🤖 AI Summary
Reddit cofounder Alexis Ohanian revealed on a podcast that about a decade ago he clashed internally with cofounder Steve Huffman over Sam Altman’s repeated requests to scrape Reddit for OpenAI. Ohanian said he “felt in my bones” Reddit shouldn’t hand over its data after Altman — who had helped fund Reddit in 2014 and cofounded OpenAI in 2015 — began pressing for large-scale access in 2015–16. Ohanian lost that debate; in 2024 Reddit and OpenAI formalized a licensing deal that grants OpenAI rights to train models on Reddit content. Ohanian praises Altman’s early recognition of Reddit’s unique value: dense, human-generated conversational data.
The exchange highlights why platform data is strategic for modern AI and surfaces technical and ethical implications: Reddit’s content offers diverse, candid dialogue that boosts language-model fluency and conversational grounding, but it also raises concerns about provenance, amplification of biases, dataset contamination with AI-generated or inauthentic posts, and user privacy. Ohanian warns of a “dead internet” trend where synthetic content dilutes human signal, pushing platforms to develop verification and provenance tools (while balancing privacy) and prompting tighter moderation, dataset curation, and transparency practices. The episode underscores both the commercial pressure to capture unique training data and the need for safeguards around authenticity, consent and long-term stewardship of public-platform corpora.
Loading comments...
login to comment
loading comments...
no comments yet