วีดีโอ

hahahaha!

@@OminousIndustries hahahahahahaha I want bigman as my personal trainer, the fucking master!! and you are e genious, i love your work!

@@mquant001 Thanks for the kind words! Yes, that guy is hilarious af hahaha

Not that I have seen. Unless you had an EXTREMELY powerful gpu it would take too long for generation to get anything running close to real time with this. The videos I generated took multiple minutes each to do on a single 3090ti.

I have heard the same, especially as these are older ones. I think I will end up using them for something similar to what you mention, though I do not yet know what.

hahah I feel that!

Thanks very much! That's exciting about the new machine, what did you get? Having gpus in a server rack would be pretty cool.

@@OminousIndustries leaning towards the Lenovo P520. New CPU and enough slots of dual NVMes. 900psu. All for very reasonable.

@@TheInternalNet That is a beastly machine, would be an awesome AI Build!

That is a very good idea!

I also experienced slowness on my 8gb M2, it just seems 16 is necessary for current models on Apple Silicon.

LOL

Thanks for the kind words!

Thanks very much, I am not really familiar with anything home-assistant related, what in specific would you like to implement gpt4o with one to do? I would be happy to at least look into the topic!

LOL

I noticed this as well, it seemed in this small sample size, gemini outperformed gpt.

Thanks very much!

HAHAHA!

Yes, that's why I mentioned it would be frustrating at the beginning !!

Ohhh, I have many videos of robots arguing but they're too NSFW to post hahaha!

HAHAHAHA

Don't give me any more ideas....

🤣 What?

Same!

Thanks for the kind words! I will likely make more tutorial related videos. I want to for windows and linux as well, but those require a bit more time as there are a lot more potential hardware configurations to be considered.

Thanks for sharing about Smart Vision. It sounds super cool how it can identify unusual elements in photos and analyze clothing fit. The examples you gave are fascinating and this is impressive work and testing!

@@OminousIndustries yeah theres another random image i uploaded and basically it identified a tiny little rabbit on a rabbit food box, but the image i uploaded was full of unusual things and it identified this when i took the pic 10 m away at-least, so crazy, but its one thing i guess open ai forgot to think of, idk

@@xd-qi6ry Wow, that is impressive especially considering the distance you took it from

@@OminousIndustries yeah but have no way of publicising it so idk, but I’ve kept the instructions secret and there is some genius tricks that I used to add more info but at the. End of the day it was a few weeks of hell and experimenting. I had already created a problem solver one that could already solve physics problems but this one is special.

@@OminousIndustries ok weirdly it just the international math olympiad question from a youtube video, (channel Tibees) titled chat gpt (gpt4) vs the world hardest exam. And base 4o got it completely wrong the same answer as old gpt 4

LOL

I'm beginning to believe Blake Lemoine....

Soon I will sell them as a BYOP kit (bring your own pi) LOL

Not in this iteration, they are sort of just left to handle the conversation loop in a more "natural" way by just sending some udp messages to help trigger the corresponding "speak" or "listen" state. If I do something more with this I will implement something more robust so they can speak longer and fully in turn.

I think gemini does that, chatgpt interrupts a lot though

The Gemini model was hilarious in the way it was analyzing the gpt4o Bobs behaviors

@@OminousIndustries yes. Kept commenting on the way it was responding. Then interrupting.

@@TheInternalNet The interrupting was my fault, I kind of let them speak "naturally" instead of strictly implementing state changes

@@OminousIndustries regardless that's really amazing stuff. Now I'm dying to create something similar

@@TheInternalNet It's a lot of fun to play with them like this and if you have an old Pi lying around it's not overly difficult to set something like this up.

I believe the consensus is still that dual 24gb cards (like the 3090) is the best "budget move", however, I would head over to: www.reddit.com/r/LocalLLaMA/ and browse/ask there as there are a lot of knowledgeable people in there who can provide good insight on this. For what it's worth, I don't believe it will make a huge difference if you get non Ti cards, I just bought them as I wanted to match the first card I had which happened to be a Ti.

I want to do a longer version with them, using more entertaining personas...

@@OminousIndustries that is really amazing you have figured out how to give them such... Personalities. I'm really impressed.

@@TheInternalNet Thanks very much!

Thanks very much!

@@OminousIndustries you are very welcome. Would absolutely love to see a thorough video of how to setup the LLM for a complete beginner. I understand that's a big undertaking but would be so amazing.

@@TheInternalNet That's a great idea, and thanks to some of the open source local llm tools it's not actually a prohibitively difficult thing to go over in full. I will gladly make that video!

This was Llama 3 70b 4.0 exl2

I was also impressed with the M2 Pro. Yeah, I use them for different purposes, sans the M1 mini. I have replaced that one with the M2 Pro mini and I keep it around for random use cases like this.

That’s cool, and I haven’t found any other videos on TH-cam doing this like you did

@@sertenejoacustic Thanks very much!

Yes I was impressed with them myself. The 8gb ram that comes standard on the base models is shameful though.

@@OminousIndustries 16 is pushing it. 32 for that effing price should be standard.

@@TheInternalNet lol. 128gb and 256 cores should be standard. and 2x bigger battery and titanium body, just like all the dells are giving ya.

Thanks very much!

I appreciate your input RE the design preference as this has been on my mind lately as well. I will in about a month to a month and a half have something like Bob that can run locally offline completely plug and play. I have been wanting to do something like that for ages but the hardware to pull it off was really not economically feasible to be included as part of a product until far more recently. I am also considering selling the parts for this build (sans the pi) as a cheap kit for any LocalLLM user to purchase and add their Pi to get an offline buddy like this. I have some work cut out for me!

@@OminousIndustries That's great to hear! I discovered your videos a little while back because I had wanted something plug and play and the R1 is the closest I had seen. Love the vids, looking forward to seeing more!

@@miketrago4561 Thanks very much! I am also very invested in totally offline, no subscription plug and play, I will post more videos pertaining to that in the near future!

I will likely do that as my next video, I will also give this guy some mouth movements.

Hahaha! Thanks for the kind words!

Thank you for the support and kind words!

I am flatttered by this, thank you!

Thanks! I actually built it as a product to sell, you can get one for $199 on www.ominousindustries.com, If you wanted to make something like that you could do it by just combining some AI back end with a TTS/STT workflow.