elexhobby

6 ms·

1.
by elexhobby 3y ago
For another possibility, see https://arxiv.org/abs/2305.15717 . The new models may not actually be better - the evaluation may be broken.
2.
by elexhobby 3y ago
GPT-4 is powerful over a diverse set of tasks. They use it to build a model which is better for a narrow sub-task. Pretty sure the model is sub-optimal to GPT-4 for everything else.
3.
by elexhobby 3y ago
Not sure if this is obvious. But its incorrect to ditch on GPT-4. The paper uses self-instruct on GPT-4 to generate the training data on which it is fine-tuned. This paper would not exist without GPT-4. Although they claim GPT-4 can be repl
4.
by elexhobby 3y ago
This! The best resource I've found to explain transformers, that made them clear to me. I wish all deep learning papers were written like this, using pseudocode.
5.
by elexhobby 4y ago
FWIW I've been told similar by Costco too, so Amazon isn't unique in this respect. My guess is that it adds some friction to the process. Refunding money is a purely digital activity, while refunding requires to you go and ship it
6.
by elexhobby 5y ago
Furthermore, follow https://twitter.com/_brohrer_/status/1425770502321283073 "When you have a problem, build two solutions - a deep Bayesian transformer running on multicloud Kubernetes and a SQL query built
7.
by elexhobby 5y ago
any twitch streams you recommend?
8.
by elexhobby 6y ago
Wow. I wish everything was explained so clearly. I understood everything in that post except this paragraph. """ Languages like Erlang must implement tail call optimizations, since persisted data is stored as "loop varia
9.
by elexhobby 6y ago
Quote: "The election comes down to a few swing states, such as Pennsylvania, Wisconsin, and Michigan. Crucially, right now all three of those states have Democratic governors and Republican-controlled legislatures … and there’s no clea
10.
by elexhobby 7y ago
Great post, thanks! Is there a reason why the training is started off with two separate matrices - the embedding and the context matrix? If the context matrix is anyway discarded at the end, why not start and work with only the embedding ma
11.
Ben Recht (optimization professor at Berkeley) starts writing a blog(argmin.net)
2 points by elexhobby 10y ago|0 comments