Artificial Intelligence (AI) has been advancing rapidly in recent years, with various applications ranging from image recognition to natural language processing. However, there is still much to achieve regarding Artificial General Intelligence (AGI), which can perform various intellectual tasks typically associated with human cognition. As a language model trained by OpenAI, ChatGPT offers a glimpse into the poatential of AGI, and its development represents a significant step forward in the field of AI. In this article, we explore the implications of ChatGPT for the future of AI and its role in the pursuit of AGI.

My unique perspective is derived from reading books on Artificial Intelligence such as - Our Final Invention: Artificial Intelligence and the End of the Human Era by James Barrat and The Singularity is Near: When Humans Transcend Biology by Ray Kurzweil - which helped me to form a distinct perspective. After these readings, I explored Yuval Noah Harari's vast historical narratives. These were not read in any particular order, but they helped bring all the knowledge together; this hardly qualifies me as an AI or Big history expert. However, my opinions diverge from the depictions of cyborg assassins in the Terminator series starring Arnold Schwarzenegger and I, Robot starring Will Smith. 

Artificial intelligence is a broad term that encompasses various types of systems that can perform tasks that normally require human intelligence, such as perception, reasoning, learning, decision making and natural language processing. However, not all AI systems are created equal. Some are more specialised and narrow in scope, while others are more general and versatile.

The field of artificial intelligence deals with building computers and machines capable of reasoning, learning, and acting in a way that would normally require human intelligence or involve a high level of data humans cannot analyse. It is a broad field encompassing many disciplines, including computer science, data analytics and statistics, hardware and software engineering, linguistics, neuroscience, philosophy, and psychology.

Artificial intelligence refers to a set of technologies that enable machines to perform tasks that typically require human-like intelligence. These technologies are primarily based on machine learning and deep learning for data analytics, predictions and forecasting, object categorisation, natural language processing, recommendations, intelligent data retrieval, and more. This operational definition provides a clear understanding of the capabilities of AI that we have given them today. It is essential to note that there are different types of AI, such as Narrow AI, designed to perform specific tasks based on human instructions, and Artificial General Intelligence (AGI), which is the ultimate goal of AI and the subject of much debate in popular culture. While Narrow AI is prevalent and used almost daily, AGI is the type of AI that could potentially be our final invention. It can think, act, and even attain sentience in the future, making it indistinguishable from humans. 

However, achieving AGI is a challenging feat. It requires solving complex and open-ended problems, such as common sense reasoning, knowledge representation, transfer learning and explainability. Moreover, it poses ethical and social challenges, such as ensuring such powerful systems' safety, fairness and accountability.

One recent AI development that has sparked interest and debate among researchers and enthusiasts is the introduction of Generative Pre-trained Transformer (GPT) models. ChatGPT is a chatbot AI developed by OpenAI that allows users to do many tasks. The user asks the chatbot a question and then responds with what it thinks is the desired result. If you do not get the desired result the first time, one of the best tips for using ChatGPT is to keep asking. A prompt-based system only gives you an output based on the prompts, i.e., inputs it is given.

Generative Pre-trained Transformer (GPT) models are excellent examples of advanced AI technologies that can push us closer towards achieving AGI. An AGI system could perform any task across any domain. Its only limitations would be those imposed by its creators.

ChatGPT is built on top of OpenAI's GPT-3 family of large language models (LLMs) that can understand and generate human-like answers to text prompts because it has been trained on huge amounts of data. ChatGPT has been fine-tuned using supervised and reinforcement learning techniques to improve its conversational skills. Its creators heavily restrained or regulated it to make it ethical and safe for public use.

Large Language Models (LLMs) are advanced computer programs that can understand human language. They use special techniques to learn from much text and understand how words and phrases relate. Its use and applications extend beyond natural language uses.

ChatGPT was launched as a prototype on November 30th 2022, and quickly garnered attention for its detailed responses across various topics. Some users praised its creativity and humour, while others criticised its inconsistency and bias. ChatGPT can also generate content like poems, stories, code snippets, etc. Other prompt-based AIs recently developed are Midjourney, an image generation tool, and DALL·E 2, an AI system that also creates images using natural language descriptions just like Midjourney.

Nevertheless, does ChatGPT qualify as an AGI system? The answer is no. ChatGPT demonstrates exceptional natural language understanding and generation capabilities that surpass many existing chatbots. It can also handle multiple domains and tasks with ease. On the other hand, ChatGPT still relies on pre-trained models that may need to generalise better to new situations or domains. It also lacks some essential features of human intelligence, such as self-awareness. It is not an AGI but has the potential to be in the future.

Some jobs are reported going to be replaced by AIs. There has been much speculation seeing that ChatGPT has passed almost all specialised exams in the medical, legal and other knowledge domains that were previously considered to have a high barrier to entry. It makes one wonder what the fate of the typical workspaces would look like as complementary technologies develop around Generative Pre-trained Transformers and gets integrated into other complex systems. The introduction of Generative Pre-trained Transformers (GPTs) is expected to impact approximately 80% of the US workforce, with at least 10% of their work tasks affected (Eloundou et al., 2023). Around 19% of workers may see at least 50% of their tasks impacted, and this influence spans all wage levels. Higher-income jobs may face greater exposure (Eloundou et al., 2023). The impact is not limited to industries with higher recent productivity growth (Eloundou et al., 2023). GPTs exhibit general-purpose technologies (GPTs) characteristics, which could have significant economic, social, and policy implications (Eloundou et al., 2023).

GPTs come with inherent biases inherited from their human creators, factual inaccuracies, and privacy concerns, which must be addressed as they are integrated into various industries and complementary technologies. However, with time and positive feedback, GPTs have the potential to improve and create better, more factually accurate, and unbiased versions of themselves through self-iterations. Such self-improvement is often a precursor to Artificial General Intelligence's (AGI) development.

The potential impact of AI on our jobs and society is a topic of much discussion and speculation. Will AIs replace our jobs entirely, will we reject them altogether, or will we merge with AIs to enhance our abilities and improve our work? This concept is commonly referred to as the singularity. There have been numerous use cases and examples of AI in various industries, with both positive and negative outcomes.

For this project, ChatGPT and Bing were utilised to support writing, coding, and formatting.

Eloundou, T., Manning, S., Mishkin, P., & Rock, D. (2023). GPTs are GPTs: An early look at the labor market impact potential of large language models. arXiv preprint arXiv:2303.10130

| blog