News for hackers. Stuff that matters.

nsky-world

Show HN: Implementing CPU Logic into LLMs

https://github.com/turing-machines/mentals-ai

For the past four months, we’ve been developing a framework that allows the creation of autonomous AI agents without programming.Inspired by the control unit in classical processors and Von Neumann architecture, we see LLMs as complex ALUs but lacking a unit for handling complex prompt sequences, including loops. We implemented a prompt that acts as the central executive system, managing...

( Read More... | no comments | 2 points | on HN )
Re:OSS AI agents with ease. Just add language. on Wednesday June 05, 2024 @01:19PM (#40587703)
For the last four months we have been experimenting a lot with LLM models in an attempt to simplify the creation of reasoning chains without using an auxiliary programming language. The idea comes from the control unit in classical processors. We see LLM as a more complex ALU, but the models lack a unit that can perform more complex sequences of prompts, including loops. In psychoanalysis there is a concept of the central executive system, which is somewhat similar to the control unit in CPUs. > The central executive is responsible for controlled processing in working memory, including but not limited to, directing attention, maintaining task goals, decision making, and memory retrieval.
it is nice that you posted datasets

Show HN: Code-Free AI Agents in Minutes

https://twitter.com/the_nsky/status/1790802387961266451

Here’s a quick demo of a team of agents that I build without a single line of code:The team is: root: Fetches 3 top stories from HackerNews researcher: Picks one story and creates a content plan for the writer writer: Writes a blog post critic: Provides feedback on the blog post The writer and critic engage in a multi-round dialogue, exchanging feedback and blog post iterations. Once complete,...

( Read More... | no comments | 2 points | on HN )
Re:Build Dreams, Not Code on Tuesday February 27, 2024 @01:00PM (#39527222)
Create AI apps and automations using just natural language. No coding required.
Re:10M Tokens LLM Context on Friday February 02, 2024 @12:52PM (#39231783)
KVQuant: Towards Enabling 10 Million Context Length For LLM Inference through KV Cache Quantization

Show HN: Turing RK1 – 32GB RAM and AI Accelerator

https://turingpi.com/product/turing-rk1/

A year ago, we launched the Turing Pi2, a 4-node cluster computer with a unique modular and heterogeneous architecture. Despite the excitement, we faced a significant challenge: a shortage of supported Single Board Computers (SBCs), including the Raspberry Pi CM4 and NVIDIA Jetson modules.In response, we made a promise to our community to deliver our own compute module, aiming to address the...

( Read More... | no comments | on HN )
Re:Cluster Computer Raises $1M in 12H on Tuesday May 17, 2022 @11:48PM (#31419050)
$400k in 30 minutes!
Re:Hosting this website on a farm – or anywhere on Wednesday February 09, 2022 @10:52AM (#30274098)
The next step would be to send Turing Pi cluster to space!
Re:Turing Pi 2: 4 Raspberry Pi nodes on a mini ITX board on Wednesday December 01, 2021 @12:20PM (#29406405)
We are waited for this for soooo loooong!!
Re:CI/CD for ARM Development on Wednesday January 20, 2021 @10:07AM (#25847242)
This article is in depth review of a CI/CD solution for ARM development. Tolls used GitLab CI, Ansible and Turing Pi cluster.
Re:The Edge at Arm's Length on Wednesday October 21, 2020 @02:05PM (#24850419)
The low cost of the Raspberry Pi has disrupted many facets of general computing, but that also extends to cluster computing. Using a series of Raspberry Pi Model B boards is possible, but with the new range of Raspberry Pi Compute Modules, we can build clusters that offer the same or greater computational power while consuming less space. The Turing Pi 1 [1] from Turing Machines used seven Compute Module 3+, but with the announcement of the Raspberry Pi Compute Module 4, the company has announced [2] Turing Pi 2, which will support up to four Compute Module 4 boards.

[1] https://turingpi.com/v1/ [2] https://turingpi.com/turing-pi-2-announcement/

V1 is $189
We are expecting the price to be cheaper than the V1, but the final price will be announced later this year.
The Turing Pi comes in a standard mini ITX form factor
That's one of the point too. We have industrial customers who deploy apps in the places with a limited access to the internet, so increased reliability something important to them. Another thing is that they can host apps with different purposes on different nodes in containers. Besides that all the infrastructure could be managed with a tool like Kubernetes and organically work with existing cloud infrastructure. And the last but not least in a cluster like that you can combine general purpose modules with accelerated compute modules.
You can. Why not?
So your question what's the advantage of cluster comparing to one processor unit?
2x Gen2x1
This is a great case study of how the Turing Pi helps with CI/CD for ARM infrastructure https://synyx.de/blog/turing-pi-gitlab-ci-ansible/

Besides that the main direction here is edge computing.

Totally agree

Older »