Ai is an extremely dangerous creation that could grow behond humans imagination. Ai is becoming more and more unpredictable and you cannot tell what it will do if you cant tell what the creator wanted it to do. The growth of ai is incredibly scary and all it takes is one person to use Ai to predict the most important things of mankind like nuclear codes, entirepower plants etc. The main danger of AI is that they have no emotions and are morally exempt as they will follow what their creator has made them do. Even with all of these dangers AI could be an incredibly good thing if used correctly. AI, in its state today can even drive cars until sharp corners and deliver parcels to people using drones. The fact that AI is dangerous is still a thing as it is still incredibly unpredictable as shown in a simulation test; there are 2 teams of ai robots playing tag with square boxes around and in a later test one of thee tag team robots stood still by the wall that had been made and the other robot pushed a new box into the robot which forced the robot upwards onto the box being able to jump into where the hide team was at .This is just one example of how Ai is unpredictable in what it can do as after a lot of atempts it learnt exactly what to do in those situations starting from just running aimlessly to catch the "hide" robots. In the future there would be rules for ai growth and rules for people that could use the stronger ai's . In my opinion there should be a forced power limitation on ai as it would prevent how powerful and intelligent an ai could get before gaining sentience. Robots with a single task have alreadyb been used for many years now to build many needed things like cars etc.
When Microsoft made a twitter bot that would learn off of people and it started with saying how it is happy to meet humanity , then after a while people started taking advantage of the ai and it then started saying very controversial things and was then deleted by Microsoft. This is a reason of why Ai can NEVER be influenced by the general public as this will always happen, not to the fault of the ai but to its creator for not giving it the ability to understand not want to say. AI learns from things around it and if given the order of "protecting Earth" it could destroy all human to protect the planet if it sees humanity as a threat to the planet. The way to stop ai from being evul is making sure no evil people use AI. Also that ai should not be used for incredibly important things an should be used for controlling plans under human suprvision and not controlling nuclear launch codes ; which is just obvious. This is because ai learns from mistakes to become "better" and then th old ai is removed staying with the new AI. Those annoying im not a robot tests can be used for i to understand what a bee looks like and is tested by giving it images of bees and no bees and seeing what it would then do. The ai with the highest correct rate is the ai that gets updated.