323 karma · joined January 16, 2023
Books3 dataset 700B text-image pairs from Laion-2B-en, filtered to only keep images with at least 256 resolution 400M text-image pairs from COYO-700M, filtered to only keep images with at least 256 resolution 10M text-video pairs from WebVid10M 3M text-video pairs from a subset of InternVid10M 73K text-video chat pairs from Valley-Instruct-73K 100K text-video chat pairs from Video-ChatGPT
Disclaimer: I am not nostalgic for the internet of the old days, especially the times when you had a few websites to access, mostly university sites, and you could memorize each and every IP address. I actually hate the fact that the major search engines (especially Bing and Google) try to deceive me with the URLs I don't need. During the times of Altavista and Yahoo, I had the regex pipeline to filter their searches; however, this tactic is now futile...
Who was responsible at AMD for this project and why is he still not fired???????? How brain dead someone have to be to reject the major market share??????
Nope. But i'm so desperate to give you a hint right now, it is almost impossible to hold myself... Stop looking into horizontal scalability. The vertical one is not exhausted yet. Btw that was not the hint.
This is not true. You have tones of models those are even better than GPT-3.5 and really close in performance to GPT-4 and you still can train them on a single GPU with 24GB video memory. There is a hint at yet better models published last year which you can train on a single GPU and have a model comparable in performance to LLaMA2 34B. The horizontal scaling which you appeal here, may fit into 10^6 performance increase, but in general I expect single node to be at least 1000 times faster than now. And it is totally feasible that you can't scale with 0.99 vertically and of course not horizontally, but I honestly expect the scaling per GPU get better than 0.75 in next 5 years.
It appears obvious that if you add a 1D dimension that is perpendicular to all three observable dimensions and is a vector of the universe's expansion where most energy resides, then the universe expands at a constant rate in that 1D dimension and our observable 3D dimensions stretch by this vector in all directions simultaneously. However, we observe that the universe's expansion rate is accelerating, which suggests two things about our observations:
1. What we observe is under the time dilation effect, and each part of the universe we observe can have different rates of the time dilation effect due to the gravitational gradient of mass and the speed of matter (not the expansion speed). Thus, the universe is really expanding at a constant rate, and we are progressively slowpokes (our local time dilation effect is increasing... What could be the reason for this?).
2. There is no hidden dimension...
Adding an additional dimension of micron size transforms it into a 3D dimension. This is too far-stretched, even for a wild guesser like me...
... permanent colonies on worlds other than Fratlat. Although still largely unable to travel outside of the Latbûut system, the distribution of Fratbac civilization across multiple worlds greatly reduces the risk that they will collapse due to any crisis of merely planetary scale.
In 3904, a swarm of self-replicating Fratbac nanobots began to replicate uncontrollably, devouring vast swaths of Fratlat at a rate which Fratbac scientists had formerly deemed impossible. After several days of rapid expansion, the swarm seems to have become dormant, but not before consuming approximately 5% of the entire mass of Fratlat and rendering Fratbac civilization completely extinct.
Also, where are the probabilities for mass-extinction-event-volcanic-eruptions are coming from? It looks like they are very likely to happen ... I have 8/10 so far...
https://en.wikipedia.org/wiki/Clathrate_gun_hypothesis
Citation: The associated period of massive carbon release into the atmosphere has been estimated to have lasted from 20,000 to 50,000 years. The entire warm period lasted for about 200,000 years. Global temperatures increased by 5–8 °C.[21]
The hot-models Sabine referred to are in range of 4.8 - 5.6 if I'm not mistaken...