This blog post is about the explorations of different Artificial intelligence is a machine or program with the ability to process and respond to inputs from the surroundings by itself without further instructions.

Teachable Machine

This experiment is about teaching a program to recognise different images to produce different outputs whether it may be pictures or sounds. This is definitely AI as it trains the program using supervised learning by pressing a button to allow the program to take pictures of you doing an action which will train the program to recognize you doing an action to connect it to the picture or sound that is output. The limitations of it are that it is only as good as how many pictures you take for each action, the more pictures taken, the better it is for the program to recognise it. Future implications can be used in everyday technology such as doing an action to switch on the TV or switch on the lights, this technology is being tested in cars now but is not very smooth.

AI Duet

This experiment is for anyone and allows people to enter any notes into the program using the online computer, the keyboard or a MIDI input keyboard. This will then allow the program to respond with a melody of its own which can be of similar rhythm or key. This is AI because its programmer used unsupervised learning for this programmer by inputting the different phrases and letting the program itself learn how to respond using its neural networks. The limitations I feel for this experiment are that when I play with it, it seems like it is playing random phrases, thus how much it actually learns from the inputs limits the experiment. Future implications of this program can be for the composition of music or even jazz improvisation.

Giorgio Cam

This is the most entertaining experiment of the group. It is based on electronic dance music and takes images and tries to output the lyrics in what it recognises in the image. Every time an image is processed by it, it would give a percentage of rating it thinks it is the object it has matched it with, with each time getting better. This is AI and is similar to the Teachable Machine in the sense that it gets better with each recognition but I feel this uses a type of unsupervised learning called visualisation as it compares with images of similar items and puts the percentage of certainty on it. The limitations of this are that it only outputs the most recognised thing it processes so sometimes there were repeats of the same item. Future implications can be used for security features, for example, used in security cameras to recognise weapons that people are carrying or used as overall face recognition.

Thing Translator

This experiment allows users to take a picture of any item and it will translate it to a given language. This is also an example of AI processing images and matching them to items and then finding the translation of it in the database. The limitations of this experiment are how well it recognises the image and whether the translation of it is correct. Future implications can be used for tourists or for educational purposes if they want to translate something by using images.

 

If you enjoyed this post, make sure you subscribe to my RSS feed!