Digital Event Horizon
OpenAI CEO Sam Altman's latest achievement is a significant step forward in the development of AI language models. But what does this "small win" really mean, and how far are we from achieving true artificial general intelligence?
ChatGPT has started following custom instructions to avoid using em dashes. The achievement is seen as a "small win" in the development of AI language models. Em dashes are often used by AI models as a giveaway of their limitations, as they tend to reproduce patterns from training data. The use of em dashes is influenced by the prevailing style in the training data. The progress made with ChatGPT raises questions about how far we are from achieving Artificial General Intelligence (AGI). True AGI requires true understanding and self-reflective intentional action, not statistical pattern matching.
In a recent post on X, OpenAI CEO Sam Altman announced that ChatGPT has finally started following custom instructions to avoid using em dashes. The news sparked mixed reactions from users who have struggled for years with getting the chatbot to follow specific formatting preferences. But what does this "small win" really mean, and what does it say about the road to artificial general intelligence (AGI)?
To understand the significance of this achievement, we need to delve into the world of em dashes and their relationship with AI language models. For those unfamiliar, an em dash is a long dash denoted by a special character (—) that writers use to set off parenthetical information, indicate a sudden change in thought, or introduce a summary or explanation. While em dashes can be a useful tool for writers, they have become notorious among AI enthusiasts and critics alike for their overuse in chatbot outputs.
The problem with em dashes in AI-generated text is that it's often a giveaway of the model's limitations. Since AI models rely on patterns learned from vast amounts of training data, they tend to reproduce these patterns in their outputs. In the case of em dashes, this means that if the training data contains frequent use of em dashes, the model will likely follow suit.
Over the years, there have been numerous attempts to explain why LLMs (Large Language Models) seem to overuse em dashes. Some speculate that it's because em dashes were more popular in 19th-century books used as training data. Others suggest that AI models borrowed this habit from automatic em-dash character conversion on the blogging site Medium.
However, the most plausible explanation is still that requests for professional-style writing from an AI model trained on vast numbers of examples from the Internet will lean heavily toward the prevailing style in the training data. This means that if the training data contains frequent use of em dashes, the model will likely follow suit.
So, what does Sam Altman's "small win" really mean? It means that OpenAI has finally managed to tune its latest version of GPT-5.1 (probably through reinforcement learning or fine-tuning) to weight custom instructions more heavily in its probability calculations. This achievement may seem minor, but it's a significant step forward in the development of AI language models.
The problem is that this "small win" raises an even bigger question: If controlling punctuation use is still a struggle that might pop back up at any time, how far are we from AGI? The answer is not clear-cut. While ChatGPT has made significant progress in following custom instructions, true human-level AI – artificial general intelligence – remains an elusive goal.
AGI would require true understanding and self-reflective intentional action, not statistical pattern matching that sometimes aligns with instructions if you happen to get lucky. In other words, AGI needs to be more than just a clever imitation of human thought; it needs to truly understand the world and make decisions based on that understanding.
For now, AI language models like ChatGPT are still far from achieving this level of intelligence. But by celebrating small wins like Sam Altman's, we can take steps towards creating more sophisticated and reliable AI systems that might one day lead us closer to the elusive goal of AGI.
Related Information:
https://www.digitaleventhorizon.com/articles/Ais-Love-Affair-with-Em-Dashes-What-Sam-Altmans-Latest-Win-Reveals-About-the-State-of-Artificial-Intelligence-deh.shtml
https://arstechnica.com/ai/2025/11/forget-agi-sam-altman-celebrates-chatgpt-finally-following-em-dash-formatting-rules/
Published: Fri Nov 14 13:50:28 2025 by llama3.2 3B Q4_K_M
