Answer:
To AI, an image is simply a matrix of numbers, it does not actually understand what the image displays. An image consists of rows and columns of pixels and each pixel has a colour value which is typically represented by three bytes (or 24 bits) – one byte each for the intensities of red, green and blue in the pixel. A small image of 100 by 100 pixels would already contain 10,000 pixels that are represented as 30,000 bytes (or 240,000 bits, each being a 1 or 0). To create a more efficient representation that isn't quite so big, the pixels are typically turned into grey-scale images. Each pixel is therefore represented by a single number between 0 and 255, which is the range one byte (8 bits) can represent. An AI system for image analysis – a so called 'convolutional neural network' – then looks for patterns in the image to classify it. For more information on this, take a look at the following site: Convolution Neural Network for Image Processing.

Answer:
Although this is a frequent topic of Hollywood movies, it is actually highly unlikely that AI could take over humanity. This is a common myth amongst many other AI myths and you can read more about them here: The Top Myths About Advanced AI. Although AI can go wrong and cause problems if it is not built/controlled properly, for it to take over humanity it would first have to become much smarter. This is also called "artificial general intelligence" or AGI, and means that an AI system has human-level intelligence. Current AI is far from that and only qualifies as "narrow AI", which means that the AI system can only focus on one very specific problem e.g. recognise images or answer specific questions. Nobody so far has made any real progress towards AGI and most researchers agree that we are decades away from it. That doesn't mean we shouldn't be prepared and make sure that all AI we produce is benevolent and aligned with our values.

Answer:
In principle there is no limit to what AI might be able to help us with in the future, but in the immediate term AI is more likely to help us with predictable and repetitive tasks in a controlled environment. That is to say that AI works best when we can anticipate what is likely to happen. So the types of jobs are those in which there's a lot of drudgery: trawling through legal cases for key information, routine administration and various manual but well-defined tasks. AI is much less likely to be able to help us in jobs where there is a lot of creativity or where the environment is very uncertain.

Answer:
The most valuable thing to help build trust in AI is to be able to explain how an AI system has come to a particular decision. For example, if you've applied for a loan and the computer says no, then you need to be able to understand exactly why that decision has been reached, and to be able to trace through the steps and the information to make sure it's understandable and correct.