Philosophy

The Philosophy of LocalCat is to make the large deep learning models, especially large language models (LLMs), easy to use (load, fine-tune, and deploy) to every artificial intelligence (AI) practitioners.

We believe that the large deep learning models are the future of AI and can be used to solve many real-world problems. We are working on making the large deep learning models accessible to everyone by providing a simple API and easy to use interface.

LLM Strategy

We are in the era of AI transformation, and the LLMs one of the most promising areas of AI. You cannot talk about your AI strategy without LLM strategy.

Basic Components

When it comes to LLM strategy, you have to think about there basic components:

  1. Model: Train your own models or use pre-trained models.

  2. Compute: Train your models on your own infrastructure or use cloud-based infrastructure.

  3. Data: Collect your own data or use publicly available data.

Apart from large tech companies or AI startups, most companies do not have the AI scientists like the OpenAI to develop their own LLM model and the resources to train their own models; but they can use their own data, which is a valuable production assets. Therefore, the best strategy for most companies is to use pre-trained models and fine-tune them on their own data.

Note

This is akin to the approach of traditional machine learning, where you utilize well-crafted ML algorithms (e.g., Random Forest, Gradient Boosting, etc.) and train them on your data to construct your models. We offer two toolkits for traditional ML task: GossipCat and BatCat.

Task Focus

In addition to the three fundamental elements mentioned, you must also consider the model type: do you require a general chatbot like ChatGPT or a specialized model tailored for a particular purpose, like translation, text classification, etc.

A more general model means more energy cost, which eventually turns into money. Therefore, you have to think about the trade-off between the model type and the cost. Besides, we should be aware of more and more big techs are providing enterprise AI solutions and these solutions are getting better and cheaper.

As a result, we believe that the best strategy for most companies is to

  1. Define the core business problem that should be solved by AI in house.

  2. Collect the data that is relevant to the problem.

  3. Use pre-trained models and fine-tune them on your own data.

In other words, task focus is the data focus and fundamentally the business focus.