This title has existed in my head for over half a year. During this time, I have tried to finish writing it several times, but it always fell just a bit short. Consequently, this became the second question lingering in my mind for the past six months: what exactly was missing?
I always attributed it to my own laziness, until this moment, while watching OpenAI's DevDay, I suddenly realized that the "missing piece" was right here.
This is not a quick review of DevDay. I won't expand on: 1. A new model called GPT-4 Turbo; 2. Longer context lengths and cheaper API rates; 3. Customizable GPTs; 4. Multi-modal Assistants that accept various types of input, accompanied by voice generation from the new Whisper V3 model...
I will only say that as these new features are presented one by one, that title—"If Everyone Has a Large Model"—is about to become a reality. And we have arrived at this "singularity" that everyone must face seriously: another world that is constantly changing with immense acceleration, whether we choose to compete or to lie flat.
It is not about whether the impact is good or bad; it is only about where we are and what we are doing. It is only about how we welcome this world and how we coexist with it:
Perhaps, starting from today, we no longer need to force ourselves to constantly challenge our physiological and psychological limits by forcibly accepting more and more nutrient-free information, because a "partner" called Assistant can help us absorb it;
Perhaps, we can suddenly reconcile with ourselves: as it turns out, we don't need 99% of the information;
Perhaps, starting from today, we no longer need to use various tools to record agendas and to-dos, or set reminder after reminder, because a "partner" called Assistant can help us;
Perhaps, we can suddenly reconcile with ourselves: as it turns out, we don't need 99% of those tasks;
Perhaps, starting from today, we no longer need to fret over report deadlines, because a "partner" called Assistant can help us;
Perhaps, we can suddenly reconcile with ourselves: as it turns out, we don't need 99% of those reports;
Perhaps, starting from today, we no longer need to worry about what kind of UI or program structure we should have, because a "partner" called Assistant only needs the simplest way to interact: a text box, a sound, a file, or a pair of "glasses";
Perhaps, we can suddenly reconcile with ourselves: as it turns out, we don't need 99% of those user interfaces;
...
Perhaps, starting from today, every one of us can have a large model;
If, starting from some day, every one of us has a large model;
That physical world we are interacting with every moment, which seems unable to carry any more, will suddenly have countless digital versions that are also evolving at an accelerated pace every moment. They can exist independently or merge; they are invisible and indefinable;
Thus, starting from some day, X1 people will keep searching for "investment opportunities," X2 people will be anxious about how to change their workflows, X3 people will try to become "faster and stronger" information intermediaries... and Xm people will say, none of this has anything to do with me...
So, it becomes a combinatorics problem: C(N-1, m-1).
Interestingly, when the number of options exceeds half the number of people, the possibilities actually decrease: the increasing "personalization" created in the physical world will eventually converge in the digital world in an extremely "personalized" way;
I can end it here. As for what I've written, 99% of it is fluff. It's just that if, starting from some day, typing on a keyboard becomes a scarce "performance art," it seems that what I am doing right now is at the lowest point of time value.