The pain points we encounter today may not be solvable through linear extrapolation.
After more than a year of hype, especially following the release of Apple Intelligence, the market has gradually begun to realize the validity of Sam Altman's perspective: minor tweaks and patches based on existing models are of little significance, as the goal of model upgrades is to become more general-purpose.
Undoubtedly, the era starting in 2023 is another "best of times," as there seem to be new tools and gadgets to experience every day; for instance, I am increasingly using Perplexity's Page feature for content creation.
Yet, it is also undoubtedly the "worst of times": every new tool seems to have a very short lifecycle. We can count the number of third-party plugins or GPTs eliminated by GPT updates over the past year, or the number of third-party apps likely to be wiped out once Apple Intelligence rolls out. Naturally, we can also look at the number of third-party tools we ourselves have quickly abandoned in the last year. "Out with the old, in with the new" might be a driver for continuous innovation, but it is certainly an enemy that kills the expansion of sustainable production scale.
Thus, as I experimented with another batch of tools and felt one step closer to the intermediate solution I desire, this title popped into my head: AI opportunities—it's all about the time window.
Whether for C-end or B-end, the core of AI applications remains having a sufficient customer base.
In Q1 of last year, when ChatGPT first exploded globally, Midjourney could provide text-to-image capabilities that GPT lacked, allowing it to quickly acquire users and monetize. However, once the multimodal GPT-4 was released and DALL-E 3 was integrated into ChatGPT, Midjourney gradually faded from the spotlight.
This is a classic example and reflects why tech giants have grown even larger over the past decade. However, this round of AI—or generative AI—has changed the time partial derivative function of productivity. The time required for an application to reach 100 million users is getting shorter. Rather than saying applications are getting better, it is more accurate to say that only those that acquire users at a faster pace can survive.
On the other hand, the accelerating evolution of base models is also increasing the speed at which "elephants" can turn. The increasingly aggressive strategies of Google and Apple, while still not fast enough for many users, may have compressed the innovation time window to less than half a year—what Sam Altman calls the "AI Law." Within this shrinking time window, if an application fails to gain a user base sufficient to challenge the "old powers," the result may be a "one-in-ten chance of survival" or, naturally, an acquisition.
I won't elaborate further; it's all about the time window.
Come to think of it, isn't the same true for the industries and jobs we are engaged in?