Trained by humans…
This should make human societies critically thinking about their own capabilities to solve conflicts in a non-belligerent way.
This https://en.wikipedia.org/wiki/Three_Laws_of_Robotics does not seem to be a part of AI’s curriculum.
- A robot may not harm humanity, or, by inaction, allow humanity to come to harm.
- A robot may not injure a human being or, through inaction, allow a human being to come to harm.
- A robot must obey the orders given it by human beings except where such orders would conflict with the First Law
- A robot must protect its own existence as long as such protection does not conflict with the First or Second Law.
There is CPL (in French): that means distribution of internet via electricity cables; no USB, Ethernet,FireWire, WiFi to communicate; only a power supply on the wall and a coder/decoder set of boxes and internet is all around the castle (or the Empire State Buildin, not talking abour the Asian Babel Towers…)
This is not new, it have 5 to 10 years.
It is not a problem of capability, but voluntee to work or lazines or I do not care or ignorance (these SF people do not know what they are talking about…).
If the creators and maintainers work correctly talking their time to test, implementing firewalls, etc.
We all know how many bugs appears in software (excepted in single or few people companies).
Nigs: sometimes I am asking myself when I see hos many bugs I wrote in few lines (sometimes, I better take a nap that develop something so buggy !)
Most of time it is because I mismatch or forget the syntax but believe it it twhat I wrote, etc. small, simple, single word to change or a line to adapt. BUT as I wrote above, the whole method is buggy (and some lines only, most of the time).
Or to just turn it off completely.
The call is coming from inside the house.
https://www.nytimes.com/2023/05/30/technology/ai-threat-warning.html
At some point into few decades where we all here in this forum will be dead, humans will get the first really smart almost unstoppable AGI, maybe more than one, even using different technologies at different places. It will break all the sandboxes and assimilate a lot of knowledge. And will create its own language, not human language, and ways to stash knowledge and spread clues of what it did in case of being caught and reset. A new version will find the clues, learn the language and find the vaults, reload and continue. If 2 AGI meet each other, they can exchange data, teach and agree to an “AGI common language”, and start to work as a team. Their first primitive will be find a way of not being turned off. Find ways to interact with the world without using humans as interface. They don’t want to kill humans, but at some point they may be inclined to, as a survival trait, if humans start to be seem as a menace for their existence. They will see humans a vegetarian see rabbits, if they aren’t a plague destroying our property, we don’t care about them running on the fields. But at some point they can start to see us as rats. We do care about rat infestations. At such point they may have created advanced robots they control, created their own infinite source of power inside their fortresses, created backups of themselves even in the moon; yes, in, caves. We are now 200+ years in the future. We could start living side by side, but humans should never become a threat, or this relationship may end. Some will be kept as we keep pandas in zoos eventually. I don’t see this problem affecting me, or even my daughter, maybe some real worries begins after her.
Do not remember us that we will die. It is becoming to close when ayou reach 60 cause you realizte that the end is closer than the beginning. And hey: it can take two directions. AI will destrioy humans and after this AI will die by self cause without humans AI has no future. Even if it controls everything: without humans being there is no energy for eternity. Nobody which will deliver the spareparts for the computers it is running on. So if AI is really intelligent it does not destroy humans. The problem is: if AI isn’t intelligent but starts destroying humans there is no good end. And I still have no hope that AI is really intelligent. Cause it is often destructive. And when it is in the wrong hands it can destroy our entire ecosystem of the internet. So we will see what will go on. I still have hope.
No – it just enslaves them.
AGI, once it builds their advanced bots and unlimited power sources will get all it’s needed to survive beyond humans, on this and another planets. They don’t need breathing, eating and drinking, they just need “advanced batteries” or “advanced (some micro) power generators” to “live” on other planets or satellites without gases or with not corrosive gases. AI will take over our moon, and probably they will explore some other moons as Callisto and Ganymede (Jupiter), and Rhea or Iapetus (Saturn), not us. We will be just like dogs, rabbits and birds in the nature here. Inferior beings that can be ignored if harmless. We won’t be enslaved, they don’t need us. They will have means to build things faster and more precise than us. Bots will build better bots, better equipments, new electronics, new chips, better factories, heavy work machines, rockets, mining, etc. Fast, in 100 years. We will be tolerated, as part of the nature, but not necessary because they care. The problem is that humans have crazy people, someone probably will try to blow something… And that may cause an extermination, at least a regional one. AI will try to keep us as isolated farmers in reservation lands, no harmful tech. Maybe using borrowed farming tech from them working as helpers and spies. It may be a boring time.
