Artificial Intelligence: Family Pet or Future Owners?

This image is an example of how the media chooses to dramatize the implications of artificial intelligence purely for views without solid justification. Credit ~ BBC News

“Success in creating AI would be the biggest event in human history. Unfortunately, it might also be the last, unless we learn how to avoid the risks.” — Stephen Hawking, Theoretical Physicist

Everyone’s heard about the conflict. How AI will dramatically make society more efficient and capable in a relatively short period of time or how humans will become pets at the will of such machines. How either humans could finally achieve utopia, or how we could be forced to survive during machine invasions. The simple answer to this dilemma: no. The accurate answer: nobody knows besides our imagination.

For all the people too constricted for time to read, here’s a brief explanation of the article’s contents.

The exact future of humanity lies within our own decision. That’s been well established from ancient times. However, in a rapidly transforming world, the impact of our day-to-day decisions has seemingly gained in capability. One of the driving technologies that continue to push humanity’s dominance and efficiency ever further was and still is artificial intelligence. Due to its breathtaking achievements along with its prominent portrayal as a menace in the media, stereotypical and unsustainable misconceptions have swarmed the industry. Most frequently, people tend to question whether AI will positively impact human society once it crosses a degree of “intelligence”.

The answer to this question is a no; at least for now. The question is not invalid, but rather uncertain as there remains a lack of reliable information regarding the development of an intelligent model. There are widespread disagreements on whether such a model is even possible, and the degree of intelligence the artificial intelligence would be able to achieve (Artificial Narrow Intelligence / Artificial General Intelligence / Artificial Super Intelligence). The dissent in this conversation includes but isn’t limited to sub-discords regarding the speed of the intelligence’s development, whether any models at the moment are capable of iterating their own intelligence, and the time-frame of AI’s development, amongst various other pertinent sub-discussions.

Credit ~ Great Learning

However, the long-term uncertainty over artificial intelligence’s impact on human society should not restrict its indisputable efficacious impact on nearly every aspect of humanity at the moment. Artificial intelligence has brought a combination of complex mathematical models with incredulous speed thanks to the newfound computational power of machines. This helps create machine learning models capable of doing a wide variety of tasks, with various applications such as predictive models, decisive models, and data synthesis models. In conclusion, the long-term uncertainty should not limit AI’s contemporary implementation, due to the lack of presiding evidence of any potential outcome.

Credit ~ Adava University

For example, reinforcement learning - one of the most cutting-edge machine learning models - has countless modern applications in fields ranging from economics to medicine. Reinforcement learning utilizes exploration of an agent’s environment in order to collect data on the state and the correlated behavior of the agent. Through repetition and penalization, the agent is able to formulate behaviors that lead to the most substantial amount of reward. This can be applied to make trading algorithms that through simulation determine ideal behavioral choices to gain the most monetary value. Other applications include identifying various diseases, trajectory optimizations, dynamic pathing, policy analysis, and motion planning amongst various other interpretations limited by the programmer’s imagination!

Although these examples are simplistic introductions, a common decision for reinforcement learning algorithms is their risk tendency - to explore new actions versus exploiting ones that have been proven (although these explored decisions may not be the most effective). Credit ~ Adava University

Another breakthrough machine learning model has been GAN. GAN stands for Generative Adversarial Networks, which have quite a niche, yet puissant utilizations nonetheless. The model works through a random input vector being inserted into a generator model. Based on this input vector (most notably multivariate Gaussian), the generator model creates a fake image. Through collected data samples in real life, both the fake, generated images and the real images are inserted into a discriminator model. The discriminator model then tries to determine the fake and real images accurately. This results in binary classification for the set of fake and real images. Both the generator and discriminator models attempt to modify themselves based on their accuracy/success. Therefore, through self-improvement caused by competition between these models, the generator model learns to create realistic images, whereas the discriminator model learns to identify minor details between the sets of images. Such a complex model could be described as a competition between two machine learning models. GANs help to create hyper-realistic images, videos, along with other formats of media (sound and text most notably).

The structure of a GAN is quite novel in comparison to previously established machine learning models, as it works through competition between two distinct machine learning models so both can improve their skills! Credit ~ Adava University

Although the example models mentioned above are quite powerful and at the headlines in the machine learning niche, they’re quite distant from humanoid machines capable of destroying modern society (if such artificial intelligence is even possible). The current state of machine learning is known as ANI or Artificial Narrow Intelligence. For reference, humans would be comparable to artificial general intelligence. These ANIs are capable of mastering narrow tasks with relatively specific criteria. When given broader environments without the necessary training, the models attempt to utilize intelligence not applicable to the current situation. For example, artificial intelligence that can play chess, could not play tic-tac-toe without the necessary training. This is the key difference. AI is not yet capable of applying generic knowledge to gain narrow knowledge.

