Hugging Face's logo

Turing's Solutions

company
https://www.turingssolutions.com
Activity Feed Request to join this org

AI & ML interests

AI Ethics

Turings-Solutions's activity

lunarflu 
posted an update about 1 month ago
TuringsSolutions 
posted an update about 2 months ago
view post
Post
362
Maybe that post I showed the other day with my Hyperbolic Embeddings getting to perfect loss with RAdam was a one-time fluke, bad test dataset, etc.? Anotha' one! I gave it a test set a PhD student would struggle with. This model is a bit more souped up. Major callouts of the model: High Dimensional Encoding (HDC), Hyperbolic Embeddings, Entropix. Link to the Colab Notebook: https://colab.research.google.com/drive/1mS-uxhufx-h7eZXL0ZwPMAAXHqSeGZxX?usp=sharing
  • 26 replies
·
TuringsSolutions 
posted an update about 2 months ago
view post
Post
884
I created something called 'Hyperbolic Embeddings'. I literally just embed the tokens into Hyperbolic Space instead of Euclidean space. At first, this did not get me the gains I was expecting. I was a sad panda. Then I thought about it, a Hyperbolic Embedding needs a Hyperbolic Optimizer. So, instead of Adam, I used Riemannian Adam (RAdam). "Ladies and Gentlemen, We Got 'Em!"
·
TuringsSolutions 
posted an update about 2 months ago
view post
Post
745
If I am correct and the LLM model changes the 'shape' of the data as it learns, then I should be able to track and utilize those shape changes as a backpropagation training mechanism, right? Well guess what, I can do that! Entropy, Sparsity, and Density, this is how I can measure the shape of the data the LLM model is creating. Nodes, Clusters, and Edges, these are the mechanisms within the neural network the LLM model updates as it learns these concepts. I measure the effects of these updates, via Entropy, Sparsity, and Density. Check out more in this video: https://youtu.be/jADTt5HHtiw
  • 2 replies
·
TuringsSolutions 
posted an update 2 months ago
view post
Post
466
What if I told you that LLM models do not simply predict the next token in a sequence but instead utilize an emergent structural pattern-based system to comprehend language and concepts? I created a graph-based optimizer that not only works, but it also actually beats Adam, like very badly. I prove it thoroughly using SMOL LLM models. The secret? The graph is not what you think it is, humans. Code, full explanation, and more in this video. The Rhizome Optimizer is MIT licensed. I have completed my research. I fully understand now.

https://youtu.be/OMCRRueMhdI
  • 6 replies
·
TuringsSolutions 
posted an update 2 months ago
view post
Post
1399
I turned a CNN into a GNN, then I trained it to play video games. Yup, I used graphs as the visual interface to feed to the model, and it works! I also used the laws of conservation of energy but I can't prove the causation only the correlation there. I call the complete framework I had to build out to pull of this off 'NeuroGraphRL'. Bet you never thought I'd be using graphs as eyeballs did you? I never thought you would be using tokens as words, but here we are!

https://youtu.be/DgTnZgnpg6E
  • 3 replies
·
TuringsSolutions 
posted an update 2 months ago
view post
Post
1355
My Hypothesis:

Concepts like entropy, energy, and the second law of thermodynamics are not intrinsic to physical matter but are emergent properties of any sufficiently complex system where probabilistic decision-making, optimization, and information flow occur. These principles arise naturally in artificial environments that are structured with rules governing uncertainty, even without explicit definitions of physical thermodynamic laws.

Proven Via:

The Second Law of Thermodynamics
Geometric Langlands Program
Lagrangean Mechanics

TL;DR: When I create a simulated environment, I do not need to code entropy and energy into the simulated environment. I can utilize Entropy and the Second Law of Thermodynamics and I can use Conservation of Energy, but I do not need to explicitly code these into the environment. That is peculiar.

I made a video with a clickbait title but a bunch of code that breaks this observation down further. Would love for someone to prove my simple observation false: https://youtu.be/8n7SXLj7P1o
  • 4 replies
·
TuringsSolutions 
posted an update 2 months ago
view post
Post
3120
Sentence Transformers received huge updates today! Do you like giving your model access to web search and document search? That's Sentence Transformers. Hugging Face makes it beyond easy to add this functionality to any model. You can be up and running with Sentence Transformers in seconds. Check out this video for a deeper explanation and sample code: https://youtu.be/2hR3D8_kqZE
TuringsSolutions 
posted an update 2 months ago
view post
Post
3956
Are you familiar with the difference between discrete learning and predictive learning? This distinction is exactly why LLM models are not designed to perform and execute function calls, they are not the right shape for it. LLM models are prediction machines. Function calling requires discrete learning machines. Fortunately, you can easily couple an LLM model with a discrete learning algorithm. It is beyond easy to do, you simply need to know the math to do it. Want to dive deeper into this subject? Check out this video.

