• 1 Post
  • 106 Comments
Joined 2 years ago
cake
Cake day: November 25th, 2023

help-circle

  • What I’m on about? I think the english term is “damning with faint praise”. If this is the best that can be done, which I am arguing, there isn’t much use to it.

    The latest one is an outlier, in that it doesn’t have a voice over, so it isn’t a radio play. Most of the other ones I have seen has a voice track that tells the story. They are also more dreamlike which matches the prediction of what kind of story can be told from one of the comment threads here (from one of the pivot videos about VEO).

    The latest one (and the only one to gone viral) is actually interesting in that he is trying to tell a visual story, but with the medium he has chosen he can’t have a novel character as protagonist really, or dialogue, which is why it is limited to a very simple story.

    I’m interested in why it is so limited, because I think that tells a lot of the limitations of the technology as such.


  • Not a sneer, but I recently saw Ari K’s AI generated video of Trump in his golden ballroom. It’s quite good, here is the channel: https://m.youtube.com/@AriKuschnir

    Looking at his other videos, he is a talented story teller. Most videos are about two minutes, has numerous short shots of a few seconds and a voice over or music connecting the shots. So presumably he generates the shots, splice them together and puts the soundtrack over the it. Most of the short stories are dreamlike. To the extent it has characters it’s famous people (getting their comeuppance), so even though they look a bit different in each shot, it’s easy to keep track.

    I think it’s interesting because by doing what can be done with the tools, it illustrates the limitations. In the hands of a good story teller you essentially get an illustration for a short radio play (and the radio play needs to be recorded separately, and you can’t show actors talking). Because of the bubble and investor bux, it can right now be done on a shoe string budget.

    But that’s all! Are illustrated radio plays replacing feature films? No, so this remains a niche use case. And once the investor bux dries up, potentially an expensive one. Not something to build a billion dollar industry on.






  • FWIW, I think he’s wrong in the causation here. During the heyday of the British Empire history was one of the high status subjects to study, and they wrote it in very plain language. Physics on the other hand was seen as mostly pointless philosophy, and in the early 19th century astronomy was a field so low in status that it was dominated by women.

    I would say the causation is money giving the field status, and lack of money hollowing out status. Low status makes the untrained think they can do it as well as the trained. You had to study history and master it’s language to make a career as a colonial administrator, therefore the field was high status. As soon as money starts really flowing into physics, the status goes up, even surpassing chemistry which had been the highest status (and thus also manliest) science.

    If one wants to look at the decline of status of academia, I recommend as a starting point Galbraith’s The Affluent Society, that goes a fair bit into the post war status of academia versus business men.

    I think the humanities were merely the weak point in lowering the status of academia in favour of the business men.













  • The ideas are in general good.

    I think the long term cost argument could be strengthen by saying something about DeepSeeks claims to run much cheaper. If there is anything to say about that, I have not kept track.

    The ML/LLM split argument might benefit from being beefed up. I saw a funny post on Tumblr (so good luck finding that again) about pigeons being taught to identify cancer cells (a thing, according to the post, I haven’t verified) and how while that is a thing you wouldn’t leap to putting a pigeon in charge of checking CVs and recommending hires. The post was funnier, but it got to the critical point of what statistical relationships reasonably can be used for and what it can’t, which becomes obvious when it is a pigeon instead of a machine. Ah well, you can beef it up in a later post or maybe you intended to link an already existing one. There is a value in being consise instead of rambling like I am doing here.