jarulraj

7 ms·

1. 31.▲
   by jarulraj 3y ago

   Yes, it should be "positive%". That's an interesting LLM-level optimization. Doesn't this token sampling optimization require using a locally-running model like Llama? I am presuming that OpenAI doesn't provide dire
2. 32.▲
   by jarulraj 3y ago

   Interesting observation :) I guess this is the SQL query you have in mind that uses the LIKE operator: SELECT ChatGPT("Respond to the review with a solution to address the reviewer's concern", review) FROM postgres_
3. 33.▲
   by jarulraj 3y ago

   Congrats on the launch, @ukuina! Neat idea. Would love to learn more about how you are querying the summaries/comments database and how you are planning to extend the queries in this app. Shameless self-plug: We are building EvaDB [1],
4. 34.▲
   by jarulraj 3y ago

   Really cool project! I am curious about the benefits and limitations of using Python in this project as opposed to C++. How well is LearnDB able to support low-level concurrency control and storage management?
5. 35.▲
   by jarulraj 3y ago

   Highly recommend Lark for Python projects -- it's easy to use :) Their IDE was super useful for debugging the grammar: https://www.lark-parser.org/ide/ We use Lark for a SQL-like language tailored for using AI mod
6. 36.▲
   MindsDB vs. EvaDB: Connecting AI Models to Database Systems(medium.com)

   12 points by jarulraj 3y ago|0 comments
7. 37.▲
   Ask HN: How has your experience been with using AI Agents?

   5 points by jarulraj 3y ago|5 comments
8. 38.▲
   by jarulraj 3y ago

   Interesting observation, @swyx. There seems to be a connection to transitive closure in SQL queries, where the output of the query is fed as the input to the query in the next iteration [1]. We are thinking about how to best support such re
9. 39.▲
   by jarulraj 3y ago

   Neat idea, @shishirpatil! We are developing EvaDB [1] for shipping simpler, faster, and cost-effective AI apps. Can you share your thoughts on transforming the output of the Gorilla LLM to functions in EvaDB apps -- like this function that
10. 40.▲
    PrivateGPT in SQL(github.com)

    7 points by jarulraj 3y ago|0 comments
11. 41.▲
    by jarulraj 3y ago

    Interesting observation, @gibsonf1. What kind of hierarchical queries do you have in mind that would be challenging to write in SQL? Would you mind sharing them in natural language?
12. 42.▲
    by jarulraj 3y ago

    We also love SQL, @potatoman22 :) BTW, EVA itself is designed for analyzing sports games -- like touchdowns in a football game. This notebook shows how an action recognition model can be run on an American Sign Language video to identify th
13. 43.▲
    by jarulraj 3y ago

    Great question, @ukuina! We will need some time to do a detailed analysis of MindsDB. But here are some key differences a first glance: 1. First-class support for unstructured data: EVA natively supports querying over videos, audio, etc. Fo
14. 44.▲
    by jarulraj 3y ago

    Haha. Thanks for sharing this reference! I had no idea about a similar comment [1] in our previous Show HN post :) For others like me, this is ChatGPT's explanation: C&C Tiberian Sun, short for Command & Conquer: Tiberian
15. 45.▲
    by jarulraj 3y ago

    Great question! We have a built-in integration test for 1 million images [1] and have benchmarked EVA on large unstructured datasets for our ACM SIGMOD paper [2]. Coming up with an open benchmark for AI-Relational database systems is essent
16. 46.▲
    by jarulraj 3y ago

    Hey friends We wanted to express our heartfelt gratitude to all of you for your support for the EVA AI-Relational Database System last week [1]. The feedback from the HN community has been truly overwhelming :) We were planning to make an H
17. 47.▲
    by jarulraj 3y ago

    We are in the process of supporting a native `LOAD PDF` command. Meanwhile, you could convert each PDF into a series of images and load them using the `LOAD IMAGE` command. You could then run any text extraction user-defined function (e.g.,
18. 48.▲
    by jarulraj 3y ago

    Thanks for the clarifications. Token summation is a cool optimization :) Query optimizers in SQL database systems typically optimize based on the time to execute the function on a local server. The token summation optimization generalizes t
19. 49.▲
    by jarulraj 3y ago

    Thanks so much for sharing your thoughts! I also felt that they are pretty similar. But, I am guessing that SQLite (similar to most relational database systems) does not automatically cache the results of functions, do non-trivial cost-base
20. 50.▲
    by jarulraj 3y ago

    Great question! Besides improving usability, the key feature of the EVA database system is the query optimizer that seeks to speed up exploratory queries over a given dataset and save money spent on inference. Two key optimizations in EVA&#
21. 51.▲
    by jarulraj 3y ago

    Great question! Yes, EVA supports NLP pipelines thanks to our recent integration of Hugging Face pipelines last month. Here is an illustrative text classification application: -- Text classification application in EVA CREATE TABLE IF
22. 52.▲
    by jarulraj 3y ago

    Hey Simon! Thanks for sharing this. I have long admired your work on Datasette :) We will check out your posts for ideas on coping with prompt injection. I just came across your recent post on the ChatGPT SQL function in SQLite [1]. We just
23. 53.▲
    by jarulraj 3y ago

    Thanks for the clarification. EVA currently does not support remote GPUs. Our ongoing integration [1] of the Ray distributed compute framework [2] into EVA will soon allow us to support remote GPUs. We would love to jointly explore how to b
24. 54.▲
    by jarulraj 3y ago

    Great questions! EVA does not currently support a weighted similarity search. We are working on creating a notebook to illustrate similarity queries. But, EVA already supports the queries of this form: -- Step 1: Extract objects in Redd
25. 55.▲
    by jarulraj 3y ago

    Yes, EVA works out of the box on an AWS server with GPU [1]. [1] https://aws.amazon.com/nvidia/
26. 56.▲
    by jarulraj 3y ago

    Thanks for the helpful suggestion! Just reordered the examples in the README. Here are the illustrative queries: -- Object detection in a surveillance video SELECT id, YoloV5(data) FROM ObjectDetectionVideos WHERE id < 20 -
27. 57.▲
    by jarulraj 3y ago

    Thanks for following up on this. We designed EVA from scratch for managing unstructured data (e.g., video, audio, images, etc.). EVA leverages relational database systems to manage structured data and widely-used libraries to manage featur
28. 58.▲
    by jarulraj 3y ago

    Thanks for the helpful suggestion! EVA uses an SQL database system for managing structured data using sqlalchemy. It runs on PostgreSQL out of the box. You only need to provide the database connection url in the EVA configuration file. Than
29. 59.▲
    by jarulraj 3y ago

    That’s correct! You can compose multiple models in a single query to set up useful AI pipelines. Here is an illustrative query that chains together multiple models: -- Analyse emotions of faces in a video SELECT id, bbox, EmotionDet
30. 60.▲
    by jarulraj 3y ago

    Honestly, we have not extensively thought about prompt injection attacks -- the equivalent of SQL injection attacks in AI-Relational database systems :) If you have any thoughts on addressing this, please do share! We will incorporate that