2 ms·
HN submissions have a bunch of examples in them, but worth remembering they were released as "Look at this somewhat cool and potentially useful stuff" rather th
by embedding-shape 7mo ago
HN submissions have a bunch of examples in them, but worth remembering they were released as "Look at this somewhat cool and potentially useful stuff" rather than what we see today, LLMs marketed as tools.
https://news.ycombinator.com/item?id=21454273 https://news.ycombinator.com/item?id=21454273 / https://news.ycombinator.com/item?id=19830042 https://news.ycombinator.com/item?id=19830042 - OpenAI Releases Largest GPT-2 Text Generation Model
HN search for GPT between 2018-2020, lots of results, lots of discussions: https://hn.algolia.com/?dateEnd=1577836800&dateRange=custom&dateStart=1514764800&page=0&prefix=false&query=GPT&sort=byPopularity&type=story https://hn.algolia.com/?dateEnd=1577836800&dateRange=custom&...
- Den_VR 7mo agoI still think of The Unreasonable Effectiveness of Recurrent Neural Networks and related writings. http://karpathy.github.io/2015/05/21/rnn-effectiveness/ http://karpathy.github.io/2015/05/21/rnn-effectiveness/
- embedding-shape 7mo agoFun to revisit no doubt, the comments make it even better. > SuckCocker 7 years ago - "in short: SKYNET is not far away. Be proud to be a part of it!"
- cindyllm 7mo ago[dead]
- dalemhurley 7mo agoWild how many people were predicting the AI slop, but was dismissing it as unlikely beyond some trolls.