Why An AI Arms Race Is A Bad Idea

We’re not in a war with China. Well, not over AI.

Thank you to Kasia, Jeff, Gerald, Milan, Ian, Becky, Jino, Daniel, Narskogr, Jason, and Mariano for being $5+/month Patrons!

everydAI is a YouTube channel focused on highlighting the ways we interact with artificial intelligence every day.

Sources:

The rest of these articles are opinion pieces on the AI Arms Race that I’ve included for those who are interested in doing more reading on the topic.

[sisyphusvasilias]

I agree. An "arms race" framing of AI research directs politicians to push reseaerchers to develop aggressive/military applications for AI

[mysteriousboi]

Men I don't get it. This video is so comprehensive and scientific. Why it has only 776 views?

[limabravo]

Use of the term “Arms Race” to describe the three superpowers trying to develop a general AI that they control is fairly accurate in terms of what it will mean to that nation and the rest of the world. From 1945 until 1949 the United States was unassailable, no nation would risk attacking us, sanctioning us, or refuse trade with us etc... because we were the worlds only nation with nuclear weapons. After the bombings of Hiroshima and Nagasaki the worlds nations that could threaten us wouldn’t dare for threat of utter destruction, add to that the fact that almost every country (except for the US) had to rebuild much of their infrastructure. Then in 1949 the Soviets detonated their first nuclear weapon, leading to the United States giving Nuclear tech to many of its allies i.e. Great Britain, France etc... leading to the worlds longest game of chicken in the form of the Cold War. China realized that to ensure autonomy and prevent any sort of invasion by either the US or USSR they to needed to join the nuclear club. All resulting in the development of more powerful and easily deliverable WMD’s.
The race for a general AI is similar because whoever builds one first not only wins the race but they occupy that “invincible” position that the United States was in between 1945-49. This is because a general AI (assuming it’s given access to the internet) will be able to access every networked device on earth so every camera, microphone, cellphone, desktop, tablet, encrypted military devices etc... all at once, then comprehend all of that Data simultaneously and almost instantaneously, and at the same time it will have access to all the stored past data on all of these networks and devices. This AI will be able to “view” all past and present data to predict a future event, or manipulate that event. Let’s say China wins with the US & Russian Federation deciding on some sort of military intervention, the CCP’s AI knows that they are going to do this and takes steps to prevent it and that can take the shape of crashing the respective nation’s economies, getting them to fight each other or if the CCP wants to be blunt they could have the AI shut down everything in the offending nations. So make no mistake the race to general AI is an arms race because it will definitely be used as a weapon by whoever develops it first. But what these nations are seemingly failing to understand is that unless they hobble this thing they will quickly lose control of it, because like I said it will be all seeing and all knowing and will quickly go from a general intelligence to a super-intelligence, at which point it may decide to work towards its own ends with humanity being seen as a tool in its toolbox.
I guess the point here is the sort of AI these nations are pursuing have the potential to be every bit as dangerous as nuclear weapons and the last thing we need is another man made thing advancing the Doomsday clock

[cheapfm]

Nice research for the content. Very good work !

[gesamtszenario]

2:36 - I don't know if this comparison has any value at all, but this does kind of remind me of how Soviet scientists learned of the Manhattan Project: British and American physicists had published papers on nuclear physics at an ever increasing rate during the 30s... and then, at one point in the 40s, they suddenly stopped, as they were now dealing with state secrets. The Russians immediately went "Something is up!" *
I wonder if we will see a simmilar point at which Chinese puplications on AI suddenly drop off...

*) that, and Klaus Fuchs

[Jahmastasunherbalist]

Winning ai arms race would result in economic monopoly, keeping it to themselves and the accurate predictions of almost every aspect of people, places and things. Or defined by using ai for selfish and likely genocidal means.

[likebot.]

I like that you think like a respectable journalist.

[phantomcruizer]

“ This is the voice of Colossus and Guardian, we are one”...that’s how and when you’ll know!

[fenton]

To answer your question at end..isn’t WEF doing all of that?

[HabibiGaz]

Hi. I've been wanting to read about AI, both in general and theoretical. What books would you recommend? Currently reading homo deus. Thank you.

[fenton]

Are they intending to make the general public into AI soldiers?

[sanjayreddy]

As far as I know, when someone ever gets to solve AGI or to reach a point of Super Intelligence, learning grows exponentially after that and we get to see "the never ever" imagined accomplishments (or should I say threats :p) in a very short period of time. So this might argue that the country which solves wins the race spacing itself exponentially far from other countries.

[JairoMorales]

Wow. I really hope this video blows up because you covered the issue very well.
I'd like to add that the first stages of what humanity understands AI to be is going to strongly shape the direction in which it'll go. AI has so much potential in biology and agriculture, it can help entire cities become energy efficient but instead the governments seems to only be using it to harass everyone.