In a different post we already introduced Machine Learning and shortly explained how it works.
In this post I’ll get into some more details about the applications. The ones which already exist as well as some new ideas.
Existing applications of AI and Machine Learning.
Lately, the areas of speech recognition are:
- Communication with mobile devices (e.g. Apple’s Siri).
- Communicate with Personal Computers.
- Dictation (speech to text) for example as an alternative to typing with a keyboard.
- As a method to give commands to other systems like car audio entertainment systems or wash machines.
- Voice command systems Like Amazons Alexis.
- Speech to text archiving systems – Google voice currently offers VOIP systems which store telephone conversations in gmail mailboxes, which are searchable. (Handy for espionage services like the NSA. And thrust me, if it’s possible it’s also done!).
But there’s more than only speech recognition:
Google recently launched RAISR, a Machine Learning technique which allows images to increase their own resolution. Not by just interpolation! It actually adds new pixels to images to improve their resolution! Think about this. It’s a revolution! Since more information is actually generated from lower resolution (less information)!
I thought about practical applications and came up with the list below:
- Criminology – Criminal cases can be solved quicker and easier since the pictures from surveillance cameras can be improved.
- Healthcare – When it’s possible to improve images, it should also be possible to improve the signal to noise ratio’s like in laboratory equipment, making it possible to detect or diagnose diseases earlier. Another great idea is to make it possible to recognize (complicated) patterns from the output of (for example Mass Spectrometers, analyzing a drop of blood) to detect pointers which could lead to later diseases like Stroke or Heart disease.
- Research – There are literally Petabytes of research data available in scientific research journals and papers. AI could make it possible to extract valuable information from all this data. Because of the large amounts of data, it is practically impossible for humans to oversee everything, let alone to make connections between different studies and their results!
- Prediction of financial markets – That sounds good right? Unfortunately, there are no serious solutions yet. But it’s being worked on. I heard of one company who are selling a product that enables real-time information from information that is 5 minutes behind! If it’s possible to predict 5 minutes then 5 hours, days, weeks and even months should be possible to.
Image recognition and classification
This area of AI, also known as Computer Vision is already in use widely. Google and Apple use it to make photo libraries searchable. When you search for example for ‘dogs’ you will find photos of dogs, without any human intervention to add tags to the photos for example. It’s also applied in biometrics, for example, to give a person access to systems or places by an iris scan. Face recognition is already applied to Smartphones.
Applications which already demonstrated the usefulness of ML and AI.
- This site is protected by reCaptcha. An AI solution to prevent brute force attacks. It can recognize if a login attempt is made by a person or (Computer) program (also called robot). Unfortunately hackers are attempting to gather a community to learn from people solving reCaptcha’s but I doubt if they will ever succeed!
- Just recently Google showed that with the technology of DeepMind and alphait was possible for a computer to defeat the world champion of the Chinese board game Go! This was so spectacular that the prominent journal Nature devoted a large article to this!