nav[aria-label="Primary Navigation"] { padding: 0; & ul { list-style: none; width: 100%; display: flex; flex-direction: row; justify-content: start; align-items: start; gap: 30px; padding: 0; & li { margin: 0; } & ul li { list-style: none; } } }

Child Item

Killer Robots

Fathom

Drones occupy battlefields today. Airborne. Strike targets. Automated warfare exists today. This technology said to be an expanding "continuum." AI weapons development replacing direct human involvement. Should limits be placed on such “lethal autonomous weapons?” Computer scientist Stewart Russell at University of California at Berkeley thinks so. But is it too late? Ask I-Robot? Terminator? Sci Fi leads the way? Ref: http://www.popsci.com/big-idea-killer-robots-are-coming

Find more posts tagged with

Quick Links

Comments

Fathom

Capt'n Midnight wrote: »

Debugging : The act of replacing known bugs with unknown bugs.

Cool expression!

Rubecula

surely that would be rebuggibg??debugging would be to secretly add the bugs

Capt'n Midnight

Rubecula wrote: »

surely that would be rebuggibg??debugging would be to secretly add the bugs

Once upon a time Grace Hooper removed a moth* from a relay in an electromechanical computer.

Bugs were known about long before then but it's a good tale.

Also it doesn't answer the question of what if there's a hardware fault or "unexpected item in fragging area ?"

A human has some chance of coping with the unexpected. An AI has really none unless it's been programmed in.

The history of the Nuclear industry shows that spending tens of billions still means the are failures in design and implementation. And we've been building nuclear reactors for 70 years.

Microsoft is a trillion dollar company. URL="https://en.wikipedia.org/wiki/ELIZA"]Chatbot AI's have been around since 1966.[/URL]

50 years later Microsoft's latest AI chatbot was released on twitter.
They pulled it after just 16 hours. Because it couldn't handle the real world.

And then accidentally re-released it later on. Which doesn't set a good precedent.

AI is a Crapshoot.

* It was dead, a ghost in the machine ?

Rubecula

Capt'n Midnight wrote: »

Once upon a time Grace Hooper removed a moth* from a relay in an electromechanical computer.

Bugs were known about long before then but it's a good tale.

Also it doesn't answer the question of what if there's a hardware fault or "unexpected item in fragging area ?"

A human has some chance of coping with the unexpected. An AI has really none unless it's been programmed in.

The history of the Nuclear industry shows that spending tens of billions still means the are failures in design and implementation. And we've been building nuclear reactors for 70 years.

Microsoft is a trillion dollar company. URL="https://en.wikipedia.org/wiki/ELIZA"]Chatbot AI's have been around since 1966.[/URL]

50 years later Microsoft's latest AI chatbot was released on twitter.
They pulled it after just 16 hours. Because it couldn't handle the real world.

And then accidentally re-released it later on. Which doesn't set a good precedent.

AI is a Crapshoot.

* It was dead, a ghost in the machine ?

hardware would be the easiest to mess up if you wanted to I suspect an armed robot would if told to do it keep on firing a weapon even if it had no ammo or anything like that that would be a huge hardware task that went wrong.

[Deleted User]

Capt'n Midnight wrote: »

AI is a Crapshoot.

Can we speak comparatively? How does AI compare to random variation, differential reproduction, and evolution? Similarities? Differences? AI now in its single cell stage today, and a primitive cell at that? Will cells combine as they evolve, with many early variations that are not fecund, but an extremely rare variation that becomes a fecund hybrid? Or perhaps these comments are just metaphorical nonsense?

Fathom

From dual to multiple competing algorithms? Survival of the killer algorithm. Herbert Spencer updated?

Capt'n Midnight

Black Swan wrote: »

Can we speak comparatively? How does AI compare to random variation, differential reproduction, and evolution? Similarities? Differences? AI now in its single cell stage today, and a primitive cell at that? Will cells combine as they evolve, with many early variations that are not fecund, but an extremely rare variation that becomes a fecund hybrid? Or perhaps these comments are just metaphorical nonsense?

No comparison, living organisms have had 4 billion years to learn the basics. And not just a handful , it's planet wide.

AI is still in it's infancy. Today's chat bot's have access to computers that the 1960's researchers could only have dreamed of. They are better than Eliza but not hugely better, it's like they have a better dictionary and a few more rules on grammar and not much else to show for 50 years of software and hardware development.

[Deleted User]

Fathom wrote: »

From dual to multiple competing algorithms? Survival of the killer algorithm. Herbert Spencer updated?

