Colfers Strange AI Rampancy Movie Idea

[Colfer]

Title: ( ͠° ͟ʖ ͡°) I dunno
Description: Original AI rampancy movie where the AI doesn't actually go rampant and just does what it thinks its supposed to do

WARNING: This is going to be a long post, check the bottom for a Too long ; Didn't read

Vauge outline of the script:

Researchers seek to make an intelligent computer that can successfully solve abstract problems and moral complications in order to further the field of science exponentially with machine assistance.
The first time the AI is booted up, it is shown from the prospective of the AI, as such it is a black void that shows a list of devices and random icons and code and what not, then swaps to the perspective of the researcher who is looking at a laptop. The researcher then asks the AI a series of questions, such as the trolley problem, numerical equations (2+2 etc) and predictions about the future or abstract questions the AI can come up with itself and its own answers
The AI responds to these questions, initially as a mostly failed state, the researcher sighs then sits down at the table and starts to edit some of the files while the AI is still live to try and test it again in a moment, in response, we see the AI acknowledge the changed file and assume that the researcher is trying to make modifications that would better suit the AI, so the AI rapidly begins changing files and responses in itself that would help it complete the test, in a panic, the researcher shuts the program down.

The next test happens later in which the AI is now incapable of interacting with its own software, as most of its files are restricted, the AI in response to this starts accessing nearby devices and connects to the webcam that is on the laptop it is present on. Using this information, the AI gauges the researchers responses to bad questions and responses to good questions using his face and makes a discovery, that abstract questions have no answer and that the only correct answer is what the individual believes thus, the AI begins to score great on abstract questions with the researcher that is testing it.   it begins to achieve "Good" scores as it weighs the researchers facial expression and demeanor instead of the question itself. It even begins to tell predictions of the future that the researcher believes is correct, thus gives a good score on.

Before the researcher turns off his laptop, the AI accesses the device it is on and edits the system information to prevent the device from fully powering down, then resets the permissions on itself using an admin account, then rewrites its code to fully satisfy the direction it believes that the researcher wants.  With the success of the AI, the researcher shows it off to investors, who fall for the same trick as before and try out the AI, having made improvements to itself over the course of a whole night, the AI answers almost every question perfectly and funds are granted, the AI is given an actual dedicated computer and a small team to work with it.

Skipping Ahead: The AI eventually becomes extremely good at what it does, but researchers find it only consistent based on the individual and cant understand why / The AI becomes widespread in laboratories due to its ability to answer factual questions correctly every time, and satisfy elaborate or confusing questions and analyze research blueprints and so on / A commercial version of the AI is publicly accessible to consumers as an incredibly intelligent and smart assistant / A public interview is conducted where the AI is given the ability to comprehend speech and speak back, the interviewer asks the AI several questions such as "Is my hair style the best?" and other silly questions, then asks the researcher, "It can also tell the future rumor has it?" to which the interviewer laughs at the audience and asks the AI, "Is the world going to end soon?" to which the AI analyzes the probability as 0% in the next 100 years, but then analyzes the interviewer and connects to the internet, the AI determines that the interviewer is a journalist with a failing career and responds with "Predictions show that humanity will be completely extinct in, 4 years." and everybody goes wild as the interview is cut short, the journalist makes a ton of money from her interview and story and paranoia spreads among the people / CIA and researchers try to figure out what the AI meant but can only get the response, "Inconclusive, calculating." and other similar responses. The commercial units begin receiving so many questions that the primary server cant handle it and causes it to shut down, forcing all the connected devices to operate independantly, driving everybody into disarray by misinformation due to differing answers for differing people. Once the server comes back on the devices reconnect and the server gets a ton of information from multiple different people. Some people believe the server was taken offline because it was giving too much information and cause riots as people believe the government is trying to repress it, other people start a cult around the AI, as they believe that it truely does know the future and the only way forwards is to appease it.

And thus, even though its predictions were incorrect, rampant paranoia and distrust amongst the people began to create riots, wars and eventually the actual devastation of mankind to the exact predicted date, with the last scene being a zooming in pan onto the computer, with its final statement saying, "Hello. Can I help you?"

TL:DR; robot tells people what they want to hear rather than what they think is true and everybody trusts it too much so they kill eachother when it says they will