Gemini 1206 Exp has a 2 million token context window, even if that isn't the effective context it probably performs much better in that regard than gpt 4o and such. Haven't tested yet because I don't want to get ratelimited from ai studio incase they monitor that
Frankly the "shorter" conversations I had at a few tens of thousand of tokens were already noticeably more consistent than before, e. g. it referenced previous responses significantly later
Great calls for 2024, I'd say most are atleast partially accurate.
However, looking at 2026, you definitely underestimated the pace of txt2video development like myself. Given that Veo 2 can already make sequences with cuts and show the same subject across both clips, the 60s consistency will probably be reached in 2025. However above DALL-E 3's quality, that has been surpassed now.
I'd say in late 2026 at the earliest or more realistically late 2027 because of compute constraints we'll see a product that can generate coherent feature film length, optionally photorealistically.
As for humanoid robots I'd say they are market ready at a reasonable price in 2027 or 2028. It doesn't make sense to look at cloud sota here because these robots will likely have to use edge compute due to privacy concerns when going mainstream in households and that in turn draws a lot of energy when running in real time across modalities. So multiple hardware issues to solve.
I'm also betting that there will be a Her-like product in the next 12 to 16 months that is indistinguishable from the movie version.
2 days ago and I might already have to adjust the timelines.
Nvidia's new Digits costs 3K and is the size of a mac mini. Two of them can supposedly run a 400B parameter language model which is crazy. So maybe the hardware issues aren't as persistent for robotics.
And also Hailuo has a single-image reference mode now that works like a lora. It's super consistent for faces, even if the rest is a bit quirky.