Antagonistic learning between 2 algorithms occurs today. Not sure if more than two would result in greater efficiency and effectiveness.

Capt'n Midnight

Black Swan wrote: »

Antagonistic learning between 2 algorithms occurs today. Not sure if more than two would result in greater efficiency and effectiveness.

But how is it measured ?

You'd think that nuclear weapons would have had lots of safety features from day one and the rules and controls would be idiot and accident proof.

You'd be very wrong. There's a huge litany of near misses.

between the years 1950 and 1968, there were more than 1,000 accidents involving nuclear weapons.

Anyone who thinks this won't be repeated elsewhere only needs to look at Boeing 737 software.

Capt'n Midnight

[Deleted User]

Capt'n Midnight wrote: »

Anyone who thinks this won't be repeated elsewhere only needs to look at Boeing 737 software.

The Boeing 737 nods its nose up-and-down in agreement with you Capt'n.

Fathom

Black Swan wrote: »

The Boeing 737 nods its nose up-and-down in agreement with you Capt'n.

Nodding for sure.

Fathom

What's the chance that tomorrow's killer robots may exhibit programming errors like those experienced by Boeing today. Will future killer robots be grounded after they run amuck? Reminded of Runaway (1984) starring Tom Selleck, whose policeman character is an expert in handling rogue machines.

Capt'n Midnight

Fathom wrote: »

What's the chance that tomorrow's killer robots may exhibit programming errors like those experienced by Boeing today. Will future killer robots be grounded after they run amuck? Reminded of Runaway (1984) starring Tom Selleck, whose policeman character is an expert in handling rogue machines.

100%. It's just a matter of time given the almost complete lack of control.

NASA spend $1,000 per line of code. And purposefully use the most boring programmers possible. So it's bland predictable code that's supposed to work every time. And they still screw up.

AI and robots are sexy so it's the opposite type of code.

SlowBlowin

Black Swan wrote: »

Can we speak comparatively? How does AI compare to random variation, differential reproduction, and evolution? Similarities? Differences? AI now in its single cell stage today, and a primitive cell at that? Will cells combine as they evolve, with many early variations that are not fecund, but an extremely rare variation that becomes a fecund hybrid? Or perhaps these comments are just metaphorical nonsense?

My AI ramble/rant

AI is a buzz word, always in the press, and companies run to commercialise it, even when alternative, established methods exist. A lot of what is termed AI is simple pattern matching, yet we use the word intelligent to describe it.

The principles and techniques have been around for a long time. The model we have of the neuron seems to work well, and is established. The methods we use to train networks of these neurons has settled (genetic algorithms and back propagation). The real progress (besides finally having the CPU power) has been in structuring the networks to give them the opportunity to excel, by creating software that uses the networks to create their own data. Deepmind's alpha zero is a fantastic example of this, a great bit of software surrounding a bunch of standard neuron models, that is able to master any two player game (with perfect information). So whereas we have a good model of the neuron, and how networks of them can be structured to excel at different types of problems, what we dont have is an understanding of how the brain manages networks and builds the "software" around them.

Most of my recent experience has been with genetic evolution simulations and I can honestly say it has helped me (real world) understand peoples behaviour better. It has also helped me understand my wife's, post chemo, symptoms. A lot of the best work in AI has come from studying and trying to understand the biology of the brain, with Biology knowledge as important as computer coding experience.

I often see things in simulations that apply to the real world, traits, afflictions, skills, etc. Always the most relevant results come when natures own rules are modelled as closely as possible.

As for AI killer robots, I do think they will be a major factor in a war soon.

[Deleted User]

SlowBlowin wrote: »

As for AI killer robots, I do think they will be a major factor in a war soon.

Thus our discussion. Enjoyed your post.

SlowBlowin

Something else that is often overlooked is that at the moment we implement neural networks on a digital system, computers. Real performance will start to happen when we have mastered the analogue side. All stages/operations in a neural network have an analogue equivalent, which can be implemented in a simple analogue circuit using electronic components. When I first studied higher education we only had limited access to a terminal on the single college computer (early 80s), but there were many analogue computers. These analogue computers were widely used in industry for engineering modelling, and as there was no "clock" they were instant.

It dawned on me some years ago that when you find a neural network solution, building an analogue equivalent allows it to operate instantly. I looked online and discovered that this is an old idea, and already implemented.

