WriterOfMinds

Atronach's Eye in Home Made Robots

Guess I haven't started a project thread for this one yet? Anyway, I put a bunch of work into the mechanical eyeball this year. It's been tougher to get it working perfectly than I expected (when is it not?), but I'm getting closer. Lots of photos and design discussion in the blog, so much I won't try to copy it all here: https://writerofminds.blogspot.com/2023/11/atronachs-eye-2023.html

Problems worked on this year and late last year include slack in the tendons, too much friction between ball and socket, limited range of motion and limit-of-motion sensing.

frankinstien

Running local AI models in AI Programming

I'm using Lm Studio as a server and have used it as an app as well, but the LLMs out there are outstanding! They are getting smaller and are competitive with online solutions like Replika! Also, the ability to operate these without NSFW filters makes them great when plotting to rob banks or murders! LoL Or at least creating a script or novel along those lines, even horror and intimacy interactions are off the charts!

So, the ability to do other types of local models such as voice where solutions like parler-tts and tortoise-tts have excellent voice abilities where you can even customize them to whoever's voice you like! Also, Whisper can do the opposite STT, and no censorship! Also, there are photo and video solutions like LLaVA-NeXT where the AI can create an impression or create images and videos based on prompts.

Here's the good part integrating these into a system that can see, hear, and imagine is a reality, taking each output and prompting the other provides for a kind of feedback approach to create...well, some might argue, but a persona. Enhancing the prompts with other types of data and even using some causality models we might just get that person from science fiction and all done from a PC!

What's required on the local machine is more than one GPU, where RTX 4070 ti supers are selling for $650, but you mix and match what you want where perhaps using an RTX 4090 for image and video is best and apply the RTX 4070 ti to do the rest. With three GPUS with just the minimum of an RTX 4070 ti that's 50GB of ram! But perhaps you need more since you may what a VR setup as well and give your bot a virtual body!

It's just freaking fantastic what is possible today and it's free from the clutches of politically correct censorship. Let your imagination go and apply your skills towards integration and you could very well build a very sophisticated competitor to ChatGPT 40 that runs at home.

Now what's a challenge is the development of a hardbody, animatronic facial expressions (the generated prompts from LLM models are freaking great, they could be used to control expressions and even position a body!)

It's a great time for the enthusiast, for a while now I thought everything was going to be locked up in the corporate cloud, controlled through a pay interface, and monitored by Big Brother, but America proves itself to be the land of freedom, and the industry has opened up to the little guy...

WriterOfMinds

Project Acuitas in General Project Discussion

Block diagram of Acuitas, the symbolic cognitive architecture:



<<< EDIT >>>
Since this is the post that pops up on the home page every time my thread gets bumped, I'm editing it to drop the most recent block diagram here. The info that was originally in this post, I will move into the second post of the thread.
<<< EDIT >>>

MagnusWootton

Hi IM BAA---AAACK!! in Home Made Robots

So this robot, is the one im going to build,   this is just an open s-cad render procedural generation cad script and it knocked my socks off how good it turned out!

Its a hydraulic one,  im not sure if my pipes and actuators are big enough,  but I'll be doing a test actuator first, (just the leg with a single actuator) this one is nearly done, just a bit more work to go before I go try to build it.

If I had access to a 2 metre 3d printer I could print it out pretty easy, (the body is only 20 centremeters but the legs are just too long) but I dont, so I've actually got a technique organized which is going to use a large big metre 2d print instead, which just mostly is just the pipes and actuators and then  build the shell off it from there,  in a top half and bottom half that just mirrors together,  should be pretty similar,   I just mainly need something that holds water, cause thats what hydraulic robots do,  just push water from the body to the arm.

So you can see the leg sorta looks like a snake, thats because on the 2d print its just flat straight out with the leg, but I actually bend and twist it into position afterwards with a heat gun and some extra braces im going to make to get it exactly the right angle, and thats how I get all the axi I need off a flat bar,    I'm not sure if it works that well for a humanoid, but since a spider is a pretty scary creature the extra twistiness to the legs I dont think makes it look less spidery, it sorta works for it.

It has a cool hinge configuration,  cause a) it can retract its leg closish to its body,  and b) its roll and yaw hinges interchange for each other as the pitch hinge changes angles.  so its always got access to yaw and roll,  but it swaps what hinge it uses for them depending on the prior pitch angle.


