The Fatalism of Artificial Intelligence Will Not Help Us Address Its Actual Risks

2023-07-03

What is Artificial Intelligence?


Artificial Intelligence (AI) is a discipline in computer science or engineering that emerged in the 1950s. Its goal is to build intelligent computing systems that are inspired by human intelligence. Like human intelligence, AI is complex and diverse, with many fields aiming to simulate various aspects of human intelligence, from perception to reasoning, planning, and decision-making.



Based on different levels of capability, AI systems can be divided into three levels:

  • Narrow or weak AI refers to AI systems that can perform specific tasks or solve specific problems, often surpassing human performance. Currently, all AI systems belong to narrow AI. Examples include chatbots like chatGPT, voice assistants like Siri and Alexa, image recognition systems, and recommendation algorithms.

  • General or strong AI refers to AI systems that can exhibit human-level intelligence, including the ability to understand, learn, and apply knowledge across a wide range of tasks, and encompass concepts such as consciousness. General AI is largely hypothetical and has not been achieved to date.

  • Superintelligence refers to AI systems that surpass human intelligence in all tasks. By definition, we cannot comprehend this level of intelligence, just as ants cannot comprehend our intelligence. Superintelligence is a more speculative concept than general AI.


Artificial intelligence can be applied to various fields such as education, transportation, healthcare, law, or manufacturing. Therefore, it is profoundly changing all aspects of society. Even in its "narrow AI" form, it has significant potential to promote sustainable economic growth and help us address the most urgent challenges of the 21st century, such as climate change, pandemics, and inequality.

Challenges Facing Current AI Systems


In the past decade, AI-based decision systems applied in various fields, from social media to the labor market, have brought important social risks and challenges that require in-depth understanding and solutions.

The recent emergence of highly reliable large-scale generative pre-trained transformers (GPT) models has exacerbated many existing challenges and created some noteworthy new challenges. These tools have been widely adopted by hundreds of millions of people globally, putting greater pressure on our society and regulatory systems.

Here are some very important challenges that should be our top priority:

  • The manipulation of human behavior by AI algorithms can have disastrous social consequences, including the spread of false information, the formation of public opinion, and the outcomes of democratic processes.

  • Algorithmic bias and discrimination not only persist but also exacerbate patterns of stereotypes, discrimination, and even oppression.

  • Lack of transparency in models and usage.

  • Infringement of privacy and use of massive training data without the consent of creators, without any compensation. Exploitation of workers annotating, training, and correcting AI systems, many of whom come from developing countries and receive meager wages. The enormous carbon footprint of building the large data centers and neural networks required for these AI systems. Lack of authenticity in generative AI systems, capable of creating credible content (images, text, audio, video, etc.) unrelated to the real world. Vulnerability of large models to making mistakes and being deceived. Replacement of jobs and professions. Concentration of power in the hands of a few giants controlling current AI systems. Is AI really a threat to human survival? Unfortunately, public discussions on AI, especially recent open letters, have mainly focused on the hypothetical existential threat of AI, rather than addressing these specific risks.


Existential threats refer to events or scenarios that pose a threat to the continued existence of human society and have potentially irreversible consequences that could damage or destroy human civilization, or even lead to species extinction. Examples of existential threats include global catastrophic events such as asteroid impacts or pandemics, destruction of habitable planets (due to climate change, deforestation, or depletion of vital resources such as water and clean air), or global-scale nuclear war.

Our world does face many risks, and the future is difficult to predict. In the face of this uncertainty, we need to prioritize our efforts. The remote possibility of uncontrollable superintelligence needs to be put into proper context, including the fact that 3.6 billion people worldwide are highly vulnerable to climate change, approximately 1 billion people live in conditions of less than $1 a day, or 2 billion people are affected by conflicts. These are humans who are truly in grave danger, and this danger is clearly not caused by superintelligence.

Focusing on hypothetical existential threats diverts attention from the serious challenges we face today, fails to encompass diverse perspectives in the broader research community, and leads to unnecessary panic.

Society would certainly benefit from the diversity, complexity, and subtle differences required to address the challenges of today's AI, including regulation. Addressing these challenges requires the cooperation and involvement of the most affected sectors of society, as well as the necessary technical and governance expertise. It is time to take collaborative action.

 

Source: https://gizmodo.com/ai-chatgpt-fatalism-wont-help-us-with-its-actual-risks-1850585699

RECENT AI NEWS

RECENT AI TOOLS