Hacker News: KhoomeiK

New comment by KhoomeiK in "Diffusion models are real-time game engines"

KhoomeiK — Wed, 28 Aug 2024 17:30:15 +0000

NVIDIA did something similar with GANs in 2020 [1], except users could actually play those games (unlike in this diffusion work which just plays back simulated video). Sentdex later adapted this to play GTA with a really cool demo [2].

[1] https://research.nvidia.com/labs/toronto-ai/gameGAN/

[2] https://www.youtube.com/watch?v=udPY5rQVoW0

Show HN: Multimodal Code Generation for Web Data Extraction

KhoomeiK — Tue, 09 Jul 2024 18:19:46 +0000

Article URL: https://twitter.com/khoomeik/status/1810727750313070647

Comments URL: https://news.ycombinator.com/item?id=40919203

Points: 2

# Comments: 0

New comment by KhoomeiK in "Introduction to Program Synthesis"

KhoomeiK — Thu, 04 Jul 2024 02:16:38 +0000

Everything relevant in "program synthesis" moved to the new buzzword "codegen"

New comment by KhoomeiK in "Show HN: Revideo – Create Videos with Code"

KhoomeiK — Tue, 11 Jun 2024 17:45:28 +0000

Interesting—LangChain seemed kinda like unnecessary abstractions in natural language (since everything is just string manipulations), but with AI video, there's so many different abstractions that I'd need to handle (images, puppeting, facegen, voicegen, etc).

Seems like there might be room for a "LangChain for Video" in this space...

New comment by KhoomeiK in "Atash Behram – Types of Fire"

KhoomeiK — Wed, 05 Jun 2024 18:49:02 +0000

Just unrolled the thread for you here!

https://threadreaderapp.com/thread/1794082465398812770.html

New comment by KhoomeiK in "Atash Behram – Types of Fire"

KhoomeiK — Wed, 05 Jun 2024 06:08:25 +0000

Thanks! I have no idea—unfortunately, very few Hindus maintain the Vedic fire rites. There are also no active central authorities on matters of Vedic ritual. The only plan of now is to use this interpretation in my own yajña practice.

New comment by KhoomeiK in "Atash Behram – Types of Fire"

KhoomeiK — Wed, 05 Jun 2024 02:52:41 +0000

Vedic Hinduism had a similar concept of eternal fire. I recently wrote up a twitter thread [1] explaining how the modern interpretation of Vedic instructions on starting these sacred fires misunderstands the text.

Etymology tidbit: "Bhārata", India's Sanskrit name, refers to the forerunner clan that established India's first historically recorded political entity—the Kuru Kingdom—around 1200 BC near modern Delhi. The clan itself was named "Bhārata" due to their ardent bearing ("bhar-" in Sanskrit) of the sacred fire.

[1] https://x.com/khoomeik/status/1794082465398812770

Show HN: Chinchilla Scaling Laws Are Not Universal

KhoomeiK — Tue, 28 May 2024 18:48:42 +0000

Hey HN! Chinchilla (DeepMind 2022) tells us that when we scale up our language model training, we should scale the parameters and data equally.

Over the last several months I've been hacking on a research project to determine if the optimal compute allocation (scaling law) for training an LLM is sensitive to training data complexity. I found that as data complexity increases, you need even more data than Chinchilla suggests!

