Artificial Intelligence (AI) and its subset of Machine Learning (ML) have become intrinsic parts of our daily lives. Some of us might not realise it but ML algorithms are almost everywhere around us – in our banking, investment or mobile healthcare apps, within our home smart sensors systems, video surveillance. They run in the background while using our social media to recommend us exciting content and promote relevant ads based on personal preferences.
Working in a bespoke software development company, I know AI and ML will be integral for our globalised and digitised society. It depends on future-oriented companies to build innovative software solutions that create value and ultimately change our lives for the better. As markets shift towards data-driven products and services, end-consumers become more demanding in terms of transparency and personalisation. That is why future ML applications should be both practically useful and respect the privacy aspect.
What Exactly is Machine Learning?
Fundamentally, machine learning algorithms are capable of learning from data based on input and the desired outcome. In the MIT Press book “Deep Learning”, Goodfellow et al. give the classic example to illustrate what it means when an algorithm learns something. Essentially, a computer program learns from experience E regarding task classes T and predefined performance indicators P if the algorithm’s task performance in T, measured by P, improves with E. The learning process doesn’t consist of simply performing the task well but attaining the ability to perform it and improve over time.
The two core types of ML are supervised and unsupervised learning. The first one requires an instructor who shows the machine-specific tasks and defines datasets with a label and target (data is tagged). Examples of supervised learning are simple pattern recognition, text classifications, weather predictions, database marketing etc. Whereas supervised learning needs a supervisor, unsupervised learning extracts patterns from not previously tagged data via methods like clustering, anomaly detection or neural networks.
Now, let’s explore five use cases of unusual ML applications that we are likely to see soon.
Food Processing Assistants
In the near future, we might eat high-quality food thanks to machine learning. Today some of the most innovation-driven companies trust ML algorithms in each step of food production to increase product quality, eliminate waste and create sustainability. Currency, the world undergoes its Fourth Industrial Revolution, or Industry 4.0, which ultimately transforms the manufacturing process, bringing more product value and reduced costs. All this is possible thanks to new technologies such as IoT (Internet of Things), AI, 3D printing etc.
The fast-paced AI adoption rates promise to improve overall food production faster than one might imagine. Every single potato or pear has to be adequately examined before it can be allowed to pass the initial selection process. Then, following strict safety and cleanliness guidelines, the yield moves forward into packaging. Machine learning can help companies during all those vital steps while improving quality and boosting their revenues.
Fake News Generators and Detectors
Unfortunately, not everyone designs AI with pure intentions. Generating fake news and distributing misinformation is rather simple using ML capabilities. Modern algorithms can compose texts that deliberately contribute to mass disinformation within seconds. These can be used for various purposes, such as suspicious sales strategies focused on misleading potential users or political propaganda via social media.
While the bad guys try to deceive their audiences, there are good guy chasing after them and trying to restoring our faith in information technology and social media. Fake news detectors try to tackle the problem by automatically classifying and reporting online content as misleading or false. Although this task can be challenging, this use case of ML seems promising as it uses advanced methods such as multilayer perceptron (aka vanilla neural networks), random forest model, linear SVM and more.
AI-B(i)ased Beauty Ranking
Beauty is attractive and sells well. Some people claim that beautiful people reach success in life more easily than those who look ordinary. The famous saying that beauty is in the eye of the beholder exposes the subjectivity related to beauty. What happens, though, when the beholder is not a real person but a machine?
New generation AI-powered apps can now rank the attractiveness of a person. Users just need to pose without makeup, take a photo of their faces and then wait for the AI verdict to tell them if they are beautiful. Elaborate solutions can even predict flaws such as wrinkles, dark circles under the eyes etc. Just as people are equipped with prejudices based on physical appearance or social rank, machines can exhibit biased behaviours as well. Future efforts will require software developers to design ML algorithms that diminish their tendency towards innate biases.
Sports Analytics and Injury Predictions
Deep learning algorithms can be useful in sports for developing team and match strategies or special training regimes that fit the uniques needs of team members. Sports aficionados around the world keen on sports betting place their bets on online platforms based on coefficients with robust ML algorithms. Sport video games such as boxing can act as a useful source to obtain data so that machines can learn punching techniques and optimise them.
In their book on ML and data mining in sports, Ulf Brefeld, professor for machine learning and his colleagues explain how sports analytics can be used in soccer, hockey, football. Individual sports such as Tour de France contestants or cyclists. Some of the most promising use cases of ML is going to be a prediction of potential injuries and prevent them. For instance, limb traction movements with ML sensors help track the direction of body movements and suggest pattern variations.
Brewing Tasty Beers
Another scenario where AI and ML can become especially relevant is beer brewing. Today, there are individual brewmasters who rely on AI to produce personalised brews. As craft beers and their enthusiastic producers gain global popularity, this application of AI seems to open new possibilities.
These algorithms can provide insightful suggestions that lead to better decisions that increase sales and revenue. For example, they can organise data into best- and worse selling products, come up with recommendations for recipe adjustments or be involved in the creative process of creating marketing campaigns.
Biography Aleksandrina Vasileva
Aleksandrina is a Content Creator at Dreamix, a custom software development company, and is keen on innovative technological solutions with a positive impact on our world. Her teaching background, mixed with interests in psychology, drives her to share knowledge. She is an avid reader and enthusiastic blogger, always looking for the next inspiration.