Github Copilot and ChatGPT alternatives

ai-codingcopilotllm-toolsself-hostingdata-privacy

Abstraction: Survey of AI coding assistants and privacy-safe ChatGPT alternatives in 2023

Key points:

Key coding tools: Tabnine (2019), GitHub Copilot (2021), Replit Ghostwriter, Amazon CodeWhisperer, Codeium, SourceGraph Cody, Tabby, FauxPilot — several support self-hosting
ChatGPT's web interface uses inputs for model training by default; Samsung employees leaked confidential data this way in 2023
Privacy-preserving LLM alternatives: OpenAI API (data not used for training), Azure OpenAI Service, Aleph Alpha (European, no customer data training), Cohere
Databricks' Dolly: open-source LLM trained on 15,000 records, cheap to build, viable self-host option
Core strategic question for companies: "buy, build, or self-host?" — depends on tolerance for vendor data exposure and cost
50+ LLMs with 1B+ parameters were available via open-source checkpoints or APIs as of May 2023

Connections: Github Copilot · Chatgpt · Codeium · Agentic Coding · Large Language Models