ukuina

36 ms·

691.
by ukuina 3y ago
The current trend for productionizing LLM-based applications is to write your own (really thin) wrappers around the actual LLM call. The majority of your code should be business logic independent of the LLM, anyway: information retrieval, u
692.
by ukuina 3y ago
So much "Yes!" for LangChain being bad practice. An unnecessarily bloated abstraction over what should be simple API calls.
693.
by ukuina 3y ago
I wouldn't be surprised if Tesla starts car buybacks for vehicles older than 15 years to offset their parts-inventory management overhead.
694.
by ukuina 3y ago
Which is heavier: a pound of feathers, or a great british pound? > Both weights are equal.
695.
by ukuina 3y ago
Could you expand on this? I am unable to see the connection.
696.
by ukuina 3y ago
Here's a packaged solution! https://github.com/imartinez/privateGPT
697.
by ukuina 3y ago
When I tried this, it made up random, non-repeatable confidence values that had little bearing to the veracity of its statements.
698.
by ukuina 3y ago
Are you saying Amazon modifies prices for items based on which user is logged in?
699.
by ukuina 3y ago
Isn't that kind of the point? I would greatly prefer it stays within the confines of the supplied documents.
700.
by ukuina 3y ago
Isn't this basically Soma.fm?
701.
by ukuina 3y ago
I am not so sure. I asked GPT4 to "optimize" English for AI-AI communication, and the modifications were accepted by another GPT4 instance for seemingly-lossless discussion. First instance: https://shareg.pt/8vgEg
702.
by ukuina 3y ago
The behavior is true, though. I just tested it with Bard.
703.
by ukuina 3y ago
I have just verified this exact behavior with Bard. GPT3.5-turbo/GPT4 is way ahead in instruct tuning and does not require such verbal gymnastics.
704.
by ukuina 3y ago
The concern with finetuning, even for specialized use-cases, is that you are binding yourself to the underlying model. Given rapid advancements in the field, this does not seem a prudent use of engineering time. Having a hierarchy of prompt
705.
by ukuina 3y ago
Thank you so much for this, especially for allowing custom LLM calls to allow testing of local models.
706.
by ukuina 3y ago
This is the way.
707.
by ukuina 3y ago
And when it underperforms relative to expectations, they claim it's not running the best model they have.
708.
by ukuina 3y ago
How does Midjourney factor into business idea generation?
709.
by ukuina 3y ago
This is very true. There is so much reinvention of processes happening behind the scenes around GPT4 as an API. These advancements will take a few quarters to show impact.
710.
by ukuina 3y ago
Some (many?) of those statements have been empirically proven false, like understanding humor. Multi-modal LLMs can explain why certain images are funny, for example.
711.
by ukuina 3y ago
There are public bathhouses still around?
712.
by ukuina 3y ago
How does Eva compare to MindsDB ( https://mindsdb.com )?
713.
by ukuina 3y ago
Yep, but your retrieval from the vector DB becomes your relevancy bottleneck.
714.
by ukuina 3y ago
No, they are playing this close to the chest, similar to how OpenAI achieved 32k context limit.
715.
by ukuina 3y ago
I run Vicuna quite well with my M1 Pro, 32GB.
716.
by ukuina 3y ago
Midjourney does this, as well.
717.
by ukuina 3y ago
Yes, you can use a local embeddings model like gtr-t5-xl alongside retriever augmentation. This can point you in the right direction: https://haystack.deepset.ai/tutorials/22_pipeline_with_promp...
718.
by ukuina 3y ago
For context stuffing LLMs with small token limits(AKA Retriever Augmentation), you will need to break up each article in few-sentence chunks. You can get to a large number of these chunks very rapidly.
719.
by ukuina 3y ago
I use Octal on iOS. Highly recommended.
720.
by ukuina 3y ago
This assumes the pendulum will be allowed to swing back and not be held in place by AI tentacles.

More ›