How has artificial intelligence influenced society?
The development of artificial intelligence (AI) is rapidly changing the way we live and work. With the help of AI, we can perform tasks previously thought impossible, such as understanding natural language and recognizing faces. However, as AI technology evolves, there is growing concerned about its potential impact on society. One of the main concerns is that AI could lead to mass unemployment, as machines could perform many of the tasks currently performed by human workers. Another concern is that AI could create “superhumans” who would be much smarter than the average person, widening the gap between the haves and have-nots.
The capability and scope of AIs are increasing daily. For example, the above write-up was generated by GPT-3, an AI capable of generating text from a statement, a question, or a title. In addition, GPT-3 can write articles, conversations, and poetry and even generate code in different programming languages given one instruction. In some cases, the content of the text may not be entirely true. Still, it manages to pass for text written by a person since it is not copied directly from a source. Furthermore, it is possible to set the “temperature” of the writing, which determines how random and creative the resulting text is.
Undoubtedly, the progress made in recent years in the development and training of AI is easily evidenced. It is increasingly common to hear news about how AIs tackle new areas, surprising with their results. Gone are the days when AIs were imagined as software for purely business use. Then, it was necessary to have large infrastructures or in-depth knowledge of programming languages. Now, everybody can use it as easily as they use Google searches.
Creating illustrations is a skill that not everyone has. However, thanks to an illustration AI such as DALL-E 2 or Stable Diffusion, generating an image very close to what you can imagine is possible. But would this image be a work of art? Would it be original? Clearly, yes, because it was generated then, and there is no other work.
These questions and many others, such as: does this make me an artist, can arise to a simple user exploring this new AI, but from the position of a conventional artist or a graphic designer, this new option of instant art can be interpreted as a risk factor for their job stability, now that anyone can get a job of equal or better quality than the one, they do, to which they dedicate several hours to complete it and for which they charge a sum of money.
These artworks can be generated in a matter of seconds, even for free. Won’t they affect the number of requests a professional can receive? Perhaps it would still be requested by those who want a personal touch to their work. Still, nothing prevents AI with enough training to replicate this “personal touch” that makes the artist unique.
Many professionals have concerns about how their work may be reduced to a simple entry in a description box. It would not be the first time, nor the last, that technological advances replace workers in their usual tasks. However, there is the possibility that new professions and jobs, such as “Input Designer in DALL-E 2”, will be born, showing that society can accept these new tools and adopt them daily.
Our standard technology has been integrated with different artificial intelligence for several years. However, in some cases, we are unaware of the processing that can occur in the background of the processes. In addition, some of the environments of use have a low margin of error, as is the case in medical analysis or the operation of transport vehicles.
In case of an accident, who would be liable? A malfunction of the AI could be alleged, and the manufacturer would have to take action, or the accident would be declared a consequence of AI misuse. At the legal level, it is not stipulated that an AI can be blamed in any situation or in the case of AI fraud or tampering offenses. It may be simple at this time to define whether a problem is due to the failure of an AI and to assign ethical and legal responsibilities straightforwardly. Still, in a future where the line separating the decisions of the AI, the operator, or the designer will be less noticeable, it will become more complex to identify the intentions of each.
When the same AI can be used in different fields, the other uses to which it can be applied grow exponentially, as is the case of Gato, an AI developed by DeepMind, capable of performing tasks of varying nature, such as playing video games with a visual input or moving a robotic arm to manipulate elements. It is also capable of holding conversations with a given text. Gato has been trained to contemplate different types of inputs and process them equally to proceed with the set task. In this way, Gato can easily be trained for new tasks with different types of inputs in a centralized model. It would be less complex to train this type of general AI for a new task such as sound recognition than for an AI designed purely for text processing, which is a step beyond purpose-built AIs.
Cases like Gato’s come a little closer to the term Artificial General Intelligence, where an AI can perform any task as well as, or better than, an average human would. Of course, this term is still a milestone to be reached, but the evidence and the speed of progress hint at the possibility of achieving an AGI.
There are still many challenges we face when trying to achieve a definitive intelligence, as is the case of the difficulty we have when facing cognitive frameworks. For example, the ability of a system to identify how relevant a piece of information is to a specific situation becomes very complex when trained in different contexts since interpreting the results of a task when you have to consider contexts not directly related to that task is a job that increases exponentially, for example, when asking for a result in the text if the AI has to consider the quality of the result. Moreover, if it generates a sound or a movement, it increases the processing of the task almost infinitely.
Is a society with all its shortcomings and imperfections possible to aspire to a cognitive and general mind that can function flawlessly? Suppose AIs are thought of as trying to mimic the same human processes. How do we avoid repeating the same mistakes that do not allow us to fulfill a fair and just society? If the data used in training is biased, then AIs will also be biased. For example, suppose artificial intelligence is learning to recognize faces and only shows faces of a specific race. In that case, it is likely to acknowledge that race more easily and not know how to identify the characteristic features of other races. Designers are responsible for avoiding such biases and not sharing with AIs all our shortcomings as a society.
The way ahead is still long, but there is no doubt that the speed at which AI design and training processes are being improved is improving daily. Their personal, social, and business uses are becoming more and more evident, and we will reach a point where they will become indispensable for the technology we currently use to meet our expectations. This same intelligence may enable a better society for all and give us the tools to evolve more creatively and freely. They may allow us to be more than just automatons.
Written by Juan Alfonso Castro Flechas SENIOR BPS DEVELOPER