There are many new ways AI is shaping the future for better and for worse. While I have been talking about the dangers of AI, I have not yet explained how one of the newest leaps in technology called deep learning works.
We have looked at machine learning before, deep learning is a subset, where algorithms in artificial neural networks learn from vast amounts of data. Artificial neural networks mimic the structure of the human brain intending to summarize complex information into tangible results. The advantage of these networks is the profound abstraction of relations between input data and the abstracted neuron values with the output data. This is done through several layers of the networks, which can solve particular problems (while traditional neural networks only contain 2-3 hidden layers, deep networks can have up to as many as 150). The superordinate name is derived from these facts: Deep Learning. Basically, in trying to copy the human brain, deep learning algorithms repeatedly perform a specific task (mimicking human learning from experience), each time adjusting the outcome a little to improve performance.
When Is Deep Learning Used?
Deep learning is used when other machine learning methods reach their limits. There are two main reasons why it has become so useful in the last years: For one, more and more labeled data is available and has been created (in 2018 estimated at 2.6 quintillion bytes) enabler for deep learning. However, the other enabling reason is much more computational power: high-performance graphics processing units have a parallel architecture for deep learning, which is efficient enough. This, combined with clusters or cloud computing, accelerates in-depth learning network development and training considerably.
"Learn" by Example
Now that we know that there are applications that learn without minimal human intervention let's find out what types of problems these applications have set out to solve.
Virtual assistants (which are mainly female): We all know virtual assistants like Siri, Alexa, or Cortana. These AI applications are based on deep learning algorithms, which help them understand your speech and human language (and also, how to understand the emotion expressed in our pattern of speech).
Translations: Similarly, deep learning algorithms can automatically translate between languages. One of the best translation tools being DeepL.
Vision for driverless delivery trucks, drones, and autonomous cars: Vision, lane-keeping, and driver-in-the-loop simulators are based on models, sophisticated algorithms, and vision processing tools based on deep learning algorithms. However (spoiler alert), a self-driving car speeding along the highway and weaving through traffic has less understanding of what might cause an accident than a child who's just learning to walk.
Chatbots and service bots: Chatbots and administration bots that give client support for various organizations can react in an intelligent and accommodating manner to an expanding measure of sound-related and message addresses on account of deep learning (so maybe we have to tax them?).
Image colorization: Transforming black-and-white images into color was an assignment done carefully by human hand. Today, deep learning algorithms can use the context and objects in the pictures to color them. The outcomes are noteworthy and exact.
Facial recognition: Deep learning is being used for facial recognition not only for (so-called) security purposes but for tagging people on Facebook posts. Soon we might be able to pay for items in a store just by using our faces (which is scary to me - what do you think?). The challenge deep learning algorithms face is not only the fact that they have insanely bad judgment as soon as they have to identify a non-white male person, but also if we change our appearance (say change hair color, etc.).
Medicine and pharmaceuticals: From diseases and tumor conclusions to customized medications made explicitly for a person's genome, deep learning algorithms are shaping the medical field considerably. Furthermore, more and more pharmaceutical and clinical organizations are investing in this area of research. And to keep adding spoiler alerts (for some reason most guides don't look at the downsides of these AI applications): this is terrible news for any woman on this planet, why? Well, the data is screwed. Don't believe me? Even the European Union says so.
Personalized shopping and entertainment: This is straight-forward. Netflix, Amazon, and other shopping/entertainment sites use deep learning algorithms to make suggestions and keep you on the site for longer (the longer you stay on a website, the higher your inclination to buy more is).
If personalized shopping and entertainment is the upside to things, what would the downside to it be? If we look at bias, sexism, and racism in AI, I think it's time to talk ask ourselves if machines can be moral. Sound good?
For this blog post, I have used the following articles and resources:
Furthermore, I suggest the following video on deep learning for more information.