61–70 of 149 posts
This does GPT-2 X 10. For anyone wondering what GPT-2 is doing look at this baffling subreddit and marvel at how one GPT-2 model trained for $70k spits out better comedy than everybody on the payroll of Netflix combined. https://www.reddit.com/r/SubSimulatorGPT2/
This does GPT-2 X 10. For anyone wondering what GPT-2 is doing look at this baffling subreddit and marvel at how one GPT-2 model trained for $70k spits out better comedy than everybody on the payroll of Netflix combined. https://www.reddit.com/r/SubSimulatorGPT2/
"I'm really starting to get worried about my Higgs Boson (HBN) after watching some videos on YouTube" [0]
"This repost and all of your posts are garbage." [1]
"It's the most random and unoriginal shit I've ever read." [2]
[0] https://www.reddit.com/r/SubSimulatorGPT2/comments/f1sqyh/an...
[1] https://www.reddit.com/r/SubSimulatorGPT2/comments/f1vfnv/wh...
[2] https://www.reddit.com/r/SubSimulatorGPT2/comments/f1o83a/fo...
Earlier quoted context omitted.
I would agree with you to a certain extent, but I still think there is a big missing component to make this a reality. Larger and more accurate general language models are great, but to enable use cases other than categorization, translation, summarization, etc., there will almost certainly have to a be contextualized knowledge graph layer. This is basically what I assume you mean when you say "the ability to create,…
Replying from my laptop. I agree, there needs to be a way to represent relationships between information. I personally don't think knowledge graphs will be the ones to do it, not because they dont work, but because of how imperfect they are, in the data quality sense. See this paper here: "DIFFERENTIABLE REASONING OVER A VIRTUAL KNOWLEDGE BASE" https://openreview.net/pdf?id=SJxstlHFPH Which is a recent effort, among…
This does GPT-2 X 10. For anyone wondering what GPT-2 is doing look at this baffling subreddit and marvel at how one GPT-2 model trained for $70k spits out better comedy than everybody on the payroll of Netflix combined. https://www.reddit.com/r/SubSimulatorGPT2/
This does GPT-2 X 10. For anyone wondering what GPT-2 is doing look at this baffling subreddit and marvel at how one GPT-2 model trained for $70k spits out better comedy than everybody on the payroll of Netflix combined. https://www.reddit.com/r/SubSimulatorGPT2/
GPT2 is still very random and quite stupid.
You start it with your love for your girlfriend as a context, she becomes a cam girl into hard core anal two paragraphs later. You start with religion, "Muslims must be exterminated". You start with software and you get a description of non existent hardware with instructions about how to setup a VPN in the middle. You start with news, and you can read than China supports the Islamic state.
That's cool because it has more context than Markov chains which usually have only 3 words of context, but it's still a long way to go before I trust anything generated by this kind of algorithm.
Those summaries look impressive, although a bit repepetive
What they don't tell you is that these summaries are always hand picked from a few that were generated.
Earlier quoted context omitted.
I'm a researcher in the field. Not in NLP anymore, but I worked on that as well, years ago, and I keep up with the research. You can't "understand the context" of "War and Peace" unless you have real, actual AGI. I doubt actually it can be fully understood at all when translated to English and read by someone without the right cultural background. This is an extreme example, chosen to make it easy to see that it appl…
Maybe. I take a really optimistic view of attention based mechanisms, like I explained and added to my original post. If you read the recent reformer paper, they produce a new way of computation to start building a model that can in some way, encode the relationships between different parts of war and piece. The bottle neck right now is computation, we don’t know how well these models can learn when that bottle neck…
This does GPT-2 X 10. For anyone wondering what GPT-2 is doing look at this baffling subreddit and marvel at how one GPT-2 model trained for $70k spits out better comedy than everybody on the payroll of Netflix combined. https://www.reddit.com/r/SubSimulatorGPT2/
It's definitely better than the original using Markov chains. It fits very well this use case, and in my opinion only this use case. GPT2 is still very random and quite stupid. You start it with your love for your girlfriend as a context, she becomes a cam girl into hard core anal two paragraphs later. You start with religion, "Muslims must be exterminated". You start with software and you get a description of non ex…
https://www.reddit.com/r/SubSimulatorGPT2/comments/f1pypf/so...