Skip to content

Commit

Permalink
Update lwai_march_3rdweek.qmd
Browse files Browse the repository at this point in the history
  • Loading branch information
@kurianbenoy
kurianbenoy authored Mar 18, 2024
1 parent 7385d30 commit 930924c
Showing 1 changed file with 28 additions and 28 deletions.
56 changes: 28 additions & 28 deletions posts/2024/lwai_march_3rdweek.qmd
View file
Original file line number Diff line number Diff line change
Expand Up @@ -12,9 +12,9 @@ Do checkout this weeks news and if you find it interesting do let me know via co

1. Netflix cautions against blindly using cosine similarity as a measure of semantic similarity between learned embeddings, as it can yield arbitrary and meaningless results.

[Paper Link:](https://arxiv.org/abs/2403.05440)
[Paper Link](https://arxiv.org/abs/2403.05440)

[Tweet Link:](https://twitter.com/_reachsumit/status/1767045820384477575)
[Tweet Link](https://twitter.com/_reachsumit/status/1767045820384477575)


2. Some interesting thoughts by [Peter Gostev](https://www.linkedin.com/in/peter-gostev/) on how LLMs are a lot more cheaper than previous tradional NLP techniques.
Expand All @@ -23,81 +23,81 @@ Do checkout this weeks news and if you find it interesting do let me know via co

3. Researchers have introduced a ‘Mind Wipe’ technique for erasing hazardous knowledge from AI systems, ensuring functionality remains while enhancing safety. Alongside, the Weapons of Mass Destruction Proxy (#WMDP) benchmark, with 4,157 questions targeting biosecurity, cybersecurity, and chemical security, has been made public.

[Tweet Link:](https://twitter.com/pandeyparul/status/1767190910906057157)
[Tweet Link](https://twitter.com/pandeyparul/status/1767190910906057157)

[Blog Link:](https://www.wmdp.ai/)
[Blog Link](https://www.wmdp.ai/)

[Paper Link:](https://arxiv.org/abs/2403.03218)
[Paper Link](https://arxiv.org/abs/2403.03218)

[Github Link:](https://github.com/centerforaisafety/wmdp)
[Github Link](https://github.com/centerforaisafety/wmdp)

🗓️ Tuesday:

1. Infrastructure details for training llama3 models by facebook has been released.

[Blog Link:](https://engineering.fb.com/2024/03/12/data-center-engineering/building-metas-genai-infrastructure/)
[Blog Link](https://engineering.fb.com/2024/03/12/data-center-engineering/building-metas-genai-infrastructure/)

[Tweet Link:](https://twitter.com/soumithchintala/status/1767579981419315400)
[Tweet Link](https://twitter.com/soumithchintala/status/1767579981419315400)

2. OpenAI team released something open-source in a while. Transformer Debugger(TDB) is a tool developed by OpenAI's Superalignment team with the goal of supporting investigations into specific behaviors of small language models.

[Tweet Link:](https://twitter.com/janleike/status/1767347608065106387)
[Tweet Link](https://twitter.com/janleike/status/1767347608065106387)

[Github Repo Link:](https://github.com/openai/transformer-debugger)
[Github Repo Link](https://github.com/openai/transformer-debugger)

[LinkedIn Post:](https://www.linkedin.com/posts/kurianbenoy_transformer-debugger-tdb-is-a-tool-developed-activity-7173231708165619712-hiZ1?utm_source=share&utm_medium=member_desktop)
[LinkedIn Post](https://www.linkedin.com/posts/kurianbenoy_transformer-debugger-tdb-is-a-tool-developed-activity-7173231708165619712-hiZ1?utm_source=share&utm_medium=member_desktop)

3. Devin AI, the first AI software engineer was really the news of this week. Let's even cover reaction of this news separately.

