Machine Learning is All Around Us
- Sam L.
- Aug 6, 2022
- 4 min read
Disgruntled, a student rolls out of bed. Their Chinese homework is due today and right now they’ve got a blank page and twenty six minutes. Half asleep they flip through the workbook. “I can’t read this” they mutter and reach for their phone. As it gets a view of their face, it is unlocked. They open their translation app and use the image to text translation to understand the Chinese characters of the questions. Shamefully they dictate their answers in English and again use the app to translate that to Chinese. Bored with the homework they tab off and watch a video in their recommended feed.
Machine learning has infiltrated our lives! Even in this extremely mundane situation, machine learning is used five times! Without this technology we couldn’t unlock our phones with our faces, obtain text from images, convert our speech to text, translate text, or even search the web with anywhere close to the accuracy of modern times. Many subtle areas of our daily routine are littered with machine learning software. Some other examples include Gmail automatically sorting our emails, predictive text, the swiping style of typing on mobile devices, synthetic voices like Siri and Alexa, image storing and sharing services recognizing faces in our photos, and social media recommendation algorithms. Machine learning obviously isn’t just limited to the small conveniences either. Agriculture experts use machine learning and images from satellites to get data on crops and investors use bots to engage with the stock market. Today almost every field and daily activity is somehow assisted by machine learning!
When I hear machine learning I can’t help but picture a Jetson’s style futuristic reality. An instance of our universe hundreds of years down the timeline. Considering this, it makes sense why I was surprised to discover that machine learning goes as far back as the 1950s!? The term machine learning was coined in 1952 by a researcher at IBM named Arthur Samuel. He used it to describe his checkers playing program that adjusted its algorithm to play better over time (effectively learning how to play checkers)! The first artificial neural network was created in 1958 by psychology student Frank Rosenblatt as an imitation of how the brain works. The program was called the Perceptron and it was designed to recognize images. Of course as the first of its kind it was also the worst.
Skipping past several decades of theoretical computer science research and board game championships concluding with machine learning programs outsmarting humanity’s best, we see Google in 2012 use machine learning to identify Cats in YouTube videos with a 74.8% accuracy. From a modern perspective this sounds almost insignificant but it was an important stepping stone in the progress of machine learning. In 2015 Google began to use Machine learning for their search engine. The program is called Google RankBrain and it is now responsible for most of Google’s search results! Also in 2015 Facebook started using machine learning technology they called “DeepFace '' to recognize faces in posts and automatically tag the accounts of those people. The paper on the program from 2014 claims a 97.35% accuracy rate which jumped ahead of older methods by an insane 27%. In fact the accuracy for DeepFace was almost equal to that of a human’s accuracy! By 2021 machine learning has completely infiltrated our lives. From the media we enjoy (Netflix’s ranking system is controlled by machine learning and several social media platforms also utilize it), to the ways we gather information from the internet, machine learning is all around us!
One of the most important things about machine learning is the data sets. In order for a machine to learn to recognize handwritten text or human faces you must have a large amount of digital images of handwritten text or human faces. This is likely the reason why the machine learning field has grown at an exponential rate in the past decade. In current times there are millions of images to parse from the web and with the internet it is much easier to coordinate with others around the world to manually create and organize data sets. We definitely haven’t seen the peak of this area of research. Self driving cars, augmenting human capabilities, more accurate weather forecasting, analyzing MRI scans, there's a lot to look forward to... and a lot to be worried about?
References
Clark, Liat. “Google's Artificial Brain Learns to Find Cat Videos.” Wired, Conde Nast, 26 June 2012, https://www.wired.com/2012/06/google-x-neural-network/.
Marr, Bernard. “A Short History of Machine Learning -- Every Manager Should Read.” Forbes, Forbes Magazine, 8 Mar. 2016, https://www.forbes.com/sites/bernardmarr/2016/02/19/a-short-history-of-machine-learning-every-manager-should-read/?sh=7e067d6d15e7.
Reardon, Sara. “Rise of Robot Radiologists.” Nature News, Nature Publishing Group, 18 Dec. 2019, https://www.nature.com/articles/d41586-019-03847-z.
Schultz, M. G., et al. “Can Deep Learning Beat Numerical Weather Prediction?” Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences, 15 Feb. 2021, https://royalsocietypublishing.org/doi/10.1098/rsta.2020.0097.
Danny Sullivan.“Google Uses Rankbrain for Every Search, Impacts Rankings of ‘Lots’ of Them.” Search Engine Land, 27 Aug. 2021, https://searchengineland.com/google-loves-rankbrain-uses-for-every-search-252526.
Taigman, Yaniv, et al. “Deepface: Closing the Gap to Human-Level Performance in Face Verification.” Facebook Research, 1 Dec. 2016, https://research.fb.com/publications/deepface-closing-the-gap-to-human-level-performance-in-face-verification/
Further Reading!
Comentarios