Category: Uncategorized

  • LLMs – How To Choose

    AI and Machine Learning. An LLM is a “Large Language Model” that is trained on a task with data and its internal structure. The language model of today is focused on learning how to generate text in the different languages of the world. A broad number of AI companies are developing these “generative chatbots” that allow a user to engage with a human-ish software service to solve complex problems. These are chatbots like ChatGPT, Gemini, DeepSeek, and Co-Pilot.

    AI and Machine Learning algorithms today are good at retrieving information, keeping a backlog of previous conversational information, presenting complex topics across the World Wide Internet, and at learning how to generate text that is more and more user-specific. All of this comes with a heavy dose of training and origination bias. The creators of the algorithms and the models today create inherent bias in their AI products.

    The GPT in ChatGPT, LocalGPT, or PrivateGPT comes from the LLMs that are “generative pretrained transformers”. Which to me sounds like these are a database full of training vertices and the products of common mathematical functions that can aid in the natural language processing task. These transformer-based models emerged in 2017.

    The LLMs that we independent software developers used are the BLOOM and LLaMA. Other Free LLMs are Llama 2, Gemma, Falcon, Mistral 8x7B, and Phi-3.

    Paid and subscription LLMs are for company software developers and those seeking to create code from their AI prompts. Some of the paid LLMs include GPT-4, Gemini Ultra, Claude 3, and Grok.

    When considering the LLM you need, look at these abilities:

    • Reasoning
    • Coding
    • Creating Video/Audio/Images
    • Mathematics ability
    • Efficiency
    • Analysis
    • Conversational Ability

    When considering the drawbacks of an LLM, look at these limitations:

    • Resource-constraints of your hardware/cloud subscription
    • Bias in the pre-trained data
    • Number of parameters used in pre-trained data
    • Constraints and deployability of the model

    Sources: Choosing the Right LLM for You: A Guide for Powerful Gen AI

  • Welcome Again!

    This is a WordPress blog. It used to be based around one project, an internet radio station, but now it is about all my projects, and sometimes will be about internet radio stations!

    I love music! If you are a musician please stop worrying about anything else and go make some more music!

  • the Server is Down – New Modem Activated By Comcast

    SO the modem was definitely fuckin’ up things at my parents house this past week. Ever since this big thunderstorm that knocked the power out last week. And even before that my router I had plugged in for them was definitely acting up and almost locking up sometimes or dropping its wifi network at random. That problem was intermittent and seemed to be solved by propping it up off the carpet onto a plastic lid. I think it was overheating or something. But nevermind that, the fuckin’ modem from Comcast wasn’t even supplying internet to the PCs, no matter what ethernet port we tried in the back of the modem. And also the xfinitywifi that comes by default on in the modem was fuckin’ up. It just kept hanging forever or maybe delivering a 1 webpage out of every 50 tries. It was weird problem. Either way, we fixed it by running to Best Buy and grabbing an Arris DOCSIS 3.0 SB6141 for $60. After about half an hour of resetting the thing and registering it with the guy on Comcast and him rebooting it a bunch of times we finally got it to work. Feels nice to have solid internet again. I can’t wait for it to go out again. I can only imagine what it’ll be next…