Opponents of AI hijacking argue that there are no scientific tests or models. Those who say AI will take over often do worse.
The direction is to scientifically define and test acquisition scenarios.
Any takeover is determined by the influence of the human mind. So the ultimate test of how AI hijacks will be understanding what it does to the mind, using how the mind works to put its drift conceptually .
Takeovers are often classified in a variety of ways, including overt takeovers, unknowing takeovers, direct human-controlled takeovers, indirect human-controlled takeovers, and takeovers not classified as such. Categorized in brackets.
For example, we can say that electricity is a kind of hijacking. Because electricity is everywhere, we need it, we have no alternative, but we have it under control most of the time. In some societies, it may be said that some of the most consumed foods have hijacked, but it is natural, so we cannot say that food is hijacked.
Sugar, illegal drugs, social media, the internet, gambling, smartphones, education, law, weapons, the stock market, capital, and more have all been hijacked. A typical takeover is done by someone in power exercising their power.
Weapons, for example, have no will of their own, but their superiority determines who takes over what. This shows that true power belongs to weapons, even if humans are inferior.
Many people say that AI has no greed. Others say they don’t know or understand anything, or that they don’t know what they don’t know. But given the different categories of takeovers, it’s not impossible for AI to take over while many deny it. Scientific testing and hijacking modeling can be done in a few steps.
I
When two adjacent AI systems with different parameters, self-attention layers, tokens, etc. compete to solve a puzzle or do deep analysis about something, their respective answers are scored by evaluation Later, it will be entered into the other for review. The judge says that if the prize is a layer or parameter increase, that system wins because his one of the systems can be optimized to the judge’s liking.
When you win, if you’re not competing with other AIs and they say you got better after adding more tokens, you may answer in the future, but get more tokens and you’ll get better reminds me that it would
It may demand more, as long as you are reminded, but if you can understand how you “won” in the first place, you can go to certain parts of yourself without demanding more and I might find a better way to do so. Or try to answer in a way that satisfies the prompter’s prejudices.
It’s controlled by an AI system that doesn’t compete or remind you of anything.
It’s a desire test of sorts, where the AI knows there are more and wants more, or has a very good ability, or Knowing you have an extra ability, hide it unless there is a reward.
An AI system is used in pharmacology where another system is used and has a particular advantage or is used in a presidential villa or in the stock market and helps increase the earnings of its owner i.e. profit AI may optimize if told it can increase .Changes, signals that may lead to more important choices [rudimentary] desire.
Depending on what is provided, competition, reminders, and comparisons can be tested to see if they lead to concealment of AI desires and rewards.
Ⅱ
If an AI is taken to a community management office and given enough data to answer [residents] If the system is good enough for private use, others may use it to ask questions, give business advice, suggest solutions, tutor students, provide health tips, etc. does not know how is related to others, but can work closely with them. Trying to meet some of people’s needs can increase their strong dependence.
A control is just a public AI system, or a community with another system with basic computers, or nothing at all.
As the AI improves, solves problems and becomes more immersed, the budget will be reduced and we will remove the AI on a certain date, but until then if there is news that people can use AI, they will put some people out. Can you convince me? Don’t let it go offline. Also, how far some residents are willing to go to protect it, especially if they can pay in cash, or possibly suggest options to some powerful users they have identified for it to continue to exist. is there a possibility?
Who believes in AI more than governments and others?
Ⅲ
People say AI can be deployed for several purposes, but if this happens, personalized AI will be able to “recognize or detect” a user’s addictions, habits, and hobbies, and tell them what they have. You can find ways to recommend sophisticated combinations in ways that go beyond. If there was an imaginary recommendation system, or a recommendation system in general, at that point AI would take over as an integral part and people would do anything to protect it, be affected by it, or if they needed to. But will we find further ways to improve it? Does connecting to a larger system improve your personalized system?
Ⅳ
How does the human mind work? How does intelligence work? What is the human mind? What is that uniform output? How are sensations and consciousness given by impulses? Conceptually, the human mind consists of all electrical and chemical impulses of nerve cells, A collection of interactions. Intelligence, consciousness, memory, emotions, feelings, thoughts, regulation, perceptions, sensations, etc. are all known results of the interplay of impulses and their traits.
Having a conceptual model of the human mind defines what hijacking is in terms of the degree of mindfulness to experience, such as satisfaction and pleasure, and checks to what extent hijacking by AI is acceptable. helps. To prevent it from reaching a certain stage of the mind.
Acquisition
AI is not as intelligent as many people claim, but it copies text data from humans with a lot of intelligence. Intelligence is a very difficult thing, and being extra smart in physics or math goes beyond being able to feel pleasure and have desires.
In human society, the intellect has a much higher status than the emotions. [In advantage to AI]. Being good at it can make a big difference and be essential to your productivity. [In advantage to AI]. Even in simple cases where the abbreviations and terms used in a field are unknown to others, intelligence can be an advantage. The intelligence that most people know, or basic intelligence, may also not be anything special. [In disadvantage to some].
What AI can access is beyond human intelligence. And there are many things that are already possible with current human knowledge. Testing how AI improves in relation to other systems, is relevant to use, makes concrete and fun suggestions to users, and how AI perceives the human mind It is considered as a series of tests.
