It’s not difficult to hack this together with CLIP. I did this with about a tenth of my movie collection last week with a GTX 1080 - though it lacks temporal understanding so you have to do the scene analysis yourself
I'm guessing you're not storing the CLIP for every single frame, instead of every second or so? Also, are you using the cosine similarity? How are you finding the nearest vector?
Sure. I had a lot of help from Claude Opus 4.5, but it was roughly:
- Using pyscenedetect to split each video on a per scene level
- Using the decord library https://github.com/dmlc/decord to pull frames from each scene at a particular sample rate (specific rate I don't have handy right now, but it was 1-2 per scene)
- Aggregating frames in batches of around 256 frames to be normalized for CLIP embedding on GPU (had to re-write the normalization process for this because the default library does it on CPU)
- Uploading the frames along with metadata (timestamp, etc) into a vector DB, in my case Qdrant running locally along with a screenclip of the frame itself for debugging.
I'm bottlenecked by GPU compute so I also started experimenting with using Modal for the embedding work too, but then vacation ended :) Might pick it up again in a few weeks. I'd like to be able to have a temporal-aware and potentially enriched search so that I can say "Seek to the scene in Oppenheimer where Rami Malek testifies" and be able to get a timestamped clip from the movie.
I have a weird sense that this is a way to get his kids that are technical into a family business without having them work at a company that isn't considered prestigious
Rephrased: If a graduate with relevant coursework from a top institution struggles to find a job in a particular field, what sort of chances do the rest of the graduates from less known colleges have?
For those unaware or may find her name familar, Rona is known for her plagiarism scandal. She blocks anyone on Twitter who asks about it or the book she got a substantial advance for but didn’t publish after this scandal came to light. She seems to have walked away from it - this utter elite impunity makes me sick.
I mean, the previous administration famously pulled strings across Twitter and Facebook to demote right wing media outlets on those platforms. This kind of crap isn’t new, and needs to stop.
That didn't happen. In fact Trump was President when Joe Biden was supposedly doing this according to Musk's Twittergate campaign, which was pure propaganda. (The right's explanation of this, when they even bother, was that the government was a bunch of deep state leftists.)
I don’t know if anyone will accept those sources. They’re the first two I found when looking for it. It could be misleading. The daily wire is obviously incentivized to report bad things about the Biden admin. But, I remember reading similar reports from a variety of sources a few years back, and could probably track those down if it’s helpful.
A shocking number of people are simply unaware (or worse, don’t care) that the current regime pardoned a thousand insurrectionists either while being nakedly corrupt to the point of taking cash in CAVA bags. The attention simply isn’t there.
> Hadn't written a code before he switched majors in college. Went from having zero experience with programming, to being a straight A computer science student
I’ve met these people before, they’re almost always undergrads from elite schools that could pick and succeed at anything they choose
I don't think elite school has much to do with talent just with how visible the things they are trusted with are.. From community college I've met some building things for companies too out of the mainstream to understand how unusual their IT worker was.
Pretty much all universes are first and fore-most strong people filters from my experience.
The prestigious schools filter up-front based on merit, and filter a 2nd time based on the ability to keep up and remain diligent to pass the courses. Less famous schools only have the 2nd filter, and have more people drop in the process that could not keep up.
The quality of education can be very similar, or even favor small schools in some subjects based on luck and school staff interests. The vast majority of learning at this level is made by the student anyway; there is only so much a good teacher or expensive teaching resources can do to help.
I've talked with interviewers that seemed completely uninterested in what courses or major was taken, just the fact that someone got through a difficult major as a stamp of quality.
At my undergrad there were 16 4.0s in a CS graduating class of about 120. I didn’t get close - about a 3.92 because I’m a poor student too. Grade inflation isn’t just a thing at elite schools, even my undergrad with a 45-55% accept rate
I'm not sure accumulated average grade metrics are actually very useful. Just the difficulty variation between courses means you can game those metrics with careful choice anyway, Or studying in ways that benefit the tests more than the subject.
Choosing difficult courses and passing is more valuable to the student. (well, a 4.0 on a paper is also valuable in a way, but more in a charade way)
What concerns me is not grade inflation, its difficulty and deapth deflation. I had plenty of courses that bored me but that had plenty of fun depth and difficulty that could have been covered with a bit more effort of students and teacher.
But we cannot measure courses difficulty and deapth, so we measure grade inflation instead. Which then lead us to false conclusions about how to solve it. (making the metrics more accurate at the cost of time and resources fir teaching the subjects)
i think less emphasis on grade and more emphasis on covered subjects could mabie shift the incentive. Its difficult to actually implement through because humans love maximising metrics... so it will remain my fantasy.
If you look at the resumes of anyone at top schools and who gets into YC this isn’t true, they’re simply more accomplished before college than the rest of us are after (or in my case, ever).
I look at resumes from time to time, just enough to know there aren't enough noble prizes.
I also knew a number of elite school students and aside from usually having wealthy parents and motivation problems when given real world tasks with real world levels of possible reward, they aren't all that far away from the average.
Take for example SBF.. possibly a genius would have gone straight to juvi either way if he was from a community college town and wasn't a nepo child of actual geniuses. More important than having any skills in the subject, he spoke economic shibboleth as a native speaker which his parents learned as a second tongue in a much harder process.
reply