I released the preprint just yesterday: https://arxiv.org/abs/2405.16684

Comments URL: https://news.ycombinator.com/item?id=40504084

Points: 1

# Comments: 0

Scaling Laws Depend on Data Compressibility

KhoomeiK — Tue, 28 May 2024 17:35:23 +0000

Article URL: https://twitter.com/khoomeik/status/1795477359933706272

Comments URL: https://news.ycombinator.com/item?id=40503250

Points: 1

# Comments: 0

Gzip Predicts Data-Dependent Scaling Laws

KhoomeiK — Tue, 28 May 2024 17:28:15 +0000

Article URL: https://arxiv.org/abs/2405.16684

Comments URL: https://news.ycombinator.com/item?id=40503161

Points: 2

# Comments: 0

SB-1047 regulates all LLM's after training a suboptimal 1e26 FLOP model

KhoomeiK — Sun, 26 May 2024 23:02:21 +0000

Article URL: https://twitter.com/khoomeik/status/1794751305615077767

Comments URL: https://news.ycombinator.com/item?id=40486183

Points: 4

# Comments: 0

How to kindle a fire: solving an 800 year old puzzle in Vedic ritual exegesis

KhoomeiK — Fri, 24 May 2024 19:09:48 +0000

Article URL: https://twitter.com/khoomeik/status/1794082465398812770

Comments URL: https://news.ycombinator.com/item?id=40469500

Points: 1

# Comments: 0

How to kindle a fire: solving an 800 year old puzzle in Vedic ritual exegesis

KhoomeiK — Fri, 24 May 2024 08:18:28 +0000

Article URL: https://twitter.com/khoomeik/status/1793917692342358338

Comments URL: https://news.ycombinator.com/item?id=40464097

Points: 2

# Comments: 1

New comment by KhoomeiK in "Show HN: Tarsier – Vision utilities for web interaction agents"

KhoomeiK — Thu, 16 May 2024 17:02:00 +0000

Maybe this project another commenter is working on?

https://news.ycombinator.com/item?id=40373310

New comment by KhoomeiK in "Show HN: Tarsier – Vision utilities for web interaction agents"

KhoomeiK — Thu, 16 May 2024 17:00:12 +0000

Awesome pics! We love tarsiers too

New comment by KhoomeiK in "Show HN: Tarsier – Vision utilities for web interaction agents"

KhoomeiK — Thu, 16 May 2024 01:04:36 +0000

Great question! See this thread:

https://news.ycombinator.com/item?id=40369713

New comment by KhoomeiK in "Show HN: Tarsier – Vision utilities for web interaction agents"

KhoomeiK — Thu, 16 May 2024 01:03:52 +0000

Yes it does work headless and we do grab a fullpage screenshot including scrolling (by resizing viewport to content height). We haven’t had to deal with infinite scrolling much but that’s an interesting feature we’d appreciate a PR for.

We haven’t tried Apple’s OCR but hopefully will integrate Azure OCR soon based on others’ advice.

New comment by KhoomeiK in "Show HN: Tarsier – Vision utilities for web interaction agents"

KhoomeiK — Wed, 15 May 2024 20:23:49 +0000

They do show textboxes with labels. From our readme:

"Keep in mind that Tarsier tags different types of elements differently to help your LLM identify what actions are performable on each element. Specifically:

[#ID]: text-insertable fields (e.g. textarea, input with textual type)

[@ID]: hyperlinks ( tags)

[$ID]: other interactable elements (e.g. button, select)

[ID]: plain text (if you pass tag_text_elements=True)"

Do you see the search boxes labeled [#4] and [#5] at the top? And before you say that the tag is on a different line from the placeholder text—yes, and our agent is smart enough to handle that minor idiosyncrasy. Are you shocked? :)

New comment by KhoomeiK in "Show HN: Tarsier – Vision utilities for web interaction agents"

KhoomeiK — Wed, 15 May 2024 20:15:36 +0000

We have a lot more powerful use-cases for Tarsier in web data extraction at the moment. Stay tuned for a broader launch soon!

New comment by KhoomeiK in "Show HN: Tarsier – Vision utilities for web interaction agents"

KhoomeiK — Wed, 15 May 2024 19:44:55 +0000

We run OCR on the screenshot & convert it to whitespace-structured text, that is passed to the LLM. The images below might make it clearer for you:

[1] https://github.com/reworkd/tarsier/blob/main/.github/assets/...

[2] https://github.com/reworkd/tarsier/blob/main/.github/assets/...