🚨 AI Industry's Open Secret: Data Heist! 🚨 Turns out, devs of mini AI models are sneakily using data from big shots like OpenAI, Google & Anthropic to craft budget-friendly alternatives. 🛠💸 Here’s the scoop: 1. Devs pay for access to GPT-4. 2. Bombard it with Qs like "What’s wrong with this code line?" 3. Use the answers their Qs to train their own rival models that can debug code. This tactic is catching fire lately! 🔥 Unsloth AI's co-founder revealed about half their clients churn some data from GPT-4 or Anthropic's Claude and throw it into their own model mix. Many are also scooping data from ShareGPT, a hotspot where devs drop answers generated by OpenAI models. These smaller models often build on open-source favs available on platforms like Meta or Mistral AI but get a major boost by incorporating responses from OpenAI models. Some devs even use a service called OpenPipe to automate this sneaky process. As more companies cook up partially borrowed models, telling them apart gets tough, shaking up the competitive edge of leaders like OpenAI who might need to rethink pricing strategies! And there’s an alternative brewing: synthetic data generated by companies' own AI models. 🌪🤖 #AITeaSpilling #CodeHeist #DataDrama

Code with more twists than a Money Heist episode 🤣 Welcome to 'Code Heist.' 😉💻 #BFCETHack #moneyheist #codingevent #techinnovation #codeheist

I’m setting up a CodeHeist to rob the #codevault and here is my team Me - Professor @reedahkh Nairobi @Bamoohaa Tokyo @tlakomy Berlin @Dutch_Guy Denver @odenfedy Rio @hackSultan Helsinki @dabit3 oslo @saniyusuf Monica @SantoshYadavDev Raquel Set up urs with 1st @ as member