AI is NOT READY for WARFARE!!!

  • Christian Chat is a moderated online Christian community allowing Christians around the world to fellowship with each other in real time chat via webcam, voice, and text, with the Christian Chat app. You can also start or participate in a Bible-based discussion here in the Christian Chat Forums, where members can also share with each other their own videos, pictures, or favorite Christian music.

    If you are a Christian and need encouragement and fellowship, we're here for you! If you are not a Christian but interested in knowing more about Jesus our Lord, you're also welcome! Want to know what the Bible says, and how you can apply it to your life? Join us!

    To make new Christian friends now around the world, click here to join Christian Chat.
Solemateleft

Solemateleft

Honor, Courage, Commitment
#1
Jun 25, 2017
14,095
4,108
113
#1
Since this falls under one of my fields of study - looking forward to going back to work and sharing amongst my colleagues - a big This is what I have been trying to tell you all along...
AI limitations are all about the DATA!
The AI Machine Learning capabilities that we are beginning to see in areas rich with DATA - are just NOT applicable or ready for use within the complexities of Modern Day Warfare - as we know it today...
Where the military can use AI is similar domains as Industry - WARFARE is JUST TOO COMPLEX - the DATA are never just black and white and the very nature of modern day warfare is being fought in the GREY AREA within the Irregular and Information and Cyber Domains...
US Marines Defeat DARPA Robot by Hiding Under a Cardboard Box
  • By Adrianna Nine on January 19, 2023 at 3:16 pm
  • 1674345327448.png
  • View attachment 247832
  • The Pentagon’s Defense Advanced Research Projects Agency (DARPA) has invested some of its resources into a robot that’s been trained—likely among other things—to identify humans. There’s just one little problem: The robot is cartoonishly easy to confuse.
  • DARPA calibrated its robot’s human recognition algorithm alongside a group of US Marines. The Marines and a team of DARPA engineers spent six days walking around the robot, training it to identify the moving human form. On the seventh day, the engineers placed the robot at the center of a traffic circle and devised a little game: The Marines had to approach the robot from a distance and touch the robot without being detected.
  • DARPA was quickly humbled. Scharre writes that all eight Marines were able to defeat the robot using techniques that could have come straight out of a Looney Tunes episode. Two of the Marines somersaulted toward the center of the traffic circle, thus using a form of movement the robot hadn’t been trained to identify. Another pair shuffled toward the robot under a cardboard box. One Marine even stripped a nearby fir tree and was able to reach the robot by walking “like a fir tree” (the meaning of which Twitter users are still working to figure out).

    While it’s funny to imagine a team of Marines using Metal Gear Solid’s cardboard box strategy to defeat what’s likely a very expensive robot, the incident detailed in Scharre’s book fortifies something we already know: AI is only as useful as the data we give it. Similar to the way AI becomes biased once it’s fed biased data, algorithms can be as ignorant as their foundational data is flat. Without being shown what a somersaulting human or a human under a box looks like in action, a robot won’t be able to discern that image from all the surrounding noise, no matter how skilled its engineers are.
 
Lynx

Lynx

Folksy yet erudite
#3
Aug 13, 2014
27,779
9,697
113
#3
So Solid Snake was right all along... And all those guards were robots.

Now it all becomes clear.

No wonder the guards had such predictable patrol routes and never deviated.
 
Billyd

Billyd

Senior Member
#4
May 8, 2014
5,249
1,669
113
#4
Unfortunately for those boxed Marines, the AI gained in knowledge from the exercise, and will be ready for similar attacks in the future. I wouldn't use the same tactic again.

In the early 1980s, I worked on a project that used AI to troubleshoot a very complex system. As the project progressed, the AI program became more efficient every time it troubleshot the system. Three years into to project, the AI knowledge base was so extensive that it was finding 95% pf the problems on the first run and almost all by the third run.

It's been almost forty years since those AI days. The AI engines and knowledge bases have progressed in complexity and efficiency.
 
Solemateleft

Solemateleft

Honor, Courage, Commitment
#5
Jun 25, 2017
14,095
4,108
113
#5
Billyd said:
Unfortunately for those boxed Marines, the AI gained in knowledge from the exercise, and will be ready for similar attacks in the future. I wouldn't use the same tactic again.

In the early 1980s, I worked on a project that used AI to troubleshoot a very complex system. As the project progressed, the AI program became more efficient every time it troubleshot the system. Three years into to project, the AI knowledge base was so extensive that it was finding 95% pf the problems on the first run and almost all by the third run.

It's been almost forty years since those AI days. The AI engines and knowledge bases have progressed in complexity and efficiency.
Click to expand...
No Doubt that their have been lots of advances... again in areas where the Data is Rich and Trustworthy... Unfortunately, only a small fraction of such data when it comes to the application of actual Military Warfare are Rich and Trustworthy ... as the Nature of War essentially continues to add complexities and dimensions that are simply significantly different than the areas that Industry are making amazing advancement with such technology...