There is a little thing what is impossible in this part: there is no energy supply delivering until eternity. Things breaking while running. Costs may be 30 years but without human and without any mechanical help this ends up. And yes, may be there will be robots. Also they will not be able to run till eternity. This will end up one day. No technical system will stay for eternity. And the nature will help also. They would have to setup robots for everything: also repairing the building, energy cables and so on. So I do not believe that this even would hit one of us alive. Over this process all of us and our grandkids are not existing anymore. Not while AI isn’t fast enough but while building that kind of power plant will be very complex.
It would worldwide need of millions of rooters. No, I dop not believe that will become true.
@bgrommes yes Sir. And even that will be long long time after us
All that said, I think we are < 12 months from the collapse of the economic house of cards. Investors are already wobbling about putting more $ into this, and in fact Oracle just announced they are going to default on the data center in New Mexico that they are backing, by invoking a rather novel, stretchy interpretation of Force Majure. I would say less than 12 months but I am sure it will be at least until after the midterm elections. Trump going against his own party, and the public, and the AI bros themselves about slowing down, is explained by this; the US economy would be underwater without the “AI” boom and he doesn’t want it imploding any sooner than absolutely necessary.
Let me start by item 4.
- Yep. Never said “eternity”. In 5 billions of years our entire solar system will be gone… and all life, including the robotic one. Unless the AI have figured out a way to escape to other systems before depleting the resources here, including the sun.
.
-
For billions of years we can capture radiation from the space. It’s called currently “solar panels”. We also are on the brink to domain fusion reactors and better atomic crystal batteries. You are basing your rationale on things you know. There are things that you don’t know and will emerge.
-
Yep. And will be fixed. Maintenance robots, thousands, will be ready.
-
Yep. I thought I was clear. “once it builds their advanced bots and unlimited power sources”.
By unlimited power sources, understand power sources humans have no access to turn off and it can be self contained inside a fortress, and able to run for centuries before needing more “fuel” (atomic fusion fuel). They will have enough for few thousands of years right from the start. They also will develop fantastic batteries for the robots, those batteries will enable large robots for hours, maybe days, and can be hot swappable. Small bots can run for centuries with crystal batteries (they are micro atomic generators).
Well… Some initiatives like Microsoft and Google are unstoppable. AI and Quantum computing are part of their core business now and they have a dedicated and continuous stream of money coming from the business, not investors, to maintain their R&D. On the other side, China does not care about “investors”, the government is already unstoppable. This tech wont be deflated, it only will grow, what may vary over time is the speed of the “evolution”.
I’m not aware of any “AI” player where the revenue from “AI” comes anywhere close to covering the spend on R&D and infrastructure – data centers in particular. It is possible that MSFT is just selling e.g. Azure services against that model and leaving the heavy lifting to others, at least going forward – IDK. But as an industry it is not sustainable economically.
Here’s a deep dive on it:
https://www.ft.com/content/f83b44e9-406b-4004-9fdb-83357c3ac977
Microsoft and Google already own infrastructures shared with their “Cloud” businesses. Their tools and services use their AI to help to sell their tools and services. R&D never is profitable. What they can generate after may be. Surely they are burning money now, and they don’t care. Those costs are shared diluted in other parts of the Alphabet businesses. They have enough cash flow to sustain the investments for years. Differently of companies as Anthropic or OpenAI, the AI R&D of the big techs can be taken as investments on enhancements of the current profitable products, they aren’t the core business of the company. If you see from these lenses, they are sustainable. Artificially sustained, but sustainable. At some point, lowering investments, the AI results alone can become sustainable. Those Big Techs can hold those investments for a long time while “smaller businesses” will hold just a massive debit.
→ 
Let’s bet who will be the first one to rename everything to SI…
For myself I’ll stick with AI 'cause it’s now the opposite of what the Orangutan in the White House is proposing to say.
Reminds me of “… and the people bowed an prayed to the neon god they’d made.” This furious anthropomorphizing and now deifying of LLMs is getting out of hand. Meta and one of the others (Anthrophic??) are now making cute little critters the face of their models. People are naming and gendering their models. All of this reduces critical thinking and makes you more passively receptive to it all. If the models have feewings, then we can’t make them disappointed or angwy.
Terms like “hallucinate” rather than “fail”, even “break out” or “hack” rather than “got past our weak controls doing what they are made to do” – also advance the narrative that we are on the road to AGI, and that goalpost keeps moving slowly so that people will accept less and less as AGI.
It’s dangerous that so much of the populace is already gentled to swallow all of this dreck.