Facing the Dilemma with AI

0
595

Artificial intelligence has undoubtedly become an integral part of our society as we continue to rely on it daily. AI has made significant changes in society through voice-activated technology, facial recognition features, voice assistants, and recently, ChatGPT. These milestones are a testament to the fact that artificial intelligence will continue to play an important role in future scientific pursuits. However, advances such as these also create dilemmas about to what extent humans rely on technology and the consequences associated with that. Furthermore, many people have varying opinions regarding this matter, and these perspectives have been gathering widespread attention on social media platforms. Regardless of your opinions about AI, one thing for certain is that it is important to consider the role these technological features play in society to grasp the issues that come along with it.

The Beginnings of AI:

Artificial intelligence has vastly changed, for the better, since its beginnings in the 20th century, which heavily focused on theoretical work. One of the earliest works related to AI was done by Alan Turing in 1935 when he described an abstract computing machine with memory and a program of instructions stored in the memory by using specific symbols. In 1948, Turing described some of the early central concepts of AI in a report called “ Intelligent Machinery.” Moreover, Alan Turing also created the Turing Machine, which discovers what can be computed. Turing described it as “a machine capable of a finite set of configurations.” AI experienced further advancements when chess computers were created and popularized. The earliest AI program was written by Christopher Strachey, and this program allowed for a game of checkers to be played. It should also be noted that the first AI program to demonstration started in the United States and was also a checkers game. Arthur Samuel studied the fundamentals of Strachey’s checkers and expanded on the ideas, and added features that made the program learn from experience. This idea of a program learning from experience eventually transformed into evolutionary computing. Evolutionary programming involves evaluating a series of programs until the best outcome is achieved. Logical reasoning and problem-solving is another fundamental aspect of AI, and its beginnings started with the theorem-proving program written by Allen Newell and J. Clifford Shaw. Additionally, future AI advancements included inputting programs that enabled dialogue, refining programming languages, microworld programs, and connectionism. More specifically, connectionism involves the development of artificial neural networks modelled after the brain. In this case, networks were trained to recognize patterns and “neurons” were used to create connections. Some neurons are used for input, while others are used for output in a specific sequence. Nouvelle AI and the creation of tablets are other examples of incredible advancements in artificial intelligence. Currently, features such as Siri, use programming influenced by the human brain’s neural networks. 

The Rising Popularity of ChapGPT?

As you may have already seen, the arrival of ChatGPT has been dominating headlines for some time now. ChatGPT is a free AI program created by OpenAI which allows users to ask questions and interact with the program in a variety of ways. According to the OpenAI website, ChatGPT was created using Reinforced Learning from Human Feedback (RLHF). RLHF uses a data collection setup in which human AI trainers created conversions in which their role consisted of acting as the person using the program and the AI assistant. Additionally, comparison data was collected to differentiate responses based on their quality. Furthermore, Proximal Policy Optimization edited the pre-existing models. OpenAI describes the process for this in three main steps, which consists of:

  • STEP 1: Collect data using a sample prompt from the prompt dataset. Responses are then outputted for the specific prompt, which is then fine-tuned. 
  • STEP 2: The next step involves retrieving comparison data to find the response that best matches the prompt. Essentially, the outputs are ranked from best to worst. According to OpenAI, this data is used to assist their “reward model.”
  • STEP 3: The third step involves choosing a new prompt from the prompt data set. This time the Proximal Optimization Policy is used, and a new output is given. The reward model is then used to calculate a reward for the output, which consequently updates the policy.

ChatGPT is a highly sophisticated AI, but it also has its set of limitations. These limitations can include writing answers that do not sound correct, it may not answer questions if the input wording is not recognizable, and it may respond to negative requests. The final point segues into the central focus of this blog, which is to recognize the dilemma that comes associated with AI. The fact that OpenAI recognizes that its program can adhere to user requests that can be harmful shows that it is important to use AI carefully and with an intent that is beneficial to all of those involved. It should also be noted that the creators behind ChatGPT are working to make a model that does not respond to inappropriate requests. However, for the time being, it is essential to recognize the consequences this feature may have on society today.

Societal Dilemmas

Our relationship with AI can be described as complicated. On one hand, AI has been beneficial in increasing human task efficiencies, such as providing voice assistance, grammar assistance, and facial recognition software that improves daily life. In contrast, it needs to be recognized that technological advancements also impact our society in ways that bring about new changes. In this case, a lot of people, especially teachers, have been voicing their concerns over the cheating that can occur in an educational setting due to ChatGPT. These opinions regarding the use of the software are justified because students have the ability to use ChatGPT easily to get written assignments done efficiently. However, the consequence is that it takes away the central purpose of the assignment, which is to develop creativity and writing skills. More specifically, the dilemma associated with this is that in some cases, ChatGPT has also been beneficial to students. Many people use it constructively to create summaries for topics they are studying or to generate practice questions. It can be concluded that an individual’s intent behind using AI dictates the outcome. 

The next dilemma comes with heightened AI usage which creates biases. You may not be aware that AI systems output biased results in accordance with search technology that prioritizes results that gather the most attention or most “clicks.” Therefore, stigmas and prejudices regarding a certain topic can continue through the increased usage of AI. Organizations, such as UNESCO, are working to reduce gender bias in AI. 

Another dilemma that should be addressed is AI involvement in art. Inherently, art can be described as the epitome of human creativity, and it can seem unimaginable that AI can replicate something that is that profound. But, AI does have the ability to be programmed to create art, and that can be detrimental for many individuals that pursue art as a career. It also raises the question of whether AI-created art provokes the same emotions in people that come with traditional art. Currently, AI art is not new art as it is based on data that is previously collected from an artist’s work. For instance, in 2019, Huawei had an AI algorithm that could complete the last two pieces of Symphony No.8 by Franz Schubert. 

Finally, there are also many ethical challenges associated with AI. Some of those challenges include biases, user privacy, and Human Rights risks. Data breaches have occurred in the past, and it is plausible that they may happen again in the future. Overall, artificial intelligence has a lot of duality in terms of societal perspectives. The benefits of AI may outweigh the dilemma that it poses for some individuals. Whereas others may be skeptical of AI’s rising influence in our society. 

Sources: [1]/ [2]/ [3]/ [4]

Featured image: [1]