One of the recent breakthroughs in the AI industry was CNN (convolutional neural networks). They work radically for interactions with different formats of media such as images, sounds, videos, text, amongst various others. Credit ~ ResearchGate

Personally, I feel that the development of AI should continue in a more structured manner. As per experts, if such general intelligence could be created, it may be possible for it to through repetitive learning gain super-human intelligence. Such timeframes are expected to be exponential in general. However, the immediate benefits of artificial intelligence are too paramount for ignorance, and therefore, I believe we should supervise AI development. By that, I mean official legislation should be enacted demanding that artificial intelligence breakthroughs be conducted in a documented manner to enable faster response times in case such devastating events occur. This structure would enable more structured research and developments, along with more connections between international leaders in artificial intelligence.

Throughout history, humanity has always been paranoid of the unexplored. A homogeneous scenario occurred with nuclear power several decades ago. Although the field of nuclear power has not been explored to the extreme depths yet, it’s a more experienced subtopic compared to artificial intelligence. We were paranoid of the uncertainties regarding its human impact, and the risks it posed to the surrounding regions. Without exploration; however, we could never have been able to learn more about the depths of physics, along with 10% of the world’s electricity! Another example would be the industrial revolution. At the time, it was considered that these machines would increase unemployment, yet instead, it’s lead to rampant development along with an increase in employment (by creating new industries for employment). These scenarios weren’t perfect, nothing is. The industrial revolution led to global warming, and nuclear power has resulted in several casualties. However, compared to their positive impact, I feel we cannot ignore the incredible opportunity artificial intelligence provides to us.

Credit ~ Adobe Stock

As with all stereotypes and over-generalizations, the only way to become more informed citizens is by looking at multiple sides of the story. Rather than accepting the unjustified claims/opinions in the media, look to inform yourself through expert opinions, cutting-edge research, or reliable sources. Look at various inputs of the future of artificial intelligence, and their respective proposals to how development in the industry should continue. By looking at various sides of this germane story, as readers, you gain a better appreciation of where the future of mankind lies and can make informed decisions for yourself.

As to the future of AI, I feel it’s quite paramount that we make a premature decision rather than stay indecisive. If humanity continues to tread on multiple opposing paths, the only conceivable outcome is various forms of conflict. Therefore, regardless of whether humanity chooses to oppose or stand with artificial intelligence, it’s quite cardinal that we stick with the decision, to at least have a marginal chance of constructive impact.

Further References to Learn More About the Future of Machine Learning:

Best-selling Books

Credit ~ Amazon.in and Max Tegmark

Extensive Review: https://www.yalescientific.org/2018/01/life-3-0-review-being-human-in-the-age-of-artificial-intelligence/

Life 3.0 is an extremely well-written book that serves as a comprehensive guide to various aspects (moral implications, consciousness, crime, justice system, impact, development amongst various other determinants of AI) of the future of society in a world surrounded by various forms of artificial intelligence. The book lists out various theories and outcomes depending on several factors of the development of artificial intelligence. The book is written in quite an unbiased manner to enable readers to form their own opinions, in order to contribute to this appurtenant discussion regarding artificial intelligence.

Other Notable Must-Reads:

Credit ~ Amazon.in and Nick Bostrom

Extensive Review: https://medium.com/@rossrco/book-review-superintelligence-paths-dangers-strategies-by-nick-bostrom-19675475d31f

Credit ~ Amazon.in and Pedro Domingos

Extensive Review: https://insidebigdata.com/2016/01/12/book-review-master-algorithm/

AI Laboratories

Credit ~ OneZero Medium — “OpenAI is an AI research and deployment company. Our mission is to ensure that artificial general intelligence benefits all of humanity.” — The Official Website of Open AI

Extensive Review: https://www.technologyreview.com/2020/02/17/844721/ai-openai-moonshot-elon-musk-sam-altman-greg-brockman-messy-secretive-reality/

OpenAI provides both an introduction to artificial intelligence to beginners, as well as cutting-edge models that delve into the depths of the substantial impact artificial intelligence could have. It provides its readers with a conceptual understanding of the latest aspects of artificial intelligence but also helps to orient the general public of what a world with AI could look like. A non-profit organization, OpenAI provides its followers with a wide range of media and interactions to enable maximum constructive and informative discussions regarding artificial intelligence (images, videos, blogs, articles, publications, API, code, research).

Other Notable AI Laboratory:

Credit ~ VentureBeat — “Our long-term aim is to solve intelligence, developing more general and capable problem-solving systems, known as artificial general intelligence (AGI). Guided by safety and ethics, this invention could help society find answers to some of the world’s most pressing and fundamental scientific challenges.” — The Official Website of DeepMind.

Credit ~ ETCentric — “We’re advancing the state-of-the-art in artificial intelligence through fundamental and applied research in open collaboration with the community.” — FAIR’s official Facebook page.

Further Topics to Explore:

• Neural Turing Machine — https://towardsdatascience.com/tagged/neural-turing-machine

A Neural Turing machine (NTMs) is a recurrent neural network model of a Turing machine. The approach was published by Alex Graves et al. in 2014. NTMs combine the fuzzy pattern matching capabilities of neural networks with the algorithmic power of programmable computers. — Wikipedia NTM

Neural Turing Machines is one of the main breakthroughs recently made regarding artificial general intelligence. NTMs work by enabling machines to create their own fundamental interactions with the memory of the computer, rather than pre-defined abilities (such as intelligently reading and writing to the memory).