Research on Machine Learning

Artificial intelligence and machine learning

Machine learning is a branch in computer science that allows the computer the ability to learn without being programmed explicitly. Artificial Intelligence is the broader concept of machines being able to carry out tasks in a way that we would consider“smart”. Machine Learning is a current application of AI based on the idea that we should really just be able to give machines access to data and let them learn for themselves.

Uses machine learning

One of the most popular uses of AI in machine learning, where computers, software, and others perform through cognition (like the human brain).Example of the area where machine learning is heading include the following: –

Virtual personal assistants

Siri, Google Now, Alexa and many more are some of the many popular examples of virtual assistants. As the name says they try to assist in finding information when asked to find something over the voice. All you have to do is ask “what is the weather today?”,” when is the Manchester United playing” or“set an alarm for 3 pm”.

Machine learning is a very important part of a personal assistant for they gather and refine the information on the basis of the previous encounter with them. Later this set of data is used to make results that are specific to your preferences. Virtual assistants are used in a variety of platforms like smart speaker like google home and amazon echo, smartphones like Samsung Bixby and google pixel as google assistant and mobile apps like google also.

Video surveillance

Video surveillance is a very hard task and a boring but with machine learning, it can be an automated process since training the computers they can handle this task.Computers can detect crime just by tracking unusual behavior using machine learning.

Social media

Social media like Facebook, Twitter, Instagram and many others use machine learning to personalize news feed, adds and many more.

On application like the camera the use of face recognition in order to identify people in a specific scene and also identify their faces so as to add effects like smoothing on their face.

Machine learning is also being used in apps like Facebook so as to identify people we may know and suggest that we add them as our friend. Also, apps like Pinterest uses computer vision to find objects in images and suggest similar pins accordingly.

Email spam and malware filtering

Apps like Gmail use machine learning to classify email into primary, social, important and spams. This is with the help of filtering being done under the hood using machine learning. Over 325,000 malware are found every day and each piece of code is 90-98%similar to its previous versions. The security program that is powered by ML understands the pattern for coding.

Online customer support

Microsoft bots are being used to provide chat rooms where people can report about the services they get and this is due to machine learning.similarly when one opens a browser the search is customized for that specific person. For example, YouTube is highly customized for each and every person according to what they like watching thanks to machine learning.

Product recommendations

When you buy something online you start receiving email relating to that product in other stores also when you browse online you see some sites suggesting things to buy which are close to your taste. This is due to machine learning which compiles your likes and taste as you browse the internet combined with an algorithm working under the hood.

Fraud detection

Online fraud detection is among the frontier that machine learning is taking head-on by trying to analyze illegal online transaction and preventing money laundering e.g. PayPal. This is done by using a set of tools that can help compare millions of transactions happening and distinguish between legit and illegal transactions.

Introduction of machine learning to smartphones

The first phone was made by Alexander Graham in 1876 and it became a revolutionary gadget as 1900 approached. The phone was used for basic services like calling and texting at the end of 20th century. As the years progressed the phone morphed from a basic phone to feature phone and later to a smartphone which was introduced in 2000 i.e. Sony Ericson R380. This was a revolutionary idea in its time since it featured a capacitive touchscreen something that has never been seen before on a phone.

As the smartphone momentum started and many companies joined namely Apple, Android and many more. Due to demand for new features, the smartphone industry has tried to outdo one another and in the process, sell more. This has made the companies making the phone to invest hugely in research and development so as to come up with new features. Artificial intelligence has always been a new frontier for the phone but the computing power has always been a constraint. The smartphone computing power cannot be able to be enough to train models which are necessary for the learning process of a artificial intelligence. Training a model entails providing a lot of data to that model until it can recognize a certain data. This is taxing on a smartphone which ishas low computing power and so training is done on a computer workstation and once that is done it is then ported back to the device via tensorflowframework.

Tensorflow

Tensorflow an open source software library for dataflow programming in a wide range of tasks. The main tasks are its application in a neural network which for the base for training data models. Its mainly utilized by Google in machine learning with which it provides through its range of application e.g. Google keyboard which has predictive typing. Tensorflow is a lightweight library which is perfect for smartphones.

In may 2017 google releasedtensorflowlite which main aim is to provide lightweight machine learning android powered smartphones (especially android 8.0 Oreo). The core oftensorflow is programmed in c++

A research design is a blueprint of methods and procedures used in collecting and analyzing variable when conducting a research study. A specific suitable question for study in a research project should be considered and then choose a suitable method of conducting the research. This is important for successful coverage of the highlighted objectives and completion of the research. Research data was gathered through the participant’s observation e.g. use of senses like eyes by examining people in a targeted population. There was also the case of examining earlier records on artificial intelligence from where we have valuable information pertaining to the inception of this technology

Target population involves the people I want to gather information from and in my case, involves any person who owns a smartphone. The features like predictive typing which involves the use of Google keyboard will be an easy task.

It is also called observational study and it is a method of getting evaluative information which entails an evaluator watching his/her subject in his/her place of living and not changing the environment. This type of data collection is used together with other data collection procedures e.g.survey, questionnaires etc.

