32 ms·

Let's be completely speculative and alarmist about the potential of AI: It could be like developing nuclear weapons first; if the US gets military AI wrong, it

Let's be completely speculative and alarmist about the potential of AI: It could be like developing nuclear weapons first; if the US gets military AI wrong, it could quickly become a poor vassal state to the world's new superpower.[0] People are alarmist and speculative because AI's potential is unknown. If the potential of the new blockchain technology is unknown then you can wait and see what happens. But given the stakes with military AI, you can't take even tiny risks; you can't wait and see if your country will be in history books as an experiment that lasted 250 years. By declining to help the US military, Google engineers take that risk to a degree. But if they participate then they gain enormous leverage: Given the stakes, the US military can't afford to alienate them. I'd hope they can use their leverage to achieve related goals: Agreements banning the use of this technology against civilians, foreign or domestic, and banning sharing the tech with law enforcement. Leverage Congress into passing privacy and civil rights laws protecting Americans against abuses of the technology. [0] Note that AI changes things in another way: For all human history, military power was tied to population size. In the future, with the right AI and some underground robot factories, potentially a small country could dominate. Maybe Singapore?

dmitrygr 9y ago
In the modern day, the difference between civilian and terrorist is so blurred (not only in factually but also in time - fon't forget, USA's founding fathers were terrorists, if you asked the UK monarchy) even humans often cannot tell. Do we trust AIs to?
- RcouF1uZ4gsC 9y ago
  > don't forget, USA's founding fathers were terrorists, This gets repeated all the time, but that doesn't make it true. When did the founding fathers deliberately do mass killings of civilians? They basically won by fighting pitched battles with the British. They had diplomatic relations with several countries (Denmark and France being the major ones). There was a command structure with clear lines of authority (Continental Congress -> George Washington -> other commanders). In addition, there was at least an attempt made to follow international laws. They were rebels, yes. However, it was more a coalition of existing colonial governments (especially the elected legislatures) deciding to part ways with the British. By and large, the terrorists of our times are not at all like that. They are not part of a legitimate civilian government fighting an enemy using a formal military organization.
  - dragonwriter 9y ago
    > This gets repeated all the time, but that doesn't make it true. You left out the “if you asked the British monarchy” part, which was fairly critical; they were leaders of a rebel movement which clearly included terrorists (both by modern standards and, even more officially, in terms of armed combatants deviating from then-current norms of warfare, by what might be imputed to be the standards of the time, though of course the term “terrorist” had not yet been applied as a proper term for those behind the still-in-the-future Terror in Revolutionary France, much less evolved to it's more general sense.)
  - trhway 9y ago
    Great debating club mental yoga. Now, what can you say about Mandela?
  - jim-jim-jim 9y ago
    There are US-designated "terrorist" groups that don't deliberately target civillians and function in a manner similar to traditional armies. Hezbollah comes to mind.
vtange 9y ago
[0] -> If we're talking robot armies, it will still depend on a "smaller" state's access to resources. You can't build robots out of thin air, unless you plan on hacking yourself one. But that depends on someone building one in the first place.
trhway 9y ago
Not a small country. A large transnational corp instead. Especially a one which has a hand/tentacles deep inside the world wide communication web plus large AI development plus great aspiration of not being evil ... Just imagine yourself in the position of wanting to make the world a better place and actually able to herd the world into that better place.
empath75 9y ago
Yeah there’s a real risk of a blitzkrieg sort of situation where some country builds a drone army that just picks a traditional army apart on the battlefield. Eventually someone is going to cut the cord and have truly autonomous armed vehicles and whoever does it first is going to have a tremendous tactical advantage.
- Razengan 9y ago
  Why do people only talk about weapons of mass destruction and dominance in these discussions? Why is Big, Bad & Heavy seen as the only way to win wars and worlds? I'd say the first group who thinks outside that box would have a major advantage: – Artificial viruses and other biological/nanotechnological infectious agents that only your side has an "antidote" or immunity for, to "cleanly" wipe out everyone else but you, without causing physical damage to cities/infrastructure/terrain. — Insect-like drones the size of flies or mosquitos that you remotely control to inject poison directly into individual enemy leaders, shutting them down before they can form or command a military opposition. – Having a permanent presence on the Moon so that some of you can survive in case of the whole world going tits up due to events you cannot control or plan for.
  - pluma 9y ago
    Sound familiar? https://www.youtube.com/watch?v=9CO6M2HsoIA https://www.youtube.com/watch?v=9CO6M2HsoIA
pluma 9y ago
Where you're wrong is that military AI isn't a traditional WMD. WMDs work as a deterrent because they can destroy a large area and kill a lot of people, sure. But the reason the Cold War never went hot is that it's immediately obvious when an atomic bomb is used and (to varying degrees) who used it. Also atomic bombs are notoriously difficult to produce -- sure, states will eventually get there (see NK) but "non-state actors" can't even begin to develop the equipment to produce them. Military AI... not so much. Sure, you might have an idea who hates you at any given moment and who has the means, but it's far fuzzier than a missile you can neatly track from point A to point B. Plus we already know how easy it is to produce cheap knockoffs. Unlike nukes, military AI isn't a deterrent. Autonomous weapon systems aren't the new H bombs, they're the new AK47s and Toyota pickup trucks.

