APX Charts, page-2762

That is just silly. Google moved the data cleaning operation inhouse precisely because of the amount of data they needed for the AI training. At the time all the big LLM's were training on pretty much raw web data. Now they have hit two problems: 1. They are all running out of data to train on, and 2. They need much cleaner data than they realised initially because of growing copy right issues, LLM hallucination, and reaching the limits of the reasoning ability of the current generation of LLM architecture trained on unfiltered text. Companies like Appen are enjoying a resurgence in the need for their services therefore which has literally just emerged in the last 6 months. The current crop of LLMs have also brought about hugely expended opportunities for customisation/specialisation with managed RAG frontends, etc. All these require customised & vectorised data bridges.