So a bit more work to go,  but I think is going to be the one!  7 years or so after I first wanted to go do this with lots of missing pieces in the jigsaw puzzle.

MikeB

LLaMA2 Meta's chatbot released in AI News

https://ai.meta.com/llama/
https://www.llama2.ai/

This release includes model weights and starting code for pretrained and fine-tuned Llama language models — ranging from 7B to 70B parameters.

Llama 2 outperforms other open source language models on many external benchmarks, including reasoning, coding, proficiency, and knowledge tests.

infurl

ollama and llama3 in AI News

Access to generative artificial intelligence just changed radically and for the better. Until recently our options were to use online services which were potentially very expensive and almost certainly heavily restricted, or to try to use open source models locally which required high end hardware to operate and which produced disappointing and mediocre results at best.

Last year we saw the release of ollama which made it incredibly easy to run just about any large language model locally no matter what platform you're on. You still needed a powerful system but at least you didn't have to learn a lot of obscure methods to use it.

https://ollama.com/

Last month the open source large language model llama3 was released. It has proven to be as capable as models two hundred times its size and is so efficient you can run it on a Raspberry Pi 5 if you want to, though it might take some patience.

I've been experimenting with it and it seems to be as good as any of the models that I have used online. I am running it on a Linux system with 24 cores, 64 GB of RAM, and 16 GB of video RAM. The smaller 8 billion parameter model responds to my queries almost instantly while the larger 70 billion parameter model can take a minute or two. Mostly the results produced by the smaller model are quite good enough.

WriterOfMinds

Attempting Hydraulics in Home Made Robots

I made a little home-built hydraulics system and did a writeup on it. It is NOT very capable yet, but it's a bare bones demo with my rather jury-rigged choice of parts.

https://writerofminds.blogspot.com/2023/10/poor-persons-first-hydraulic-system.html

infurl

Server Upgrade in Welcome to AI Dreams forum.

Hi folks, I'm in the process of upgrading the server that's hosting this forum so it may be temporarily inaccessible for short periods of time in the coming days. I'll keep the disruptions to a minimum.

ivan.moony

Reasoner.js: a framework for generalized theory synthesis in General Project Discussion

the idea

Experimentation with different custom made programming languages related to theorem proving has been one of my interests for a long time. Why theorem proving? Because it seemed like a logical step towards AI. In a meanwhile, deep learning ANNs (artificial neural networks) took over the world, and it could be a good thing. Similarity between an ANN and a theorem prover is that the starting point and ending result are the same in both cases, but the process in between is different. ANN utilizes pretty obscure weighted graph connecting input and output while theorem proving (in cases I'm considering) utilizes potentially human readable set of constructive theorems that need to be combined to connect input to output.

The theorem proving is still an area of my interest, but with addition of one missing link that ANNs seem to do pretty well (or at least in some extent). ANNs use the nowdays-all-present deep learning techniques to automatically form the connections between input and output. Analogously, the link I'm trying to explore is an alternative to deep learning in a form of automated semi-axiom construction. Here, I'm making a distinction between theorem proving and semi-axiom construction. In theorem proving, axioms need to be manually fed into the system to be used in further automated theorem construction. In semi-axiom construction, I'm seeking for a way to automatically construct semi-axioms, with semi-axioms being both axioms or theorems.

Googling around the web about this process (in 2022) gives somewhat rare and specialized results, so the space for a progress may be opened. I believe that appropriate term for the process I'm trying to explore could be "generalized theory synthesis". There are already some achievements on this process in a form of algorithm synthesis, and that is exactly direction I'm trying to head at, only applied also to theorems.

To finally explore the blurry cloud of theory synthesis and to see it in more focused features, the plan would be to fuse a selection of my existing languages, and present them in gradual form under the same language named Tricosm. The language should be applicable to theory synthesis, as well as to algorithm synthesis. In a simple example, from provided data:

input -> output
---------------
    2 -> 4
    4 -> 8
    8 -> 16
   16 -> 32

one of the generated theories (given that we already know how to handle integers and multiplication) would be:

input -> output
---------------
    x -> 2 * x

frankinstien

AI controlled F-16, for real! in AI News

Interesting article and the implementation of 1,000 aircraft controlled by AI is the tipping scale. Those nations that can harness AI don't need to sacrifice lives to win wars. Welcome to the game of World Domination, or more like the 1983 movie WarGames: "Shall We Play a Game?"