dmitrygr 9y ago

In the modern day, the difference between civilian and terrorist is so blurred (not only in factually but also in time - fon't forget, USA's founding fathers were terrorists, if you asked the UK monarchy) even humans often cannot tell. Do we trust AIs to?

RcouF1uZ4gsC 9y ago
> don't forget, USA's founding fathers were terrorists, This gets repeated all the time, but that doesn't make it true. When did the founding fathers deliberately do mass killings of civilians? They basically won by fighting pitched battles with the British. They had diplomatic relations with several countries (Denmark and France being the major ones). There was a command structure with clear lines of authority (Continental Congress -> George Washington -> other commanders). In addition, there was at least an attempt made to follow international laws. They were rebels, yes. However, it was more a coalition of existing colonial governments (especially the elected legislatures) deciding to part ways with the British. By and large, the terrorists of our times are not at all like that. They are not part of a legitimate civilian government fighting an enemy using a formal military organization.
- dragonwriter 9y ago
  > This gets repeated all the time, but that doesn't make it true. You left out the “if you asked the British monarchy” part, which was fairly critical; they were leaders of a rebel movement which clearly included terrorists (both by modern standards and, even more officially, in terms of armed combatants deviating from then-current norms of warfare, by what might be imputed to be the standards of the time, though of course the term “terrorist” had not yet been applied as a proper term for those behind the still-in-the-future Terror in Revolutionary France, much less evolved to it's more general sense.)
- trhway 9y ago
  Great debating club mental yoga. Now, what can you say about Mandela?
- jim-jim-jim 9y ago
  There are US-designated "terrorist" groups that don't deliberately target civillians and function in a manner similar to traditional armies. Hezbollah comes to mind.

vtange 9y ago

[0] -> If we're talking robot armies, it will still depend on a "smaller" state's access to resources. You can't build robots out of thin air, unless you plan on hacking yourself one. But that depends on someone building one in the first place.

trhway 9y ago

Not a small country. A large transnational corp instead. Especially a one which has a hand/tentacles deep inside the world wide communication web plus large AI development plus great aspiration of not being evil ... Just imagine yourself in the position of wanting to make the world a better place and actually able to herd the world into that better place.

empath75 9y ago

Yeah there’s a real risk of a blitzkrieg sort of situation where some country builds a drone army that just picks a traditional army apart on the battlefield. Eventually someone is going to cut the cord and have truly autonomous armed vehicles and whoever does it first is going to have a tremendous tactical advantage.

Razengan 9y ago
Why do people only talk about weapons of mass destruction and dominance in these discussions? Why is Big, Bad & Heavy seen as the only way to win wars and worlds? I'd say the first group who thinks outside that box would have a major advantage: – Artificial viruses and other biological/nanotechnological infectious agents that only your side has an "antidote" or immunity for, to "cleanly" wipe out everyone else but you, without causing physical damage to cities/infrastructure/terrain. — Insect-like drones the size of flies or mosquitos that you remotely control to inject poison directly into individual enemy leaders, shutting them down before they can form or command a military opposition. – Having a permanent presence on the Moon so that some of you can survive in case of the whole world going tits up due to events you cannot control or plan for.
- pluma 9y ago
  Sound familiar? https://www.youtube.com/watch?v=9CO6M2HsoIA https://www.youtube.com/watch?v=9CO6M2HsoIA

pluma 9y ago

Where you're wrong is that military AI isn't a traditional WMD. WMDs work as a deterrent because they can destroy a large area and kill a lot of people, sure. But the reason the Cold War never went hot is that it's immediately obvious when an atomic bomb is used and (to varying degrees) who used it. Also atomic bombs are notoriously difficult to produce -- sure, states will eventually get there (see NK) but "non-state actors" can't even begin to develop the equipment to produce them. Military AI... not so much. Sure, you might have an idea who hates you at any given moment and who has the means, but it's far fuzzier than a missile you can neatly track from point A to point B. Plus we already know how easy it is to produce cheap knockoffs. Unlike nukes, military AI isn't a deterrent. Autonomous weapon systems aren't the new H bombs, they're the new AK47s and Toyota pickup trucks.