Serious question - how do we know the Deepseek APIs ARENT chatGPT? Couldnt they have just registered for chatGPT APIs through intermediaries for their consumer facing app, release a open source model that in no way replicates the performance of the consumer app (“because of no RL”), and tell everybody its just really innovative training? They are losing a lot of money for now, but building a worldwide brand that im sure the CCP would be more than happy to foot the bill for mobing forward given its damage to the US economy and ability to steal IP from any coder dumb enough to use it.
The source of the traffic to ChatGPT API servers would probably be obvious and if not, anomalous.