Can we even achieve AGI with LLMs, why do AI bros still believe we can?

*The AGI Conundrum: Can LLMs Really Deliver?*

The concept of Artificial General Intelligence (AGI) has been a topic of interest in the AI community for decades. While some experts believe that Large Language Models (LLMs) are the key to achieving AGI, others are more skeptical. In this article, we'll examine the feasibility of using LLMs to reach AGI and why some experts remain unconvinced.

*The Case for LLMs*

Proponents of using LLMs for AGI argue that these models have made tremendous progress in recent years. They have been able to process and generate vast amounts of human-like text, demonstrate a level of common sense, and even show some degree of creativity. Some of the most notable examples of LLMs include Google's BERT, Microsoft's Turing-NLG, and Meta AI's OPT. These models have been able to achieve state-of-the-art results in various natural language processing (NLP) tasks, which has led some to believe that they can be scaled up to achieve AGI.

However, critics argue that these models are still far from achieving true intelligence. They are limited to a narrow scope of tasks and lack the ability to generalize to new situations. Moreover, the complexity of human cognition and the limitations of current computational power make it difficult to scale up these models to achieve AGI.

*The Reality Check*

While LLMs have made significant progress, they are still far from achieving the level of intelligence that would be required for AGI. AGI is not just about processing and generating human-like text; it requires a deep understanding of the world, the ability to reason, and the capacity to adapt to new situations. Currently, LLMs lack these capabilities, and it's unclear whether they can be scaled up to achieve them.

Moreover, the complexity of human cognition is not fully understood, and it's unclear whether it can be replicated in a machine. Human intelligence is not just a matter of processing power; it requires a deep understanding of the world, emotions, and social interactions. LLMs, on the other hand, are narrow AI systems that are designed to perform a specific task.

*The AGI Bro Problem*

Despite the lack of evidence, some experts continue to believe that AGI will come from LLMs. This optimism is not based on any concrete evidence but rather on rhetoric and a desire to see AGI become a reality. Some experts are so convinced that they ignore the limitations of LLMs and the complexity of human cognition.

However, this optimism is not shared by everyone. Many experts in the field are more cautious, recognizing that AGI is a complex and multifaceted problem that requires a deep understanding of human cognition and the limitations of current computational power.

*Conclusion*

While LLMs have made significant progress, they are still far from achieving AGI. The complexity of human cognition and the limitations of current computational power make it difficult to scale up these models to achieve true intelligence. The AGI conundrum remains a mystery, and it's unclear whether LLMs or any other approach will be able to achieve it. Until we have more concrete evidence, it's essential to remain skeptical and focused on the limitations of current AI systems.