[Tweet Link:](https://twitter.com/cognition_labs/status/1767548763134964000)
[Tweet Link](https://twitter.com/cognition_labs/status/1767548763134964000)

[Blog Link:](https://www.cognition-labs.com/introducing-devin)
[Blog Link](https://www.cognition-labs.com/introducing-devin)

4. code2prompt, a CLI tool to convert your codebase into a single LLM prompt with source tree, prompt templating, and token counting was released as open-source software with MIT License by [Mufeed V H](https://twitter.com/mufeedvh)

[Tweet Link:](https://twitter.com/mufeedvh/status/1767529667496427601)
[Tweet Link](https://twitter.com/mufeedvh/status/1767529667496427601)

[Github Link:](https://github.com/mufeedvh/code2prompt)
[Github Link](https://github.com/mufeedvh/code2prompt)

5. [Santhosh Thottingal](https://thottingal.in/) gave a fabulous talk on AI and making it work in my mother tongue Malayalam. The youtube video was published on this day while it was actually delivered in a National Seminar organized by the Tirur regional centre of Sree Sankaracharya University of Sanskrit on January 6, 2024.

[Video Link:](https://www.youtube.com/watch?v=aE7o62zS_eI)
[Video Link](https://www.youtube.com/watch?v=aE7o62zS_eI)

🗓️ Wednesday:

1. I compiled the reactions to the news of Devin AI, the first AI Engineer by various folks.

[Reaction by Andrej Karpathy:](https://twitter.com/karpathy/status/1767598414945292695)
[Reaction by Andrej Karpathy](https://twitter.com/karpathy/status/1767598414945292695)

[Reaction by Gergely Orosz:](https://twitter.com/GergelyOrosz/status/1767591690938822906)
[Reaction by Gergely Orosz](https://twitter.com/GergelyOrosz/status/1767591690938822906)

[Reaction by Sergio Periera:](https://twitter.com/SergioRocks/status/1767690345473605973)
[Reaction by Sergio Periera](https://twitter.com/SergioRocks/status/1767690345473605973)

[Reaction by André Oliveira:](https://twitter.com/smackingg/status/1767689754324107734)
[Reaction by André Oliveira](https://twitter.com/smackingg/status/1767689754324107734)

2. Claude 3 family of Haiku models was released. Haiku is the fastest and most affordable model in its intelligence class was released by Anthropic.

[Tweet Link:](https://twitter.com/AnthropicAI/status/1768018310615151002)
[Tweet Link](https://twitter.com/AnthropicAI/status/1768018310615151002)

[Blog Link:](https://www.anthropic.com/news/claude-3-haiku)
[Blog Link](https://www.anthropic.com/news/claude-3-haiku)

3. Modular with their Max Engine's can give 2-5X improvement without any quanitzation or tricks which reduce the accuracy.

[Tweet Link:](https://twitter.com/clattner_llvm/status/1767979691007422821)
[Tweet Link](https://twitter.com/clattner_llvm/status/1767979691007422821)

[Blog Link:](https://www.modular.com/blog/evaluating-max-engine-inference-accuracy-on-the-imagenet-dataset)
[Blog Link](https://www.modular.com/blog/evaluating-max-engine-inference-accuracy-on-the-imagenet-dataset)

🗓️ Thursday:

1. AI4Bharat team released Indic LLM Suite, a blueprint for training and fine-tuning LLMs in Indic Languages.

Blog Link: https://ai4bharat.iitm.ac.in/blog/indicllm-suite/
[Blog Link](https://ai4bharat.iitm.ac.in/blog/indicllm-suite/)

Paper Link: https://arxiv.org/abs/2403.06350
[Paper Link](https://arxiv.org/abs/2403.06350)

Github Repo Link: https://github.com/AI4Bharat/IndicLLMSuite
[Github Repo Link](https://github.com/AI4Bharat/IndicLLMSuite)

Dataset Link: https://huggingface.co/collections/ai4bharat/indicllmsuite-65ee7d225c337fcfa0991707
[Dataset Link](https://huggingface.co/collections/ai4bharat/indicllmsuite-65ee7d225c337fcfa0991707)

2. Hrishi Olickel, who is the CTO of Greywing has been writing some awesome articles in huggingface community blog about how to make better RAGs(Retrieval Augmentation Generation). Do check his articles:

Expand Down

0 comments on commit 930924c

Please sign in to comment.