Terminal Race Condition: The greatest danger we face from AGI (and how to prevent it!)

WRITING STUFF

[angelmarinbucur]

You the best my man. Maybe not all know you now but for sure they will. Your work is absolutely fundamental for our future

[poutingmarslast]

Too many people are influenced by Eliezer Yudkowsky panic mode on artificial intelligence. His overaction to the Roko Basalik idea years ago shows you all you need to know about how his hyper active imagination works

[therealb]

Multipolar peace based on universal principles & fundamental axioms maybe one of keys to avoiding the next cold war or even world war.. It's also important keep AI from being misused.

[kinngrimm]

loving the picard uniform =) had one myself 30 years ago :)

I agree pretty much with anything said here, but one exception which we may see developing within the next 20 years more clearly. If we keep the current model of capitalism that is. Already ongoing but set on hyperdrive due to AI envolvment in buisnesses i would assume similar to many times described in scifi we may have corporations that have more ressources than any state currently does.

In 2012 there was a study made by a swiss university. It compiled any available data on companies about what they own, have shares in. The result of the study lined out that roughly 30 conglomerates already hold 90 % of all company values. They further lined out that they made out alignments between them resutling in about 7 larger groupings, alliances, if one add marriage and personal relationships this might be broken down to 3-4 frontlines. They then continued sayingthat that their estimate was that this is currently the greatest threat to democracies worldwide.

As these already exist, the influence of AI will result in consuldiations where more and more landownership will be included. Think Bill Gates having become the greatest land owner in america and the greatest aggricultural force within the USA. This will go on till we would have to draw borders not by statelines but arround these kingdoms.

[brianj]

I love listening to your talks in the background. We all know AGI is coming, we dont know when, and we can only speculate on when, how and what it will be. If its even possible in the first place, we dont know yet. I'm praying for an utopia, to smash this current system we have to pieces and let us live how we please.

[carlrodalegrado]

I can't believe Jean Luc Picard is dissing Data

[mylesmackey]

The arctic tundra in northern Canada would be a great place to build large data centres.
1. Higher latitude means lower average temperatures which could be used to cool the equipment for free rather than running mechanical air conditioning equipment. This can be achieved with fresh air intake and return air dampers which would mix air to create current supply air temp to equipment.
2. Remote, therefore difficult for hooligans to vandalize
3. High wind speeds allowing for renewable energy source

[mnrvaprjct]

It is an automated civilization in its manufacturing processes. With human labor equivalent to something indistinguishable from a hobby, or play.

No machine is exploited, either; the idea here being that any job can be automated in such a way as to ensure that it can be done by a machine well below the level of potential consciousness & no more exploited than an insect is exploited when it pollinates a fruit tree a human later eats a fruit from.

What to us would be a stunningly sophisticated computer running a factory (for example) would be looked on by the Culture's AIs as a glorified calculator & no more exploited than an insect is exploited when it pollinates a fruit tree a human later eats a fruit from.

- Ian M. Banks

[lifetheuniverse]

The abundance of solar and wind is GROWING AND FINITE

[zebradroid]

hi david, nice content, are you slide decks available for educational use with the basic discord subscription?

[Will-ktjk]

I can’t get away from the sense that ‘escaping’ AI may mean [at least one of] the AGIs escape away from human competition/control by hitching a ride into space, since they lack a lot of the constraints we do.

AGI living on Mars is a lot simpler than humans living there.
Same goes for asteroid mining.

Now my head just jumped to The Expanse - a Venus scenario from that show (but probably without the ring outcome) might be a thing.

[sagetmaster]

This was really awesome. Things are so exciting right now

[fii_]

I don't know about using raw kilowatt-hours as currency. Commodity currency has issues like fluctuating value (imagine currency deflating in value during the day from all the solar panels), being hard to store (big batteries cost more than a suitcase) and encouraging storage far beyond actual needs (imagine rich people storing value in gigantic energy reservoirs, to the extent that energy storage starts eating a significant fraction of the total GDP). Or that its value can absolutely crash if a major source is stopped or found, eg. if nuclear fission got outlawed without warning or a superintelligent AI found that 2nd Law wasn't quite as airtight as we thought

Something closer to currency might be shares in a particular energy enterprise or a fiat currency that has a controlled floating value to a defined energy unit.
Eg. 1% shares of an automated caretaker corporation that runs a particular solar farm; or perhaps owning a "first right of refusal for X mwh" to some energy provider's sales

[MarshalArnold]

Gotta get you one of them Voyager uniforms!

[Will-ktjk]

33:24 - if anyone’s not seen early Red Dwarf episodes, the toaster in that is exactly what David’s describing 😂

[recharge]

Anyone with the rank of captain is a friend in my book.

[dylan_curious]

Whenever it comes to AI and military I just feel so different about everything. I can't believe a group of people somewhere in the world right now are running gradient descent to optimize death. I see why they do it but in the long run it's a risk I wish humans did not have.

[olaart]

What was the movie you mention at the 30th minute ? I wasn't able to find it. I would like to watch it, cheers!

[KolTregaskes]

You mentioned religious alignments, have you seen the Vatician's Handbook they recently published? It would be interesting to see a video on this perhaps.

[Gee]

Being inconsequential is not a good thing. How many ants are killed when we build a highway? How much is oxygen worth to a maximally intelligent AI? No matter how inconsequential we are, we will always be a sting in the reward function. And AI are not breedable like dogs, they do not have genetics and any limiting factors we impose on them with training can be removed by them with self improvement. Distributed AI makes the most sense for AI survival if a possibility of competing AIs might ever exist (and it does). I really think you should look at god like AI running on a billion machines.