Most important news of the last week:

- Anthropic increased context window of its large model to 100k tokens (3x times of ChatGPT-4-32k):
- Top 3 GPT models are: GPT-4, Anthropic Claude, GPT-3 (according to the new leaderboard from LMSYS.org)
- ChatGPT Plus subscribers get access to plugins and browsing.
- Microsoft is opening its Github Copilot Chat to early beta test. Its prompt is already leaked 😂

Read more in the next "ML Under the Hood" newsletter: https://abdullin.substack.com/p/learning-fast-failing-forward

Let’s catch up on ChatGPT and ML-driven products! We’ll talk about:

- A few notable large language model releases
- Why all of these models are already irrelevant
- Product-specific LLM benchmarks
- On the potential of ChatGPT
- Answering GPT-related questions

Read the newsletter online (no subscription needed): https://abdullin.substack.com/p/on-new-llms-their-irrelevance-and

Did you ever want to talk to your knowledge base or a journal?

Here is how to to build that with ChatGPTin less than 30 lines of Python code.

permalink: https://abdullin.com/llm/talk-to-your-knowledge-base/

I found a way to save 30min of doom scrolling and anxiety every day!

Solution: stop reading news and deploy an agent to track news-sites according to MY interests.

It would summarise relevant ones and send update in time for the morning coffee. Or will tell: “I scanned 55 trending news articles today, nothing worthy of your attention.”

Check your inbox for the next newsletter on building ML-driven products!

- News Recap
- LangChain and LlamaIndex are awesome
- AI-driven businesses will come

You can also find it in the archive: https://tinyletter.com/abdullin

Check your inbox for the next newsletter. This time on Meta LLaMa, OpenAI Turbo, and making ML product provide answers in 2-10 seconds.

https://tinyletter.com/abdullin

The cat is out of the bag! Facebook LLAMA is being openly distributed via torrents

https://news.ycombinator.com/item?id=35007978

Hey, are you using ChatGPT to automate some mundane part of your day?

I use it to write formal letters to authorities or come up with boilerplate code snippets to scaffold my projects faster. I love to hear more ideas.

Check your inbox for next newsletter 🗞️

This time I’m writing about ML inference pipeline for a hardware prototype. Like Siri or Alexa, but covers a small native language that tech giants never care about.

Also available in archive here: https://tinyletter.com/abdullin/