[flagged]
Imagine having a patent on 'fire' and then suing everybody who tries to cook a meal.
Leave us behind whom or what?
I agree with gp. It may not be LLMs, but we will certainly create a technology at some point that can't be openly shared due to existential danger, aka The Great Filter.
We can't just naively keep frolicking through the fields forever, can we?
We have to be able to at least agree on that, theoretically, right?
1. either lock everything down and accept the control of a small unaccountable group to dictate the future of humanity according to their morals and views - and I believe that AI will fundamentally shape how humanity will work and think, or 2. continue to uphold the ideas of individual freedom and democratic governance and accept a relative increase in the chance of a great filter event occurring.
I, like many here, am firmly against ggp's position. The harm that our spices sustains from having this technology controlled by the few far outweighs the marginal risk increase of some great filter occurring.
I will continue to help ensure that this technology remains open for everyone regardless their views, morals, and convictions until the day I die.
The only benefit I can imagine of less players having control of a technology is that there are less chances for them to make a bad call. But when you democratize something you hit the law of large numbers.
https://en.wikipedia.org/wiki/Law_of_large_numbers
disclaimer: this goes against so much of what I believe, but I can't escape the logic.
Whom: The corporations with enough money to burn.
What: Technological progress.
Here's a nice video that showcases the same patterns in history and how having free and open tech + breaking monopolies helped move society forward - https://youtu.be/jXf04bhcjbg
AI, on the other hand, will be dangerous by default, once it's powerful enough.
It’s like balancing a piano on a ledge above a busy street and saying “well if no one pushes it then it’s not dangerous!”
Nuclear war and climate change rank far higher as threats than rogue AI to me right now.
Here is what mine had to tell you:
It’s not like we don’t already have nuclear weapons, biological agents, chemical agents etc...
AI is simply another tool which can be used for good or ill. It doesn’t matter how much regulation/control you put on it - if someone really wanted to use it maliciously then they will find ways around your safeguards. The best thing to do is educate yourself as much as possible.
What would you think of a world where only one country has nukes (due to a monopoly on radioactive fuel, rather than a monopoly on knowledge)?
This is more like giving every individual on Earth the nuclear launch codes. It only takes one stupid or malicious person to press launch. Giving more people buttons is not how you avoid thermonuclear war.
To do serious harm or have broad social control requires concentrating that power with an infrastructure that a small group does not have, it requires coordinating the resources of a broad social base. And at that point the incentives to use them are affected by the needs of many people.
>requires concentrating that power with an infrastructure that a small group does not have, it requires coordinating the resources of a broad social base.
These are all things intelligence (artificial or otherwise) can help acquire. It listens, thinks, and responds. Genghis Khan, Adolf Hitler, and Albert Einstein are all intelligences that resulted in dramatic tangible, physical changes to our world almost entirely by listening, processing, and responding with human language.
A small number of slow and weak apes came to have absolute unilateral control over the destiny of all other lifeforms because of intelligence. The power, infrastructure, and resources you speak of were not available in 15,000 BCE, yet somehow they exist today.
In that case, this current batch is not "advanced AI". It is a big autocomplete panel which retrieves content already present in its training corpus, and presents it in new shapes; it is not ready to define new goals with a purpose and follow them to completion.
There is a really big step to go for it to be self-learning which is what is one of the things it will need to be self-aware. Right now the tech is just a static model - it will not learn from being corrected. You can often argue with it saying "Hey this is wrong because..." and it will admit you're right. And then it will give the wrong initial answer back the next time.