When Altman celebrates finally getting GPT to avoid em dashes, he’s really celebrating that OpenAI has tuned the most recent version of GPT-5.1 (probably through reinforcement learning or fine-tuning) to weight custom instructions more heavily in its probability calculations.
There’s an irony about control here: Given the probabilistic nature of the difficulty, there’s no guarantee the difficulty will stay fixed. OpenAI constantly updates its models behind the scenes, even throughout the same version number, adjusting outputs based on user feedback and recent training runs. Each update arrives with different output characteristics that may undo previous behavioral tuning, a phenomenon researchers call the “alignment tax.”
Precisely tuning a neural network’s behavior will not be yet an actual science. Since all concepts encoded within the network are interconnected by values called weights, adjusting one behavior can alter others in unintended ways. Fix em dash overuse today, and tomorrow’s update (geared toward improving, say, coding capabilities) might inadvertently bring them back, not because OpenAI wants them there, but because that’s the character of attempting to steer a statistical system with thousands and thousands of competing influences.
This gets to an implied query we mentioned earlier. If controlling punctuation use continues to be a struggle that may pop back up at any time, how far are we from AGI? We are able to’t know needless to say, but it surely seems increasingly likely that it won’t emerge from a big language model alone. That’s because AGI, a technology that will replicate human general learning ability, would likely require true understanding and self-reflective intentional motion, not statistical pattern matching that sometimes aligns with instructions in the event you occur to get lucky.
And speaking of getting lucky, some users still aren’t having luck with controlling em dash use outside of the “custom instructions” feature. Upon being told in-chat to not use em dashes inside a chat, ChatGPT updated a saved memory and replied to 1 X user, “Got it—I’ll stick strictly to short hyphens any more.”
