Robots used to be machines that performs certain functions when given directions. Through a computer, a person is able to give directions for robots to function, robots interacts with the world through sensors and motors. Motors spins objects connected (usually gears, wheels and axle). With more motors connected, the robot can perform more complex movements, but more moving part can also be more parts where error can occur. Problems to be considered are frictions and power of the motor. Sensors are used by robots to detect certain parameters such as humidity, heat and colour. Now robots are more complex than before, through the use of machine learning and big data, robots can recognise certain object with accuracy. This requires large amount of memory and electricity for the robot to iteration upon the original algorithm, the algorithm can often be very complex and difficult to study. Through these new technologies an AI can be created and can be trained to perform certain complex task with proficiency such as play games, create digital art (writing, music and image). In many cases with enough training an AI is capable of outperforming a human, but AI can often only excel in one task and to make the AI excel in other task would require more training and data.
Overall, robots have become more and more advance over time, but so far there is no robot that excels at everything and perhaps this may never occur. Robots could have many applications in real life to assist with a person’s daily life, but there should be a limitation to what a robot can do in the future either physically or through programs. There could be the possibility that robots can overtake many functions in ones’ life and future generations may become more lazier, there may also be the possibility of an AI consuming misinformation or biased opinions which could have an effect on the output.
Through experimenting with Stable Diffusion, I discovered that while the AI is capable of creating great artworks, the AI does not understand how objects interact. This makes AI great at drawing sceneries and backgrounds but bad at details. Another problem is that everything is in one layer, which made editing difficult.
A way to solve this is to prompt background and foreground separately and merge the 2 together afterwards. AI is also not very good at understanding basic human anatomy and by using a roughly drawn but accurate proportion draft in img to img, the process can be speed up and some minor edits can be made to finalize the image.
Kommentare