11–20 of 166 posts
I wonder how they tested that
> (CTEC) found that extremist groups can use GPT-2 for misuse, specifically by fine-tuning GPT-2 models on four ideological positions: white supremacy, Marxism, jihadist Islamism, and anarchism. CTEC demonstrated that it’s possible to create models that can generate synthetic propaganda for these ideologies I wonder how they tested that
I would bet it gets the talking points but can’t convey the subtlety.
> (CTEC) found that extremist groups can use GPT-2 for misuse, specifically by fine-tuning GPT-2 models on four ideological positions: white supremacy, Marxism, jihadist Islamism, and anarchism. CTEC demonstrated that it’s possible to create models that can generate synthetic propaganda for these ideologies I wonder how they tested that
wow I’d love to read the Marxist and anarchist texts it has produced. I wonder if they used good source material. I would bet it gets the talking points but can’t convey the subtlety.
Sorry for asking but is there an example output and an example input?
Surely we are not far off models capable of submission-quality essays that will enable a new generation of cheating.
Sorry for asking but is there an example output and an example input?
It's a subreddit which consists entirely of posts and comments by GPT-2 bots (with votes by humans). There's a variety of different bots fine-tuned on posts and comments from different subreddits, so depending on which bot is posting you can get wildly different results.
Paper: https://d4mucfpksywv.cloudfront.net/papers/GPT_2_Report.pdf Code: https://github.com/openai/gpt-2
notebook: https://twitter.com/theshawwn/status/1191800180192010246
code: https://github.com/shawwn/gpt-2
It's a fork of nshepperd's gpt-2 codebase (https://github.com/nshepperd/gpt-2) which lets you fine-tune 117M and 345M on GPUs.
For a tutorial on how to fine-tune GPT-2, see http://gwern.net/GPT-2
> (CTEC) found that extremist groups can use GPT-2 for misuse, specifically by fine-tuning GPT-2 models on four ideological positions: white supremacy, Marxism, jihadist Islamism, and anarchism. CTEC demonstrated that it’s possible to create models that can generate synthetic propaganda for these ideologies I wonder how they tested that
Surely we are not far off models capable of submission-quality essays that will enable a new generation of cheating.
From my observation, even the largest GPT-2 model has difficulty retaining any long-range relationship information. In the "unicorn" writing example that was published originally, the model 'forgets' where the researchers are (climbing a mountain versus being beside a lake iirc) after just a few sentences. Because of this, it's hard to imagine models of this type being able to write long-form coherent papers. Now if…
But I can’t remember the last time I relied on sequentially remembering the ordering of tokens in order to complete an essay or hell even reply to an email.
Structurally we retain some kind of hierarchical information (topic, places, names, events) about text.
Is there any active research looking into text generation models which do this? Maybe some kind of query that is made in a learned vector space and which is not temporarily dependent but rather “spatially” - as in these are the facts about the text being generated so far.