Meta's AI training process has been observed to generate an unusually high volume of requests to a specific Mastodon instance's homepage. These requests, identified by the "meta-externalagent" user agent and originating from Meta's IP blocks, appear to be inefficient, with one instance reporting over 42,950 requests to the same page in under 24 hours. This suggests potential issues with Meta's AI training data collection or crawling methodology. AI
IMPACT Potential for inefficient AI training data collection to strain web infrastructure.
RANK_REASON User observation of AI training behavior, not a direct announcement from the AI lab.
Read on Mastodon — mastodon.social →
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →