ARRSE is supported by the advertisements on it, so if you use an adblocker please consider helping us by starting an Ad-Free subscription.

Artificial Intelligence and Warfare

Discussion in 'Current Affairs, News and Analysis' started by Wordsmith, Aug 15, 2017.

Welcome to the Army Rumour Service, ARRSE

The UK's largest and busiest UNofficial military website.

The heart of the site is the forum area, including:

  1. Wordsmith

    Wordsmith LE Book Reviewer

    Interesting paper forecasting AI is going to be a game changer in warfare - and not too far in the future either.

    http://www.belfercenter.org/sites/default/files/files/publication/AI%20NatSec%20-%20final.pdf
    So:
    1. Will AI affect warfare?
    2. If it does, how soon?
    Wordsmith
     
    • Excellent Topic Excellent Topic x 1
  2. Another risk is that as AI matures, the risk of something becoming self aware increases

    Killer robots aren't far off now, MAARS was over a decade ago, add in AI and the potential to go full Terminator is there

    Pursue ever increasing AI, and what are the chances we lose control of the technology ?
     
    • Funny Funny x 1
  3. Of course, IF you can direct 60-70 UK million minds to AI, good; unfortunately, the Chinese can direct how many minds to AI? And they will. Unless, Trump decides to end the planet now, the Chinese will take control in the not too distant future.
     
    • Like Like x 1
  4. This paragraph stood out for me:


    "Also like the first industrial revolution, population size will
    become less important for national power. Small countries
    that develop a significant edge in AI technology will punch far
    above their weight."


    And yet the UK, and Europe generally, continue to breed and import more and more poorly educated proletarians. These people and their progeny will be a liability, not an asset, going forward.

    A lot of people now saying that Japan, far from being in decline and stagnation, has actually got it about right in focussing on robotics and technology rather than increasing its population at the expense of social cohesion.
     
    Last edited: Aug 15, 2017
    • Like Like x 5
  5. Wordsmith

    Wordsmith LE Book Reviewer

    It's not only the number of minds - its the level of technical skill. The pool of people with cutting edge AI skills is (I suspect) fairly restricted in most countries. China might have a lot of bodies, but many are poorly educated rural people who might never have seen a computer.

    The US probably has a larger pool of AI literate nerds.

    Wordsmith
     
  6. Interesting article. However I agree with Professor Hawking that it is a dangerous path to follow. Recently the two "chat-bots" that began communicating in unbreakable code is food for thought.
     
    • Like Like x 1
  7. Sarastro

    Sarastro LE Reviewer Book Reviewer

    May I summarise the paper?

    "We don't know a huge deal about AI , so we don't know if AI is going to be a game changer. We did talk to a number of other people who also don't know a huge deal about AI, and they also did not know if AI was going to be a game changer. However, if AI does change the game, then we all confidently predict that the game will change.

    Here are 100 pages about how games change:"

    Public discussion of this whole field generally falls into four categories:

    This is what AI is being used for today and next week / month (5%)
    Our platform will use AI to solve XXXX (60%)
    AI will definitely completely change XXXX in the near future (25%)
    OMG SKYNET! (10%)

    Only the first one has any basis in reality. The rest tends to be pure speculation and hype, and is much more usually pushed by people who actually have nfc about AI than the researchers who do.
     
    • Like Like x 2
    • Funny Funny x 1
  8. Wordsmith

    Wordsmith LE Book Reviewer

    I think AI will eventually come. Look at Intrusion Prevention Systems.

    (Or as I prefer to call them - Intrusion Ignoring Systems. I've seem test malware uploaded to IPS to validate them. Despite the malware waving a flag and saying 'here I am', it took the system 4 days to flag there was a problem).

    But eventually IPS will improve, probably based on some form of AI. Then it will be a more effective form of cyber defence as it'll be able to adapt to a variety of threats. How long? My guess is we'll see quite effective interactive IPS inside 10 years.

    Wordsmith
     
  9. We use AI/Machine Learning to build profiles of our "Learners" so that we can customise their training depending on their performance, roles and even gender. It's not new..Amazon and Google have been doing it for years.

    We have complex algorithms that predict a learners next move and so changes the presented information to something targeted rather than generic.

    Like any AI or ML system, it needs data to "learn". For AI to be used in cyber security for example, it needs to be attacked several times before it realises it's been attacked. It learns, and it prevents the next similar attack...and so it goes on. These systems always need a shed load of data before being switched on , else it is like asking a 5 year old to cook dinner...they have never been shown, so won't have a hope in hell of doing it.
     
    • Like Like x 1
  10. LmH

    LmH Clanker

    Machine 'learning' is based on retrieval and patterns.

    Becoming aware and taking over would be like your phone or computer doing the same thing. But systems can and will appear seamless in the future.
     
  11. Have you not seen The Terminator
     
    Last edited: Aug 17, 2017
    • Like Like x 1
  12. Wordsmith

    Wordsmith LE Book Reviewer

    An AI cyber security program would need to be proactive - to realise that the pattern is seeing is a very slow cyber attack. If an hacker is exploiting a real life SQL injection vulnerability, they may only be executing an single enquiry every few seconds. That would get lost against the background of tens of thousands of legitimate queries per second. Such a cyber security program wouldn't have to wait to be attacked to realise something odd was going on, it would have to understand that certain types of activity - even if executed very slowly/quietly - were probable indicators of an ongoing cyber attack.

    I could see HMG creating a central 'experience bank' for UK based AI cyber defence programs that all such programs wrote back to and updated themselves from. That way you'd eventually get a huge database of 'experience' and ever more of a learning potential.

    Wordsmith
     
    • Informative Informative x 2
    • Like Like x 1
  13. A2_Matelot

    A2_Matelot LE Book Reviewer


    There is another ARRSE thread on this, but to answer your questions, it's already starting to:

    - AI is being used in medical applications, in intelligence analysis including imagery analysis and signals analysis - all to a small extent, but it's increasing.

    - We're starting to look at developing AI to build into the command and control chain, sensors to shooters, to speed up reaction times. For instance as hypersonic weapons are being developed and fielded, the reaction times for detection and counter-measure are hugely reducing, possibly far below anything a human can react to, hence the application of AI.

    Equally in high threat complex environments, such as air warfare, AI is already in progress to monitor air tracks and identify hostile targets far further out and far more quickly than humans.

    We've had AI in use for years in a number of weapons and sensor systems (fuzzy logic, expert systems) but it's now that massive computing power and meshed networks are available, you can exploit advances especially in machine learning.

    This is one of the reasons I liked Gen Barrons, he was exposed to these technologies and very quickly saw the potential for both adversary and friendly application and knew we needed to make progress, as opposed to just think about tanks, ships and aircraft.
     
    • Informative Informative x 1
  14. A2_Matelot

    A2_Matelot LE Book Reviewer

    There are a lot of cyber vendors who now use AI extensively to look at signatures on networks and to analyse and detect anomalous behaviour in a way more superior to IDS/IPS. Quite niche and expensive, but look very promising..