Because most of what you said applies to AID as well. I think what Gwern is proposing is perhaps better seen by you as an optimization for the current AID rather than a "true" CYOA novel implementation. I think it's possible that if some future generation of GPT could generate a good-enough novel, then it could also generate a good-enough interactive novel, but this thing has gotta learn how to crawl before it can learn how to walk. If AI Dungeon were to auto-generate an interactive novel, particularly in the way Gwern describes here, with the choices already pre-computed and crowdsourced in advance, the result would be indistinguishable from a (bad) hand-written interactive novel. In other words, when GPT-3 computes an entire novel, it's indistinguishable to the reader from a (bad) hand-written novel. Presenting a pre-computed menu of options is what hand-written interactive novels already do to succeed, it would have to compete with hand-written interactive novels on quality. The feature that makes AI Dungeon interesting and unique is its ability to improvise in response to player's actions. Writing an interactive novel with interesting, dramatic choices is harder than writing a non-interactive novel, and GPT-3 can't even do that. (And we have professional editors who read and edit their work.)Ĭould we use GPT-3 to avoid paying professional authors? No, of course not. Our approach is to pay professional authors to write these interactive novels. (The author of TFA links to one of our blog posts in the footer of this article. Our novels use hidden state to provide an experience that's longer, richer, and deeper than traditional paper-based CYOA books. I run a company, Choice of Games, that publishes hand-written interactive novels. People have, but the novels it generates aren't worth reading. Today, GPT-3 can generate text, so you could try to use it to generate a whole novel. GPT-3 won't be able to generate a good CYOA until it can generate a good non-interactive novel. I don't think it will work, at least not with the current generation GPT-3. 06:48:37 (162 KB/s) - ‘model_v5.torrent.zip’ saved ġ2/08 06:48:37 Downloading 1 item(s)ġ2/08 06:48:37 Exception caught while loading DHT routing table from /root/.cache/aria2/dht.datĮxception: errorCode=1 Failed to load DHT routing table from /root/.cache/aria2/dht.In this article, Gwern proposes replacing AI Dungeon's free-text parser with a menu of CYOA options. Resolving github-production-repository-file-5c1aeb.s3. (github-production-repository-file-5c1aeb.s3.). Remote: Total 48 (delta 2), reused 20 (delta 0), pack-reused 0 status=403Īdditionally, the workaround now fails with: > errorCode=22 The response status is not successful. Remote: Total 48 (delta 2), reused 22 (delta 0), pack-reused 0ĭownloading AIDungeon2 Model. Remote: Compressing objects: 100% (39/39), done. I "**" the ip adresses since i don't know if it save to share them. There is 403 error while downloading and this causes errors while initializing the AI Dungeon. See '-l' option in help/man page for details. If there are any errors, then see the log file. =+=+=+=Īria2 will resume download if the transfer is restarted. URI= /model_v5/model-550.data-00000-of-00001Įxception: errorCode=22 URI= -> errorCode=22 The response status is not successful.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |