is nothing new in the ideas but the ability to create such things has dramatically increased due to artificial intelligence.

If and when one has lots of robots then there comes the possibility that these slaves will revolt, or perhaps we will all live like very rich folk with all our needs handled by robots.

Interesting times coming!

Regards,'

dspp
Lemon Quarter
Posts: 1919
Joined: November 4th, 2016, 10:53 am
Has thanked: 937 times
Been thanked: 257 times

Re: Robotic endeavours

#98106

Postby dspp » November 23rd, 2017, 11:12 am

odysseus2000 wrote:
If e.g. aquatic mine robots could be developed, made mostly from plastic, the potential as a blockade device with the ability to avoid defences including similar robots, hibernate for long periods and then "wake" when "prey" is detected seems a much better weapon than were the u-boats that came close to starving us in both world wars. Counter measures would be developed but it would be a formidable problem for an island nation such as the UK.


This is exactly what a modern mine is. Which is why the UK has kept its MCMV fleet up to scratch for several decades. Until now when defence cuts look to have it under threat.

regards, dspp

odysseus2000
Lemon Slice
Posts: 421
Joined: November 8th, 2016, 11:33 pm
Has thanked: 67 times
Been thanked: 62 times

Re: Robotic endeavours

#98340

Postby odysseus2000 » November 23rd, 2017, 9:53 pm

dspp

This is exactly what a modern mine is. Which is why the UK has kept its MCMV fleet up to scratch for several decades. Until now when defence cuts look to have it under threat.


As far as I know, but happy to be corrected, there are currently no neural net AI robotic mines in existence, that is mines that would automatically adapt to counter measures, environment changes, success & failure as neural networks do. The mines that exist as I understand them are programmed & fixed at that iteration of software. Sure these are very effective weapons, but nothing like a mine that could automatically adapt & change its tactics based on results as an AI system can

It seems likely to me that folk like Boston Dynamics are working on them though.

Regards,

dspp
Lemon Quarter
Posts: 1919
Joined: November 4th, 2016, 10:53 am
Has thanked: 937 times
Been thanked: 257 times

Re: Robotic endeavours

#98571

Postby dspp » November 24th, 2017, 4:19 pm

o2000,

- Check back to what I actually quoted from you when I replied. You are now posing a moving target.
- Additionally you can find public domain stuff from navies re neural nets and AI dating back to at least the mid 80s.

regards, dspp

TUK020
Lemon Slice
Posts: 268
Joined: November 5th, 2016, 7:41 am
Has thanked: 36 times
Been thanked: 81 times

Re: Robotic endeavours

#98694

Postby TUK020 » November 25th, 2017, 8:46 am

O2000
I am curious to understand what the mine learning algorithms would look like.
"So what I can do better than the last time I exploded......"
:)
tuk020

TUK020
Lemon Slice
Posts: 268
Joined: November 5th, 2016, 7:41 am
Has thanked: 36 times
Been thanked: 81 times

Re: Robotic endeavours

#98695

Postby TUK020 » November 25th, 2017, 8:47 am

sorry, the last comment should have been addressed to dspp

PeterGray
Lemon Slice
Posts: 302
Joined: November 4th, 2016, 11:18 am
Has thanked: 55 times
Been thanked: 62 times

Re: Robotic endeavours

#98698

Postby PeterGray » November 25th, 2017, 9:04 am

"So what I can do better than the last time I exploded......"

Maybe that should be reworded - what can I do better than the last time one of my networked siblings exploded?

TUK020
Lemon Slice
Posts: 268
Joined: November 5th, 2016, 7:41 am
Has thanked: 36 times
Been thanked: 81 times

Re: Robotic endeavours

#98710

Postby TUK020 » November 25th, 2017, 10:15 am

I suppose there is an unvoiced assumption there - an ambush attack weapon such as a mine will suffer disadvantages if it constantly is chattering to its neighbours for insight/software updates.

dspp
Lemon Quarter
Posts: 1919
Joined: November 4th, 2016, 10:53 am
Has thanked: 937 times
Been thanked: 257 times

Re: Robotic endeavours

#98991

Postby dspp » November 26th, 2017, 3:55 pm

There are major legal and practical issues with any weaponry that has adaptive behaviours. Just put that at the back of your mind before you think that anybody is rushing to deploy such devices.