The main aim of this is to evaluate a happening behavior process, event or when results can be seen. When observing the subject one should not make them aware of your purpose since this can alter the observation and for that reason, the subject should not be aware. There are two types of direct observation i.e. structured direct observation and unstructured direct observation. Structured direct observation is used when we want to get standardized information and result in quantitative data while unstructured direct observation involves looking at natural happenings and get qualitative data.

This involves observing the functions offered by some smartphones that range from facial detection in photos to predictive typing in keyboards. When one uses the google keyboard it learns a person typing patterns and makes out the words a person types this intern stores those words in a database and later reproduces them when one is writing a text of a Facebook post.

Online data collection was among the main research methodology used to come up with this information. Sending online interviews can be a tricky since then evaluator and the participant have never met and this makes sharing private information very hard. Therefore, one must first come with ways for the participant to trust you and accept to share that information

It is estimated that more than 80% of all households in the United States now have computers in their home, and of those, almost 92% have internet access. As computers became more prevalent in American society, the next natural advancement in communication was through the internet. This is rising trend is due to the invention of the smartphones. Smartphone features are the main selling point today i.e. the one who outsmarts the other in terms of providing better features that customer is willing and able for such a smartphone then he/she takes the day.

Apple a tech company is a leading competitor in this field, innovating every year to come with gadgets that impressed everyone and this has created a group of apple royalists.

They are willing to spend more than a1000$ on a smartphone.

The company latest innovation is powered by machine learning to provide security to its flashy iPhone x. The innovation could not have happened without machine learning, no matter how much you code i.e. control statement, methods, API it couldn’t be done. Machine learning is about training the computer to think like a person without explicitly coding it. This would then involve training a model by showing it a lot of data and later it would be able to make guesses about that specific data.g. imagine how can you code a program that distinguishes between apple and oranges, one would argue that you can by saying that apple an apple is between red and maroon and orange is yellow and so you can develop a program that analyses the pixel in those pictures and if the color coincides with the color of yellow.This can be true but what about if the program is fed of black and white photos then the program cannot determine.This is where machine learning comes in and by training a model it can be able to make pretty close guess of a black and white picture of apple and orange since it was trained with a wide range of data e.g. black and white pictures of apple and oranges, shape of apples and many more.

Apple use in machine learning was informed of face id in which the phone would make a mess of your face when you want to unlock it and compare it to the stored mesh in the phone and if it coincides it would unlock the phone. The problem is with face id the person could just hold a picture of your face and the phone would unlock. So they came with system that would project dots on your face and the camera using this dots would be able to make a 3d mesh of your face and using machine learning a lot of your facial characteristics would be taken for comparison and the model would be able to guess it was that actual person close to probability of million to one. This made it possible even to unlock your phone wearing glasses, wearing a lot of makeup or changing ones look by maybe growing a beard.

The possibilities with machine learning are endless as long as the computing power of the smartphone keeps up. Apple had to develop A11 fusion chip (the most powerful chip in a smartphone to this date) to handle such computing requirements needed by the smartphone to train models of one’s face.

The bottom line is that everything related to machine learning and artificial intelligence is tasking to smartphones as of this date, but as companies continue innovating and coming up with better CPU and GPU architecture then this task will run buttery smooth and even open new frontiers in this specific field.

Works Cited

1. Beede, D., Julian, T., Langdon, G., McKittrick, G., Khan, B., & Doms, M. (2018). The AI economy: Insights from Big Tech's latest acquisition spree. U.S. Department of Commerce, Office of the Chief Economist. https://www.esa.doc.gov/sites/default/files/the-ai-economy.pdf
2. Bostrom, N. (2014). Superintelligence: Paths, dangers, strategies. Oxford University Press.
3. LeCun, Y., Bengio, Y., & Hinton, G. (2015). Deep learning. Nature, 521(7553), 436-444. https://doi.org/10.1038/nature14539
4. Li, F. F. (2018). The age of AI: How it will impact jobs, skills, and wages. Brookings Institution. https://www.brookings.edu/research/the-age-of-ai/
5. McCorduck, P. (2004). Machines who think: A personal inquiry into the history and prospects of artificial intelligence. A.K. Peters.
6. Mitchell, M. (2019). Artificial intelligence: A guide for thinking humans. Farrar, Straus and Giroux.
7. Nilsson, N. J. (2010). The quest for artificial intelligence: A history of ideas and achievements. Cambridge University Press.
8. Russell, S., & Norvig, P. (2016). Artificial intelligence: A modern approach (3rd ed.). Pearson.
9. Silver, D., Huang, A., Maddison, C. J., Guez, A., Sifre, L., van den Driessche, G., Schrittwieser, J., Antonoglou, I., Panneershelvam, V., Lanctot, M., Dieleman, S., Grewe, D., Nham, J., Kalchbrenner, N., Sutskever, I., Lillicrap, T., Leach, M., Kavukcuoglu, K., Graepel, T., & Hassabis, D. (2016). Mastering the game of Go with deep neural networks and tree search. Nature, 529(7587), 484-489. https://doi.org/10.1038/nature16961
10. Tegmark, M. (2017). Life 3.0: Being human in the age of artificial intelligence. Vintage.