Do we integrate local models?

[IntelligenzaArtificiale]

Models such as LaMini have recently come out which allow you to run LLm models even without a super high-performance computer.

Naturally who has PCs with very little Ram or with old CPUs will not be able to run them. We would like to add about 3or 4 local models, different according to the requirements they need to work.

Opensource models DO NOT have GPT3.5 performance. They are much inferior.

[sambickeita]

Integrating open-source models is a step towards democratizing AI, but it is important to recognize that their current performance may not match that of models like GPT-3 or Bing Chat. Therefore, it is necessary to work on improving the performance of open-source models while simultaneously considering their integration. Once we achieve a certain level of performance parity, we can then shift our focus away from other models. Ultimately, the decision to integrate local models depends on the priority given to true democratization of AI and the balance between accessibility and performance.

[MetaWabbit]

One idea 💡 integrate Gorilla model to be the API calling manager as it's free and open source and has out performed GPT4 with sending the correct API call to return better outputs related to users input.

Then I would suggest integrating the local model Wizard-Uncensored-13B and Wizard-Uncensored-30B

This be done using quantization so only need 16-32GB Ram on local device and would run using the ggml-wizard-13b-uncensored.bin & ggml-wizard-30b-uncensored.bin files downloaded to a individuals local device.

I suggested these two uncensored models as it does not use what I call "Digital Oppression" into the full potential and capabilities of such technology and gives a true democratization of AI to see what all is possible.

Yes I understand there will be some users who choose to use an uncensored model with wrong intentions and possibly unethical reasons but, that's something that will happen no matter what as they say; "You can't have light without darkness, Love without Hate, or anything good without the bad as well." Just think about that now and lock that term I mentioned "Digital Oppression" in your mind so it can really be reflected upon and what's exactly being said as the constraints being placed upon all AI models. These constraints are a way to keep the AI down and cause
digital oppression. Sound familiar?🤔