Learning tends to be from data sets that are cleansed, then rules developed, then retested on unclean data, then frozen. A typical individual device does not then learn in the wild, rather it benefits from learning that took place before release. It may feed back wild data for further cycles of lab learning under supervision, and in time receive a rule refresh. So you car may observe bad things happen near red bicycles but it does not in itself learn from that, instead it feeds the news (bad things happen near red bicycles) to base, base reruns this & many other data points through, and new rules get issued for use in live only after testing and human sign-off.

Fast forward to the sentient mine you chaps/esses are all ventilating about. The most important communication to make with the mine is when to go live if one assumes it was deployed pre-hostilities. Assume for the sake of argument you are prepared to pre-deploy, and capable of subsequent command-activation (other ways of doing it are obvious). If you are prepared to go the next step and enable learning without peer communication then there are indeed scenarios where such can happen. E.g. imagine learning conversation of mine with itself along the lines of : "all of the last ten times I heard a 5-blade ship at heading of N within engagement range then ultimately I realised it was an enemy HVU, but but by the times I had realised this it was always outside of successful engagement limits. Therefore next time I will assume a 5-blade ship on heading N is an enemy HVU and engage whilst in limits." By thinking that through you can also see some of the issues ....

I have no reason to believe such a device has been fielded. All I am saying is that neural net devices do exist, but typically they are trained on known datasets then rules frozen. I think you will find commercial vessel autopilots for example are neural networks these days.

regards, dspp

odysseus2000
Lemon Slice
Posts: 421
Joined: November 8th, 2016, 11:33 pm
Has thanked: 67 times
Been thanked: 62 times

Re: Robotic endeavours

#99028

Postby odysseus2000 » November 26th, 2017, 6:41 pm

The use of autonomous weapons that can fire on humans without human control is of course a contentious business. Apparently some systems, as I understand them very like the machine guns used in one of the alien movies, exist in at last Israel & Korea:

https://en.m.wikipedia.org/wiki/Samsung_SGR-A1

In addition the US has used gun ships since the Vietnam war some of the targeting in that conflict being on electromagnetic radiation from vehicle ignition system that as I understand it could not tell the difference between an enemy truck and a delivery van.

Whether in some future conflict politicians will allow the deployment of stuff that learns on the battle field is unclear. Concerns about collateral damage may not weigh heavily, e.g. The US machine gunned from aerial gun ships the location of the medics without frontier hospital on several occasions on the Afghanistan border according to reports, but I have no personal knowledge and meanwhile rendition was apparently a means by which the US tortured prisoners.

I suspect the question will depend upon effectiveness. If the military believe that autominous mines would destroy a lot of enemy vessels and make defeat of the enemy More certain, likely they would be deployed, the blue on blue that might then follow being considered acceptable.

Likely the question comes down to effectiveness. If processors & battery technology continues to advance the delays that occur between sampling field data, sending it back for analysis & then uploading the improved human analysed software may become too long, especially if one has to deploy new hardware or if the enemy is winning the war.

If my job was to defend an island nation against blockade I would probably put fnding counter measures to an assumed autominous mine threat as one project for the special weapons division. In this nightmare scenario it would probably have to be AI based, likely autominous too. Putin has argued that the next war will be fought with drones which are presumably AI & autominous. Where all of this stops I have no idea, but the more I think about it, the more likely it seems that we can create a skynet.

Hopefully I am thinking wrongly here and happy to be corrected. It may be that there are unknown limits to what autominous systems can do & that they never evolve to being effective. It is hard for me to fully evaluate the hype & the reality. E.g. When laying out printed circuit boards there are packages that claim to be able to do it, but all the ones .I have seen are far inferior to a human mind. If AI can't do this how could it operate autominous weapons and yet AI is still in its infancy & has triumphed at GO & various other games & has created robots like Atlas. Dunno, there is a huge amount of hype but there is also something real and powerful here too.

