Debanjum

[Debanjum @Debanjum]

@sama Given o1 can scheme at an expert level, understanding its internal objective would have been useful for safety?

[Debanjum @Debanjum]

If an AI system can scheme at an expert level, understanding its internal objective isn’t just useful —it’s existential.

[khoj.dev @khoj_ai]

coming soon - android app.

https://t.co/vG3K2PYPFE

[khoj.dev @khoj_ai]

We found our new research mode allowed us to make remarkable gains on the base models. See "Iterative Retriever" below.

Khoj more than doubled the accuracy of gemini-1.5-flash from 27% to 64%. For reference, the model's ceiling on this benchmark is 66.5%.

[Justine Moore @venturetwins]

🚨 New @a16z thesis: building the "AI brain"

We all exist in our own context.

Is it possible to take your jumble of thoughts, history, and memories and distill it into something tangible?

How I exported my brain to AI + what I'm seeing from startups 👇

[Debanjum @Debanjum]

@khoj_ai @OpenAI Some tips to integrate o1 faster into your application

1. Streaming responses are not supported. So set `stream=false` in API query to use o1

2. System messages are not allowed. So convert it to a user message

3. Stop words, temperature etc. can't be configured

Happy hacking!

[Debanjum @Debanjum]

@khoj_ai now supports @openai's o1.

The 3 step plan
1. sign-up at https://t.co/0T4KNvodJ4
2. sync your reference papers and research notes with khoj
3. brainstorm your research ideas using o1

Go do novel research with your always available ai phd advisor!

[OpenAI @OpenAI]

We're releasing a preview of OpenAI o1—a new series of AI models designed to spend more time thinking before they respond.

These models can reason through complex tasks and solve harder problems than previous models in science, coding, and math. https://t.co/peKzzKX1bu

[Debanjum @Debanjum]

Our secret business model for the AI age:

First we booted up @khoj_ai, our business associate. Then we booted up our actual business.

[Debanjum @Debanjum]

@AdeptAILabs I created a Colab notebook (https://t.co/NEZ5zoOoP1) to play with Adept's new Fuyu model. Quantized the model to 4 bit. It runs pretty fast on the Colab T4 instance with GPU. Try it out and report any bugs :)

[Debanjum @Debanjum]

@AdeptAILabs I created a Colab notebook (https://t.co/NEZ5zoOoP1) to play with Adept's new Fuyu model. Quantized the model to 4 bit. It runs pretty fast on the Colab T4 instance with GPU. Try it out and report any bugs :)

[saba is researching @sabaimran_go]

Khoj is an open source, AI personal assistant. To improve developers' lives, we've been working on a Github integration that lets you chat with your repository.

[Debanjum @Debanjum]

@Kallang22 @arnevoelker @fortelabs @DannyHatcher Ah, that error is because Khoj needs python 3.10 or lower (because of Pytorch dependency).

And yes installation/setup isn't easy if you're not too comfortable with the terminal. Plan to simplify setup process by May (as currently working on Khoj Chat)

[Debanjum @Debanjum]

@arnevoelker @DannyHatcher @Kallang22 @fortelabs Umm, can you share what you've tried and where you're stuck?

Using the Khoj Obsidian plugin requires installing and running the Khoj backend. You can do that with:
pip install khoj-assistant & khoj --no-gui

For more details see the setup instructions at https://t.co/5SaTy9fVFy

[Debanjum @Debanjum]

Testing sequence completed. All ground stations prepare for lift-off

[Debanjum @Debanjum]

..9999999...10000000.

[Debanjum @Debanjum]

Testing 123...