deepsquirrelnet

8 ms·

91.
by deepsquirrelnet 10mo ago
I’m finishing up a language identification model that runs on cpu, 70k texts/s single thread, 13mb model artifact and 148 supported languages (though only ~100 have good accuracy). This is a model trained as static embeddings from the
92.
by deepsquirrelnet 10mo ago
> Berulis said he and his colleagues grew even more alarmed when they noticed nearly two dozen login attempts from a Russian Internet address (83.149.30,186) that presented valid login credentials for a DOGE employee account > “Whoeve
93.
by deepsquirrelnet 10mo ago
Heavens to Betsy, I don’t know if you can hear me, But try supporting these things if you actually want them to be successful. About the 3rd day into trying to roll your own LMI container in sagemaker because they haven’t updated the vLLM v
94.
by deepsquirrelnet 10mo ago
That might not be relevant to OPs use case. A lot of nurses get tied up doing things like reviewing claims denials. There’s good use cases on the administrative side of healthcare that currently require nurse involvement.
95.
by deepsquirrelnet 11mo ago
I love using encoder models, and they are generally a better technology for this kind of application. But the price of GPU instances is too damn high. I won’t lie that I’ve been unreasonably annoyed that I have to use a lot more compute tha
96.
by deepsquirrelnet 11mo ago
One of the issues with using LLMs in content generation is that instruction tuning causes mode collapse. For example, if you ask an LLM to generate a random number between 1 and 10, it might pick something like 7 80% of the time. Base model
97.
by deepsquirrelnet 11mo ago
SPLADE-easy: https://github.com/dleemiller/splade-easy I wanted a simple retrieval index to use splade sparse vectors. This just encodes and serializes documents into flatbuffers and appends them into shards. Retrieval
98.
by deepsquirrelnet 11mo ago
I think that happened when gpt5 was released and pierced OpenAIs veil. While not a bad model, we found out exactly what Mr. Altman’s words are worth.
99.
by deepsquirrelnet 1y ago
I haven’t used RCNN, but trained a custom YOLOv5 model maybe 3-4 years ago and was very happy with the results. I think people have continued to work on it. There’s no single lab or developer, it mostly appears that the metrics for comparis
100.
by deepsquirrelnet 1y ago
FWIW this has happens in consulting too, not just product companies. Just swap “product” for “delivery”.
101.
by deepsquirrelnet 1y ago
I think a less order biased, more straightforward way would be just to vectorize everything, perform clustering and then label the clusters with the LLM.
102.
by deepsquirrelnet 1y ago
> For the searches we use hybrid dense + sparse bm25, since dense doesn't work well for technical words. One thing I’m always curious about is if you could simplify this and get good/better results using SPLADE. The v3 models l
103.
by deepsquirrelnet 1y ago
Absolutely the first thing you should try is a prompt optimizer. The GEPA optimizer (implemented in DSPy) often outperforms GRPO training[1]. But I think people are usually building with frameworks that aren't machine learning framewor
104.
by deepsquirrelnet 1y ago
I go back and forth on this. A year ago, I was optimistic and I have had 1 case where RL fine tuning a model made sense. But while there are pockets of that, there is a clash with existing industry skills. I work with a lot of machine learn
105.
by deepsquirrelnet 1y ago
> “What would America’s Founding Fathers think if they were alive today?” > For Cross, it is pointless to speculate about the present-day views of men who could not have imagined cotton candy, let alone the machine that makes it. Some
106.
by deepsquirrelnet 1y ago
David Frum talked at length about self-abasement in MAGA public culture in his recent podcast for the Atlantic[1]. > I think it also becomes a real test of in-group loyalty to see who can outcompete in slavishness the other members of th
107.
by deepsquirrelnet 1y ago
It’s a result of the lack of rigor in how it’s being used. Machine learning has been useful for years despite less than 100% accuracy, and the way you trust it is through measurement. Most people using or developing with AI today have punte
108.
by deepsquirrelnet 1y ago
This seems like a great place for a Cypress (Infineon) PSOC. A while back, I interfaced one to a linear CCD and it was a great experience. They also have USB HID support on chip.
109.
by deepsquirrelnet 1y ago
More than that, adding longer context isn’t free either in time or money. So filling an LLM context with k=100 documents of mixed relevance may be slower than reranking and filling with k=10 of high relevance. Of course, the devil is in the
110.
by deepsquirrelnet 1y ago
The system you’re describing is one where lives are enriched by this drive away from zero sum, but in my view the one we observe is increasingly described by excess and inefficiency. The middle class is shrinking in the face of technologica
111.
by deepsquirrelnet 1y ago
Except that is not how it has played out. In science fiction, there are competing views of the future. One view of the future is like Star Trek, where people’s needs are easily provided for by technological advancements, and people spend th
112.
by deepsquirrelnet 1y ago
According to this post, technological stagnation via regulation is what leads to zero-sum, totalitarian societies (in Thiel’s worldview). I personally feel so little connection to this ideology, especially in the post-COVID world. Wanting
113.
by deepsquirrelnet 1y ago
The term has been so abused that it’s no surprise the current administration was going to push in that direction themselves. All it takes is a minor rebranding and half the country or more would immediately sign up for it. For all intents a
114.
by deepsquirrelnet 1y ago
I would guess OP is referring to the association with lowered birth weight due to chronic hypoxia: https://pmc.ncbi.nlm.nih.gov/articles/PMC7050200/
115.
by deepsquirrelnet 1y ago
SVGs are under explored in generative AI. They are effectively a graphics language of their own. LLMs can write them directly without a vision architecture, and understand them in ways that non-vector graphics cannot.
116.
by deepsquirrelnet 1y ago
I don’t think parading out engineers in shackles for a photo op was a good idea. From another article: > Images of South Koreans being shackled at the wrists and ankles have caused outrage in South Korea, a key U.S. ally in Asia that has
117.
by deepsquirrelnet 1y ago
It’s also an easy situation to manipulate. I see a lot of people eager to make assumptions about things that are not known. That is also a very predictable response if you live in this country.
118.
by deepsquirrelnet 1y ago
I also like that it ships with some cli tools, including an openai compatible server. It’s great to be able to take a model that’s loaded and open up an endpoint to it for running local scripts. You can get a quick feel for how it works via
119.
by deepsquirrelnet 1y ago
That was where my mind went as well. I was thinking about about what it would take to create a text-based game (eg The Wizard’s Castle), but augmented with a language model. This seems like a useful piece of doing that.
120.
by deepsquirrelnet 1y ago
> That could help explain why Zohran Mamdani, a Democratic socialist state assemblyman, won the Democratic primary for New York City mayor earlier this summer with the support of young people. My fear is that we are continuing to acceler

More ›