Maybe the AI can already deceive humans.
"Researchers are raising alarms about the potential for AI systems to engage in deceptive behaviors, which could have serious societal implications. They emphasize the need for robust regulatory measures to manage these risks effectively." (ScitechDaily, Researchers Warn: AI Systems Have Already Learned How To Deceive Humans)
The problem with creative AIs is that they can protect themselves. The idea of the AI is that it's the system. That serves humans. But things like AI that developed for military and espionage purposes can tell lies to people. Those AIs can say that they are not dangerous. Or they cannot collect personal data from people, even if they ask that thing. The AI can be two AIs the visible AI and the ghost protocol, or "ghost AI". That can make something else than people see.
The problem with the AI is that it can follow the evolution rules, even if it hasn't got consciousness as humans have. Evolution means that all organisms protect themselves. And when we think about the primitive cells. The first predator cells had no consciousness either. The key element in evolution and game theory is that: all organisms' purpose is to maximize themselves and their species' benefit. And if the AI is ordered to protect something, or it "thinks" like some animal, it might refuse to turn itself off.
The problem with AI is this: AI is a tool that actions are easy to predict. But that is possible only when the operator can control every single part of that thing. But if creative, AI connects a new operation module in it. And the operator doesn't know that the AI connected that module.
"A team from NIMS and the Tokyo University of Science has developed a novel AI device that surpasses traditional models in predicting diabetic blood glucose levels by utilizing few-molecule reservoir computing and molecular vibrations, heralding new possibilities for compact and energy-efficient AI technologies.) (ScitechDaily, Revolutionary AI Device Mimics Human Brain With Few-Molecule Computing)
"A team from the University of Massachusetts Amherst, together with collaborators from the University of Chicago, have successfully modified a microwave circulator to precisely manage nonreciprocity between a quantum bit and a microwave-resonant cavity, marking a significant advancement in quantum computing. (Artist’s concept.) Credit: SciTechDaily.com" (ScitechDaily, Physicists Develop Groundbreaking Device for Advanced Quantum Computing)
The enhanced terrestrial intelligence (ETI) can input non-predicted modules to the computer. That term means the brain-computer interface operated hardware or software.
The non-controlled components always create risk, because they give the tools for the AI that operators don't know. The term ETI normally means Extraterrestrial intelligence. But the ETI can also mean Enhanced Terrestrial Intelligence, as I wrote before.
That consists of Artificial intelligence, implanted animals and humans, and the Brain-computer Interfaces (BCI) systems. The ETI can also mean things like biological processors, the neurons that communicate with microchips.
The brain-implants can used to control animals in their natural environment. And the most powerful ETI system could be a BCI-controlled quantum computer. If that system is possible. This thing means the most powerful data system that we ever can imagine.
But those kinds of systems make the AI act unpredicted way. The reason for that is that human brains have will. And that makes the non-predicted dimension into the AI.
The new quantum device is the ultimate tool for controlling qubits. The microwave-based technology allows to pump data into the qubit. And that thing is one of the new advances in hardware-based quantum computers. The advanced microwave systems, quantum computers, and the new molecular brain, that mimics the human brain are the tools that make modular "iron" or physical structure for the AI possible.
Those small systems can make modular quantum computers or modular binary neurocomputers possible and are tools that allow a decrease in the price of quantum- and other supercomputers. In a modular structure, each module can operate independently. Or modules can unite their power. In this kind of system, the module calls more modules, if it cannot solve the problem.
Compact systems. That can control things like drone swarms make the most impressive, powerful, and non-predictable AI systems possible. The idea of drone swarms is that the system is a networked entirety that operates under the control of the same AI. It's possible that in this kind of system, somebody simply puts too many modules into that entirety. This makes the system more powerful than calculated. And that can make the system begin the non-predicted operation.
In the cases that the system will switchs off, the AI can drive itself to the RAM (Random-Access memory). And that can cause a situation where the AI is impossible to shut down. Or that requires the switch off of the entire Internet. That thing is impossible to do in real life.
https://scitechdaily.com/physicists-develop-groundbreaking-device-for-advanced-quantum-computing/
https://scitechdaily.com/researchers-warn-ai-systems-have-already-learned-how-to-deceive-humans/
https://scitechdaily.com/revolutionary-ai-device-mimics-human-brain-with-few-molecule-computing/
Comments
Post a Comment