Searching...
12 results for “language models”
...the the the language, sequential model of LSTM. We we might be able to learn,
14:12 / 15:16
from transcript14:12 – 15:16
...models that match larger models, that's when you can really unlock some really interesting on device use cases. And I think a trend that we're noticing now is the trend of training smaller models longer. For example, if you compare how much how long
21:00 / 22:01
from transcript21:00 – 22:01
...to understand language models, which doesn't really have a, like, a Chinese company that is has an analog. While I'm talking, I'll say that the Chinese open language models tend to be much bigger, and that gives them this higher peak performance as M
44:58 / 46:11
from transcript44:58 – 46:11
...models, inside LLMs. And so I think they'll be very helpful as things like orchestrators. The problem is when it comes to the generalization, I think, text as a generalization backbone. When most of the the, when most of the the the pre training is i
47:13 / 48:16
from transcript47:13 – 48:16
...small models can only do so much. And the problem with learning about LLMs at scale is, I would say, it's exponentially more complex to make a larger model because it's not that the model just becomes larger. You have to now think about sharding your
2:13:17 / 2:14:33
from transcript2:13:17 – 2:14:33
...of the models are really close. So I think we've made like hugely forward in terms of small models. Of course, that's one just one data point but there's more. For example, if you look at this chart from the QUIN 2.5 blog post, it shows, that today w
17:58 / 19:05
from transcript17:58 – 19:05
...models, Pix 12 b. It says open source and Pix 12 large, just amazing model models for not understanding, images, but also great at text understanding. So, yeah, a lot of the image models are not so good at text understanding, but pixel large and pixe
33:48 / 34:55
from transcript33:48 – 34:55
...language models and the multimodal language models that we're seeing nowadays, their underlying representation under the hood is a one dimensional representation. We talk about context lengths. We talk about transformers. We talk about sequences, att
24:45 / 25:47
from transcript24:45 – 25:47
...and world models. I think speech and text generation will just be
48:01 / 49:04
from transcript48:01 – 49:04
...learning for language models and the growing post training and mid training stack with hot takes on everything from constitutional AI to DPO to rejection sampling, and also previewed the sea change coming to the Allen Institute and to InterConnex, hi
0:49 / 2:04
from transcript0:49 – 2:04
...models, the one domain where you have the most training data is probably coding. Right? And coding is where you can also have the most structure. And yet, anyone who has used,
44:15 / 45:14
from transcript44:15 – 45:14
...language models are very difficult to serve. Base ten was the first of the Inference NeoCloud startups to get DeepSeq v three online because of their h 200 clusters, their close collaboration with the DeepSeq team, and early support of SG Lang, a new
0:36 / 1:58
from transcript0:36 – 1:58
Have a podcast?
Run your podcast through Clypt
Get ranked clips, hooks, and ready-to-post copy from your own episodes. Free to try.