I attended the Military Operations Research Symposium (MORS) Special Session a few years ago where the focus was on AI and it's potential uses in the military... Oh sure, DoD is investing a lot of money in this area of research (I have been involved from a study and analysis and Data Assessment perspective) and lets just say the very nature of experimentation means there are lots of trials and error and failed experiments... The hosts were professional associates and friends of mine, and I was asked to participate on the Synthesis Group... Essentially, we were the ones who were responsible for generating the report that came out of this DoD AI Special Session...
The event Host brought in the Pros from Dover - EXPERTS - on the topic of Motor Vehicle AI & Autonomy... And these Experts were here to tell the Military Operations Research Society that while we are making huge advancement in this field with motor vehicles, at amusement parks, airports, and Logistics - these applications all include some form or fashion of HUMAN in the LOOP as a risk mitigation feature... These Experts cautioned the Military that such AI in the Department of Motor Vehicles would require increased attention to Testing and Evaluation and Verification to eliminate MISHAPS... as they have experienced lots of SIMILAR incidents that the AI has just NOT be TRAINED to RECOGNIZE... These Experts projected that it would be at least another 50-100 years before fully Autonomous Vehicles (without Human in the Loop) would be operating openly within our Road Ways - without incident or risk of life...
Now, comparing the DMV AI requirements for ACCURATE AI RECOGNITION to applications in WARFARE is like comparing Apples to Jack Hammers... as in Warfare unlike DMV AI - there is always an Enemy deliberately attempting to deceive its opponents.

While this single DARPA Experiment merely suggested a few additional means that a small few number of Marines were able to come up with... Now imagine all the Outfits seen on the Masked Singer, every Halloween Costume every imagined and every Jim Carry or Robbin Williams comedic walk etc etc... and that someone is now going to get paid to image new and novel things that their enemy Robots have not seen before...

While there are useful and productive applications that are being gained, there is a long way before the Military Data is both Rich and Trustworthy...
 
HealthAndHappiness

HealthAndHappiness

Well-known member
#6
Jul 7, 2022
10,623
4,549
113
Almost Heaven West Virginia
#6
Solemateleft said:
No Doubt that their have been lots of advances... again in areas where the Data is Rich and Trustworthy... Unfortunately, only a small fraction of such data when it comes to the application of actual Military Warfare are Rich and Trustworthy ... as the Nature of War essentially continues to add complexities and dimensions that are simply significantly different than the areas that Industry are making amazing advancement with such technology...

I attended the Military Operations Research Symposium (MORS) Special Session a few years ago where the focus was on AI and it's potential uses in the military... Oh sure, DoD is investing a lot of money in this area of research (I have been involved from a study and analysis and Data Assessment perspective) and lets just say the very nature of experimentation means there are lots of trials and error and failed experiments... The hosts were professional associates and friends of mine, and I was asked to participate on the Synthesis Group... Essentially, we were the ones who were responsible for generating the report that came out of this DoD AI Special Session...
The event Host brought in the Pros from Dover - EXPERTS - on the topic of Motor Vehicle AI & Autonomy... And these Experts were here to tell the Military Operations Research Society that while we are making huge advancement in this field with motor vehicles, at amusement parks, airports, and Logistics - these applications all include some form or fashion of HUMAN in the LOOP as a risk mitigation feature... These Experts cautioned the Military that such AI in the Department of Motor Vehicles would require increased attention to Testing and Evaluation and Verification to eliminate MISHAPS... as they have experienced lots of SIMILAR incidents that the AI has just NOT be TRAINED to RECOGNIZE... These Experts projected that it would be at least another 50-100 years before fully Autonomous Vehicles (without Human in the Loop) would be operating openly within our Road Ways - without incident or risk of life...
Now, comparing the DMV AI requirements for ACCURATE AI RECOGNITION to applications in WARFARE is like comparing Apples to Jack Hammers... as in Warfare unlike DMV AI - there is always an Enemy deliberately attempting to deceive its opponents.

While this single DARPA Experiment merely suggested a few additional means that a small few number of Marines were able to come up with... Now imagine all the Outfits seen on the Masked Singer, every Halloween Costume every imagined and every Jim Carry or Robbin Williams comedic walk etc etc... and that someone is now going to get paid to image new and novel things that their enemy Robots have not seen before...

While there are useful and productive applications that are being gained, there is a long way before the Military Data is both Rich and Trustworthy...
Click to expand...
Do you mean, like these car driven by killer robots?


Oops! What was the title of this video again?
Oh.
Never mind. 😁


That was my bad joke for the weekend.
 
You must log in or register to reply here.