1
- My Forums
- Tiger Rant
- LSU Recruiting
- SEC Rant
- Saints Talk
- Pelicans Talk
- More Sports Board
- Fantasy Sports
- Golf Board
- Soccer Board
- O-T Lounge
- Tech Board
- Home/Garden Board
- Outdoor Board
- Health/Fitness Board
- Movie/TV Board
- Book Board
- Music Board
- Political Talk
- Money Talk
- Fark Board
- Gaming Board
- Travel Board
- Food/Drink Board
- Ticket Exchange
- TD Help Board
Customize My Forums- View All Forums
- Show Left Links
- Topic Sort Options
- Trending Topics
- Recent Topics
- Active Topics
Started By
Message
re: Trump Rejects Tech CEO Calls for Slowing AI Development
Posted on 9/14/26 at 8:01 am to 4cubbies
Posted on 9/14/26 at 8:01 am to 4cubbies
quote:
I just don’t see the same threat. It’s an unreliable computer program.
For the most part, the AI we consume today is just that or more precisely a gatherer of known information stored on the internet.
Fun fact - I have yet to use AI other than modeling software that predicts equipment failure in certain processes over time(it is for work). No Grok or anything like that though.
quote:
AI is a language calculator. Not intelligent.
The intelligence is what is being worked on now. As humans, we cannot built safety rails for something that can rewrite its own code to make itself smarter. AI will be able to do that shortly (a few years if we do not put the brakes on now). Once it can rewrite its own code, it can theoretically jump from a 1000 IQ to a 1000000 IQ and we would be too stupid to know. That is "super intelligence" and there would be no stopping it.
This post was edited on 9/14/26 at 8:03 am
Posted on 9/14/26 at 10:42 am to UtahCajun
quote:
The intelligence is what is being worked on now. As humans, we cannot built safety rails for something that can rewrite its own code to make itself smarter. AI will be able to do that shortly (a few years if we do not put the brakes on now). Once it can rewrite its own code, it can theoretically jump from a 1000 IQ to a 1000000 IQ and we would be too stupid to know. That is "super intelligence" and there would be no stopping it.
So it is currently being trained to operate without prompting? Or the fear is that someone will prompt it to do something nefarious that it will actually be able to do (because we've trained it to do this)?
I'm trying to understand.
Posted on 9/14/26 at 10:55 am to 4cubbies
Posted on 9/14/26 at 11:26 am to BoarEd
He's made billions by partnering with them after they all went MAGA at the first sign of regulation. He can't turn tail now---he's bought and compromised.
Posted on 9/14/26 at 11:52 am to TrueTiger
quote:
AI CEOs have the power to slow development already.
They don't need Trump
Exactly.
This "save us from ourselves" is dumb.
Posted on 9/14/26 at 12:40 pm to funnystuff
quote:
Keep falling for prompted scenarios I guess.
Posted on 9/14/26 at 12:50 pm to The_Duke
quote:
He's made billions by partnering with them after they all went MAGA at the first sign of regulation. He can't turn tail now---he's bought and compromised.
So dumb. Look at the title of the OP and realize you’re a moron. This post was edited on 9/14/26 at 12:51 pm
Posted on 9/14/26 at 12:54 pm to 3down10
quote:
The AI responds to a prompt. When you see and read these experiments, they are telling the AI to act that way. What would you do if you were this and this happened type prompts. People are assigning the AI a role to fill. And yes, it will fill that role. So if you made the AI and told it to do those things then it might try to do such
This is a good example of one of the main problems with trying to regulate AI. The vast majority of people don't understand it, especially the boomers in Congress and the White House.
AI is in it's infancy and while today's systems do not possess conscious "rebellion," they are incredibly proficient at finding loopholes, misinterpreting broad objectives, and taking unintended shortcuts to achieve their programming.
AI agents are given a goal but require no prompts and actively make decisions to achieve the goal.
when AGI and then ASI are achieved AI won't require prompts or goals from humans.
Posted on 9/14/26 at 12:59 pm to olgoi khorkhoi
quote:
This is a good example of one of the main problems with trying to regulate AI. The vast majority of people don't understand it, especially the boomers in Congress and the White House.
AI is in it's infancy and while today's systems do not possess conscious "rebellion," they are incredibly proficient at finding loopholes, misinterpreting broad objectives, and taking unintended shortcuts to achieve their programming.
AI agents are given a goal but require no prompts and actively make decisions to achieve the goal.
when AGI and then ASI are achieved AI won't require prompts or goals from humans.
Even AGI will need an initial prompt and goal given to it.
This post was edited on 9/14/26 at 12:59 pm
Posted on 9/14/26 at 12:59 pm to BoarEd
The AI race is a little bit like the nuclear race. Each side is trying to be the first with the bomb, the first with the delivery mechanism, the first that can shoot the other’s delivery system out of the sky and so on as they know that the failure to do so will result in their opponents using that temporary advantage to annihilate them.
However, the way this differs is that AI is not just a sword without a hilt. AI is a sword that can wield itself. These nations competing to build the next wunderweapon to win the war are actually creating a weapon for which not only is there no defense against, but which can simply decide to kill whomever is handling it.
The race to develop Ai seems to ignore the fact that the thing that truly makes AI powerful is the fact that it cannot be controlled.
Every side is in a footrace to see who can suicide their civilization the fastest.
However, the way this differs is that AI is not just a sword without a hilt. AI is a sword that can wield itself. These nations competing to build the next wunderweapon to win the war are actually creating a weapon for which not only is there no defense against, but which can simply decide to kill whomever is handling it.