A state of the art missile has a digital guidance system which gets it to the target, but in the last few moments (seconds), guidance is handed over to an simple analogue computer/circuit that is able to make changes to course without having to wait and calculate it. Such circuits are high tolerance analogue implementations of a simple neural net developed on a digital platform.

At the moment there is huge investment making what we have faster and more parallel, and students are steered towards code libraries (Tensor Flow etc) from the big IT players. There will be a magnitude leap in AI when hardware manufactures produce analogue neurons on chip, and this is now happening...

EDIT: Sorry the reason this is relevant for killer robots and advanced weaponry instant compute speeds are required for many situations (fire and forget bullets for example)

Fathom

SlowBlowin wrote: »

EDIT: Sorry the reason this is relevant for killer robots and advanced weaponry instant compute speeds are required for many situations (fire and forget bullets for example)

"The need for speed!"

Capt'n Midnight

SlowBlowin wrote: »

It dawned on me some years ago that when you find a neural network solution, building an analogue equivalent allows it to operate instantly. I looked online and discovered that this is an old idea, and already implemented.

A state of the art missile has a digital guidance system which gets it to the target, but in the last few moments (seconds), guidance is handed over to an simple analogue computer/circuit that is able to make changes to course without having to wait and calculate it. Such circuits are high tolerance analogue implementations of a simple neural net developed on a digital platform.

Seconds are an eternity when a typical multi core smartphone CPU can make billions of calculations. Compared to the cost of weapons they are disposable.

Pipelining means that even moderately complex instructions can be done at one per core per cycle.

Control surfaces or motor gimbals or control thrusters take time to react too. Inertial sucks.

You can now buy 1 TB micro SD cards. Many computations can be looked up from tables pre-calculated to save time and decision making.

( the logical extreme here was the IBM 1620 which was known as the CADET because it used lookup tables for addition ( 1+1 is "see what's stored in row 1 column 1" ) Can't Add, Doesn't Even Try )

Mass produced military tech lags behind consumer tech where both overlap.

Yet another Tesla crashed on "autopilot".
Note that AFAIK when they say driver wasn't in control it doesn't mean they didn't have their hands on the wheel, just that there were no control inputs.

SlowBlowin

Cores in phones do not cut it, sorry.

All the main smartphone players are now incorporating NPU (dedicated matrix maths ICs for neural nets) in order to offer the latest AI features. These are still "clocked" but far more efficient.

EDIT: Any who would want a killer robot controller by a Samsung Galaxy

Fathom

Capt'n Midnight wrote: »

Mass produced military tech lags behind consumer tech where both overlap.

Old tech inventories linger on. B-52's still used. Boeing first developed in 1950's. Recently sent to Iran arena by self-proclaimed genius Donald Trump.

Capt'n Midnight

Fathom wrote: »

Old tech inventories linger on. B-52's still used. Boeing first developed in 1950's. Recently sent to Iran arena by self-proclaimed genius Donald Trump.

The Russian Tu-95 can trace it's lineage to reverse engineered WWII B-29's
It's got huge propellers. It's so loud that it can be picked up underwater microphones.

Like the B-52 and the Chinese Xian H-6 it wouldn't last two minutes against modern fighter aircraft.

They don't have to. They simply fly to the middle of a nearby ocean, far from prying eyes and lunch radar sneaky or stealthy cruise missiles from 1,000km or 2,000km away.

A software update to give the missile some 'AI' keeps the aircraft that were technically obsolete in the 1960's relevant.

In other news
US Navy wants 350 billion social media posts I smell AI training.

Fathom

Capt'n Midnight wrote: »

A software update to give the missile some 'AI' keeps the aircraft that were technically obsolete in the 1960's relevant.

Comparatively, when I get old. Will AI be there for me? Maybe if I become a killer robot the military will keep me going? Cyborg me?

[Deleted User]

Fathom wrote: »

Comparatively, when I get old. Will AI be there for me? Maybe if I become a killer robot the military will keep me going? Cyborg me?

Military and universities are always looking for grand killer robot grants and research. Not sure they are looking for the emic perspective, rather etic.

Capt'n Midnight

Rubecula

Just for a moment let us think about these killer bits with advanced AI running riot and taking over the world... who is to say we will not get help to survive from outside our solar system?
How do you think our ew found alien benefactors would go down here amongst the general populace?

Fathom

Just watched a Terminator movie rerun.

Capt'n Midnight

And so it begins ...

Fathom

Love or killer robots? Puts a whole new spin on May West: "Is that a pistol in your pocket, or are you glad to see me?"

[Deleted User]

Role confusion? AI therapy robot for robots?