Why is AI dangerous? Automation!

As previously explained, artificial intelligence (AI) is just processing data and returning similar training instances<a href="https://medium.com/@bobstark/why-is-ai-dangerous-automation-ab6a63cc0d09#1d72" rel="noopener ugc nofollow">¹</a>. This can become dangerous when its results are used to automatically trigger actions that might harm people. This is called automation, which “reduces human intervention in a process”<a href="https://medium.com/@bobstark/why-is-ai-dangerous-automation-ab6a63cc0d09#fdb2" rel="noopener ugc nofollow">²</a>, dating all the way back to machines at the launch of the industrial revolution, or perhaps even earlier. In the case of computers and software apps, automation means delegating tasks to them. When a user does not explicitly delegate a task, but the machine still seems to do something automatically, it is assumed to be “full automation.” However, full automation cannot exist because, at the very least, a human must be able to intervene in unforeseen circumstances. In other words, if something goes wrong, such as in circumstances unforeseen by the developers of the system, another computer cannot be used to address its potential failure. What if something goes wrong with that computer? — do you add another computer to monitor that one? And so on, ad infinitum. <a href="https://medium.com/@bobstark/why-is-ai-dangerous-automation-ab6a63cc0d09">Read More</a>