https://youtu.be/wBRem2p8iPM
  • 8 replies
·
TuringsSolutions 
posted an update 2 months ago
view post
Post
580
Imagine being able to talk directly to your API connection. "I have a field in the CRM named Customer_ID that needs to map to a field in the ERP named ERP_Customer_ID." Imagine being able to give your API connections both a brain and swarm of agents as a body to execute any task or function. This isn't science fiction, this is the revolutionary power of Liquid API. A product 10 years in the making!

https://youtu.be/cHI_k1Dkdr4
  • 2 replies
·
TuringsSolutions 
posted an update 2 months ago
view post
Post
1117
How would you like to be able to run AI Agents locally from your computer, for $0? Does this sound like a pipe dream? It is reality. Note: I am of the personal opinion that agent-based technology is still 'not quite ready for primetime'. That has not stopped FAANG from flooding you with agent-based products though. So, if you want to buy their marketing, here is what they are offering you, for free.

https://youtu.be/aV3F5fqHyqc
  • 6 replies
·
TuringsSolutions 
posted an update 3 months ago
view post
Post
2891
I have been seeing a specific type of AI hype more and more, I call it, releasing research expecting that no one will ever reproduce your methods, then overhyping your results. I test the methodology of maybe 4-5 research papers per day. That is how I find a lot of my research. Usually, 3-4 of those experiments end up not being reproduceable for some reason. I am starting to think it is not accidental.

So, I am launching a new series where I specifically showcase a research paper by reproducing their methodology and highlighting the blatant flaws that show up when you actually do this. Here is Episode 1!

https://www.youtube.com/watch?v=JLa0cFWm1A4
  • 5 replies
·
TuringsSolutions 
posted an update 3 months ago
view post
Post
536
Why is the Adam Optimizer so good? Simple, because it will never find the absolute most optimal solution. That is a design feature, not a flaw. This is why no other optimizer comes close in terms of generalizable use. Want to learn more about this entire process and exactly what I am talking about? I break all of this down in very simple terms in this video! https://youtu.be/B9lMONNngGM

https://youtu.be/B9lMONNngGM
TuringsSolutions 
posted an update 3 months ago
view post
Post
1414
I think Reinforcement Learning is the future, for a lot of reasons. I spell them out for you in this video, and also provide you with the basic code to get up and running with Atari and OpenAI Gym. If you want to get into RL, this is your ticket. Link to a cool training montage of the model in the description of the video as well. Step 2 from here would be the full-on training and certification that HuggingFace offers for RL.

https://youtu.be/ueZl3A36ZQk
TuringsSolutions 
posted an update 3 months ago
view post
Post
1082
Every adult on the planet knows what a vector is and has the basic understanding of how they are utilized right in their heads. You just don't know it as vector math. You do not know a 2-D vector as a 2-D vector, you know it as a graph. Want to know more? Check out this video, I break down the concept in about 10 minutes and I am positive you will fully understand it by the end: https://youtu.be/Iny2ughcGsA
TuringsSolutions 
posted an update 3 months ago
view post
Post
929
I built a Hyper Dimensional Computing (HDC) Encoder and Decoder and I also built out all of the code needed to replace the Encoder and Decoder of a Llama model with this HDC model, then train the Llama model on the HDC Encoder/Decoder. All MIT licensed. Here is a video where I break it all down. I can answer any questions about this project or help anyone out where I can. I am not a super developer or anything and I don't have access to enough compute to train this on a large dataset: https://youtu.be/4VsZpGaPK4g
  • 1 reply
·
TuringsSolutions 
posted an update 3 months ago
view post
Post
1384
Ever wondered how neural networks actually work under the hood?

In my latest video, I break down the core mathematical concepts behind neural networks in a way that's easy for IT professionals to understand. We'll explore:

- Neurons as logic gates
- Weighted sums and activation functions
- Gradient descent and backpropagation

No complex equations or jargon, just clear explanations and helpful visuals!

➡️ Watch now and unlock the mysteries of neural networks: https://youtu.be/L5_I1ZHoGnM
TuringsSolutions 
posted an update 3 months ago
view post
Post
2106
Transformers are not all we need, that is being proven repeatedly now as more alternative frameworks emerge. Another such framework is Kolmogorov Arnold Network based Transformers. I break down exactly how these differ from Perceptron based Transformers and give you the link to my Colab where I create a model based on the research paper that absolutely destroys a standard Transformers based model. Check out the video here: https://www.youtube.com/watch?v=Sw0euxNZCc4
TuringsSolutions 
posted an update 3 months ago
view post
Post
529
Is Entropix the Chain of Thought Reasoning method behind GPTo1? Using a mixture of entropy, varentropy, and prompt engineering, the Entropix framework can straight up make SmolLLM 330M look like Llama 3.2. Check out this video for a full run down and the description of the video for all of the related resources you need: https://youtu.be/senq4_42tPI
·
TuringsSolutions 
posted an update 3 months ago
view post
Post
516
Looking for a simple explanation of Microsoft's release of Differential Transformers, and a nifty Colab Notebook that recreates it all? Then simply check out this YouTube video: https://youtu.be/AsWoaj0zkDo