i agree with the bubble sentiment that apparently many people have. i recognize how it would be at my own career's expense. but i feel that many arguments made here miss the forest for the trees.
applied statistics or statistical learning has been around long enough and we have seen its innovations and rebranding over the decades. i clearly see the theoretical point to it and hence decided to find my place in this field.
however, the "AI" movement as of late, including the generative AI bits, fall into the bubble territory for me. just like those who are serious about blockchain and its wide implications will still toil towards it, so would companies serious about machine learning.
however, most people are riding the wave are for short-term gains, just like many in crypto space were there for speculative money-making.
the LLMs to me are an evolution (albeit a macro one) to predictive functionality of smartphone keyboards of the past, but they are touted to be the holy grail. their capabilities are impressive, but it only scales up so much in its current form. those just making an app on top of api provided by these services will not last. moreover, the explosion of advancements mean there will be no stability for those maintaining the infrastructure in the near future.
at least with the pursuit of making the largest models have shed light on the need to optimize the deep learning stack, which is the only silver lining for me.
i would love to be wrong and see what comes next, but i believe the general public will lose interest soon and we will have another winter before a major breakthrough. the "AGI" claims are just like those made by vr enthusiasts in the 2010s...i mean the 1980s.