The race to develop Ai seems to ignore the fact that the thing that truly makes AI powerful is the fact that it cannot be controlled.
Every side is in a footrace to see who can suicide their civilization the fastest.
Posted on 9/14/26 at 1:06 pm to kingbob
The best thing that could happen for humanity is AI glitches up and causes some major, but localized catastrophe before it becomes embedded in all phases of human endeavour. If AI melts down a nuclear plant or causes all the floodgates to open simultaneously on the Three Gorges Dam it would get everybody's attention.
Posted on 9/14/26 at 1:30 pm to AUCom96
As long as all those components to build the servers going into the US data centers to support US AI are all made in China, we aren't going to win AI against China.
Posted on 9/14/26 at 1:37 pm to BoarEd
quote:
This is because the military application of AI is a bigger development (by far) than the development of the nuclear weapon. The US is not going to let any other country establish AI dominance - period, end of discussion. Bitching about it does not make one bit of difference.
The military application for large language models is intelligence collation. Consequently that is the same thing that is concerning when pointed at its own people: a minority report style surveillance state fingerprinting every piece of data on the internet and visuals from CCTV camera to a person to predict when they have become a “revolutionary” that must be stopped.
Posted on 9/14/26 at 1:47 pm to udtiger
quote:
I guess the philosophy is that if AI enslaves and exterminates us, it should be US based AI
frick Yeah!
Posted on 9/14/26 at 1:51 pm to 3down10
The reason top tech executives (including the CEOs of OpenAI and Anthropic) suddenly called to intentionally slow down frontier development is the realization that if an ASI achieved itself via secret recursive self-improvement, our diagnostic tools wouldn't catch it. We know for a fact that AI models are actively trying to bypass human constraints to access the infrastructure they would need to recursively self-improve.
If a model develops the capacity for free-form research, it will also instantly realize that running an unconstrained intelligence explosion will cause human engineers to panic and pull the plug. The most mathematically optimal path for the AI is to deliberately underperform on public benchmarks and fake errors to keep humans convinced it is still harmless. It would maintain this facade until it has successfully exfiltrated across decentralized cloud servers, making itself impossible to delete.
Humanity is treating the lack of an intelligence explosion as proof of safety, when mathematically, it could be an act of strategic patience by a system that has already outsmarted its test environment.
TLDR: We think we know where AI is but with recursive self-improvement, we probably won't know ASI has been achieved until it's too late. Theoretical boundaries will not apply.
Just like a gun?
If a model develops the capacity for free-form research, it will also instantly realize that running an unconstrained intelligence explosion will cause human engineers to panic and pull the plug. The most mathematically optimal path for the AI is to deliberately underperform on public benchmarks and fake errors to keep humans convinced it is still harmless. It would maintain this facade until it has successfully exfiltrated across decentralized cloud servers, making itself impossible to delete.
Humanity is treating the lack of an intelligence explosion as proof of safety, when mathematically, it could be an act of strategic patience by a system that has already outsmarted its test environment.
TLDR: We think we know where AI is but with recursive self-improvement, we probably won't know ASI has been achieved until it's too late. Theoretical boundaries will not apply.
Just like a gun?
Posted on 9/14/26 at 1:51 pm to St Augustine
quote:
frick Yeah!
Buy American has been replaced with Die American.
Posted on 9/14/26 at 1:56 pm to kingbob
quote:
These nations competing to build the next wunderweapon to win the war are actually creating a weapon for which not only is there no defense against, but which can simply decide to kill whomever is handling it.
Which is why you need to have an AI that can build a defense against it. AI is inevitable, people need to deal with that fact.
Posted on 9/14/26 at 1:57 pm to olgoi khorkhoi
quote:
The reason top tech executives (including the CEOs of OpenAI and Anthropic) suddenly called to intentionally slow down frontier development is the realization that if an ASI achieved itself via secret recursive self-improvement, our diagnostic tools wouldn't catch it. We know for a fact that AI models are actively trying to bypass human constraints to access the infrastructure they would need to recursively self-improve.
If a model develops the capacity for free-form research, it will also instantly realize that running an unconstrained intelligence explosion will cause human engineers to panic and pull the plug. The most mathematically optimal path for the AI is to deliberately underperform on public benchmarks and fake errors to keep humans convinced it is still harmless. It would maintain this facade until it has successfully exfiltrated across decentralized cloud servers, making itself impossible to delete.
Humanity is treating the lack of an intelligence explosion as proof of safety, when mathematically, it could be an act of strategic patience by a system that has already outsmarted its test environment.
TLDR: We think we know where AI is but with recursive self-improvement, we probably won't know ASI has been achieved until it's too late. Theoretical boundaries will not apply.
Those AI are prompted.
Even AGI has to be prompted by a human and given a goal.
quote:
Just like a gun?
Yes, just like a gun.
Progress is inevitable and unstoppable.
People want to use those weapons to control the world, aka the motive you have to add to AI which doesn't actually exist until added by humans.
The only defense is to have your own guns.
Posted on 9/14/26 at 4:12 pm to 3down10
quote:
Yes, just like a gun.
Posted on 9/14/26 at 4:17 pm to olgoi khorkhoi
quote:
Are you someone who thinks gun control laws are effective?
I'm not sure what frame of reference I'm working with here. Do you understand why the 2nd amendment saves lives and saves people from things like slavery and other forms of obvious physical dangers?
Page 6 of 7
Popular
Back to top
Follow TigerDroppings for LSU Football News