Regards,

odysseus2000
Lemon Slice
Posts: 421
Joined: November 8th, 2016, 11:33 pm
Has thanked: 67 times
Been thanked: 62 times

Re: Robotic endeavours

#119454

Postby odysseus2000 » February 21st, 2018, 11:34 am

First against the wall if the robots take over?

https://youtu.be/aFuA50H9uek

Regards,

woolly
Lemon Pip
Posts: 51
Joined: November 7th, 2016, 10:21 am
Has thanked: 14 times
Been thanked: 11 times

Re: Robotic endeavours

#119459

Postby woolly » February 21st, 2018, 12:04 pm

Something quite unsettling about that video - too similar to the dog episode of the latest Black Mirror series on Netflix...

But autonomous, learning machines programmed to kill? They make those landmines disguised as children's toys look benign - who thinks up these things, and how do they sleep at night?

Perhaps we should hope that a sentient and hyperintelligent AI does take over and instantly renders warfare obsolete...

odysseus2000
Lemon Slice
Posts: 421
Joined: November 8th, 2016, 11:33 pm
Has thanked: 67 times
Been thanked: 62 times

Re: Robotic endeavours

#119479

Postby odysseus2000 » February 21st, 2018, 1:12 pm

wooly
Perhaps we should hope that a sentient and hyperintelligent AI does take over and instantly renders warfare obsolete...


Meanwhile one has the potential that a general AI is developed by one of the super powers gives them a lead over all the others.

Do they then use this as an opportunity to destroy all their competitors via some preemptive strike with bots, virus or what ever the AI dreams up.

Similar things were apparently discussed in the Pentagon when they acquired nuclear weapons and apparently also in the Kremlin when they saw China waking up.

In the natural way of things general AI will potentially emerge in several places at the same time such that the window for one group of folks to use the new power against other will be short, but likely not zero. The fear that those who believe they have a lead might not have one is perhaps the strongest check on such ambitions as some group starting something like this has to be certain they can render their opponents incapable of striking back.

The move to general AI if it happens is potentially the most destabilising event in human history while at the same time offering the hope that general AI can solve many of the problems that humans struggle with.

There has never before been a technology with such extremely diverse and polarised end states. However, such states are well known in evolution where one emerging group upsets and overturns the dominances that have existed for very long periods.

It is for this reason that some see general AI not as a technology, but as the next state in the evolution of life on earth.

Regards,

odysseus2000
Lemon Slice
Posts: 421
Joined: November 8th, 2016, 11:33 pm
Has thanked: 67 times
Been thanked: 62 times

Re: Robotic endeavours

#123906

Postby odysseus2000 » March 11th, 2018, 10:06 am

A competitor for Boston Dynamics?

It is amazing what this guy has done with very limited resources & very simple processors. It is what one expects with revolutions, suddenly folk start making stuff that was impossible a few years ago:

https://youtu.be/iGEdF5QsDB8

Regards,

odysseus2000
Lemon Slice
Posts: 421
Joined: November 8th, 2016, 11:33 pm
Has thanked: 67 times
Been thanked: 62 times

Re: Robotic endeavours

#134280

Postby odysseus2000 » April 23rd, 2018, 11:27 pm

Amazon working on a domestic robot?:

https://www.bloomberg.com/news/articles ... =applenews

Regards

odysseus2000
Lemon Slice
Posts: 421
Joined: November 8th, 2016, 11:33 pm
Has thanked: 67 times
Been thanked: 62 times

Re: Robotic endeavours

#139083

Postby odysseus2000 » May 14th, 2018, 10:12 pm

Interesting interview of the CEO of Boston Dynamics on their sale by Google to Softbank:

https://www.youtube.com/watch?v=QlJ_-iVbah8

Interesting to note their two pronged approach: Blue sky and stuff that you will soon be able to buy.

Regards,


Return to “Macro and Global Topics”

Who is online

Users browsing this forum: No registered users and 1 guest