By 2025 I expect language models to be uncannily good at mimicking an individual's writing style if there's enough texts/emails/posts to train on. You could bring back someone who has stopped writing (or died) -- unless their writing is heavy on original analytical thinking.
Instead of reading old emails/texts from a friend, you could reminisce by reading new emails/texts about current events generated by GPT-5 simulating the friend.
Instead of re-reading Orwell's 1984 and Animal Farm, you could read the "1984 reboot", a GPT-5 version of 1984 updated for the 2020s.
Sorry! To clarify: The prediction for 2025 is matching an individual style's up to 1-2 pages of text unless the person is doing novel analytic reasoning (e.g. math or working out an innovative cake recipe). Novels (which require coherent plot) would also come later.
GPT-3's ability to ape styles is already impressive but 1-2 pages of text will often be incoherent / non-truthful. I expect this to improve significantly in 2 years from scaling model size, RLHF, and info-retrieval (see WebGPT, LaMDA, InstructGPT).
• • •
Missing some Tweet in this thread? You can try to
force a refresh
