IMO, the algorithm is overhyped, and the secret to tik toks success is its scale. If you gave metas algorithm the same amount of data to train on and the same amount of content to recommend, it would be equally as addictive. It might suck at first but give it a couple months and people won’t be able to tell the difference between the old algorithm and whatever oracle creates.
It’s just like chatgpt, given the same set of public data any company with sufficient engineering and compute resources can make there own model that performs very similarly.
In an ideal world we’d have algorithmic choice, like in bluesky sort of, and we could actually compare different ones, but that would cut into the profits of the data monoplies.
IMO, the algorithm is overhyped, and the secret to tik toks success is its scale. If you gave metas algorithm the same amount of data to train on and the same amount of content to recommend, it would be equally as addictive. It might suck at first but give it a couple months and people won’t be able to tell the difference between the old algorithm and whatever oracle creates.
It’s just like chatgpt, given the same set of public data any company with sufficient engineering and compute resources can make there own model that performs very similarly.
In an ideal world we’d have algorithmic choice, like in bluesky sort of, and we could actually compare different ones, but that would cut into the profits of the data monoplies.