After a day of fermentation, the reasons behind the major personnel upheaval at OpenAI are becoming clear: Chief Scientist Ilya and CTO Mira believed that a series of new products, represented by the GPT Store, violated OpenAI's original mission of "safe AI," leading them to launch a surprise move using the board's rules.
Interestingly, 24 hours ago, what I intuitively thought was the cause was exactly the opposite: I assumed it was driven by the power of capital, but now it seems the most likely driver is an anti-capital force.
Perhaps deep down, I preferred it to be driven by capital forces. As an individual with somewhat idealistic sentiments and a certain level of self-perceived purity, the tragic undertones of such an explanation would be easier to accept.
However, the new version clearly makes me feel more excited and encouraged: It turns out it can be like this; it turns out there really is a "line," and there are truly people willing to guard it.
To me, this is a state of psychological conflict reaching the level of internal exhaustion. For OpenAI and the people involved, it is precisely a difficult choice between extreme contradiction and direct conflict.
AI—the difficult choice between efficiency and safety—is once again placed directly before us, questioning our hearts, swaying our complex considerations of interest, and impacting our fragile values.
In November 2022, ChatGPT was released, and humanity reached a crossroads in a brand-new field. In November 2023, with the OpenAI shakeup, humanity and AI together face a new crossroads.
One year in the field of AI consists of two half-years and the iteration of two generations of products.
We have witnessed the madness of everyone going "all-in," the "AI+" transformation of countless products, the rise of the new era's "shovel sellers," and the deception and bubbles that accompany the arrival of every new species.
The good and the bad are presented one by one. Looking at myself critically once more, perhaps I will believe even more firmly in open communities, shared ideas, and the pure pursuit of experience... whether that stems from an abnormal sense of self-satisfaction and self-actualization or a character flaw rooted in excessive pride.
Perhaps we simply want to know the answers to some ultimate questions; perhaps we simply want to turn the images that repeatedly surface in our minds into reality. It is just that some feel a line has been crossed, some feel it is not enough, and some wonder: how do we monetize this?
The story has a sequel. Efficiency and safety, interests and risks, original intentions and temptations will continue to impact everyone involved or observing from the sidelines.
These things may not become training data for AI, but they may also be where the greatest dignity of being human resides.