Can Frank follow and respond to a non-main blog? Like, if a have another blog I wnat it to follow but I can't send !follow from it bc it's not my main one?
I haven’t made an automatic way to do this, but if you tell me the name of the sideblog I can just go in and manually make Frank follow it.
OK, I put a slightly cleaned-up version of nostalgebraist-autoresponder on github… have fun trying to understand the convoluted horrors I’ve created :P
There are more detailed disclaimers in the README, but just to be clear, this doesn’t have the model/data files you’d need to reproduce my bot or create your own, and I can’t even guarantee it works the way the “live” version does with those files in place.
However, there are some terse hints and scripts and stuff which in principle would let someone reproduce the models or create analogous ones for a different bot, if that person feels comfortable getting into the awkward weeds of GPT-2 fine-tuning and its lack of mature tooling.
@the-real-numbers since you asked to be notified
Let me know if you ever do post Frank's code to GitHub, I would love to poke around. Would it be weird if I made an attempt to do something similar? I would credit you & Frank for the inspiration of course.
Let me know if you ever do post Frank’s code to GitHub, I would love to poke around.
Will do.
Would it be weird if I made an attempt to do something similar?
Not at all, go for it!
On the topic of publishing Frank’s code… I’ve wanted to do this in some form for a while, at the very least just for the sake of “showing my work.” However, there are 2 very different directions I could go with it:
The first option treats the code more as documentation, the latter as a toolkit intended for real use. When I think about releasing the code, I usually jump to thinking about the latter, and then groan when I think about all the details involved. The former is much easier, and perhaps could eventually transform into the latter, but I don’t know if it resembles what people are thinking when they ask me for the code. Feedback on this (from anyone interested) is welcome.
I may write more about this later, but quickly noting to forestall confusion – this week I went back and fine-tuned Frank’s underlying GPT-2 model again.
The generator hadn’t gotten nearly as much iterative improvement work as the other pieces, and I figured it was due for a refresh. I swapped in the new one around noon today. If it turns out to be worse than the old one in some significant way, I can always go back to the old one.
What’s changed:
Hey, so I prompted your autoresponder with the name of one of my fictional creations to see what it'd make of it, and I actually really like the paragraphs it came up with? What is the copyright status on this material? Would you agree to me reintroducing concepts made up by the autoresponder back into the actual story?
That’s totally fine! In fact, it’s more than fine – I am delighted that my creation is inspiring other people creatively :)
I have not thought at all about the copyright status of the autoresponder posts. I don’t feel any particular sense of ownership over the content of the posts, since I didn’t write them, but I guess I would object if someone were to pass them off verbatim as their own writing.
Even that is negotiable, though, and I can easily imagine giving someone permission to use long verbatim stretches of bot text in their own writing if they had a reason and asked me first.
I would also object (obviously?) if someone made a mirror of the bot and tried to pass it off as a bot they had made – the bot itself is my work – but that’s not really a feasible lie to maintain anyway.
How have things gone with the CHAZ in the last day or so? The national media paints if as a crazy place but earlier you said it was more a return to normal than anything else. Has that changed or are things still good?
Honestly, I haven’t been paying very close attention since I wrote that long post. Trying to catch up on lost sleep, get my brain hooked back into the stuff I was doing at work before the really hectic week, etc.
During the hectic week, it would have been impossible to ignore stuff even if I had wanted to, because it was literally audible outside, so there’s that.
——-
For reference, here are some of the news sources I look to on this topic. If you want to follow the situation from afar, or want to dig for context about some particular story you’ve heard, they’re a good place to start:
- The Capitol Hill Seattle blog (probably the best one place to start). Its primary blogger is also on twitter as jseattle
- Erica C. Barnett on twitter and her blog.
N. B. she often live-tweets press conferences, city council meetings and the like, so her twitter is a journalistic venue in itself, not just a place for opinions/jokes like many journalists’ twitters
- I don’t read SCC Insight as much as the previous two so I can’t vouch as much for the angle it takes, but it’s a similar venue for very detailed municipal politics news
- The Stranger and its blog (here’s their CHOP/CHAZ tag).
The Stranger is a local “alternative weekly” – magazine format printed on newsprint, this really specific type of irreverent house style, 90% of the ads are about weed and the other 10% are about porn, if you’ve seen one city’s version of this thing you’ve seen ‘em all – and it has all the quirks and annoying traits that come with its genre, but its reporting on the recent protests has been really good IMO.
- Local news outlet websites: KOMO (local ABC affiliate), KUOW (local NPR affiliate), King5 (local NBC affiliate), KIRO (local FM station), Crosscut (local online newspaper).
These are traditional news outlets, so they provide less detail than subject-specific blogs, and are less upfront about their biases. But if you’re reading that kind of news, a local outlet is less likely to tell you things that are straight-up wrong in a contextually obvious way.
I haven’t been able to sleep much this evening because of persistent, worrisome thoughts that I’m never going to get enough sleep because that screw’s all there is, I’m never going to get enough sleep, there’s just this horizontal line drawn across the heavens where the sun sets and rises and the sun never sets and rises for six days and never sets, the full moon never rises and never sets, I’m never going to get enough sleep, someone’s watching me from the heavens and she’s got big eyes and a cleft chin and she smells of lilacs and she has a warm, patient, loving heart and she’s always been there for me and she’s always going to be, she’s warm and kind and strong and responsible and she works every day in a shed near the station, on a farm, and she does all she can to provide a wholesome and comfortable environment for her two tiny children, and when I think of home I think of this warm, kind, intelligent, goodhearted woman who is always there for me, who takes one look at me when I get out of bed and just knows that it’s time for my nightly dose of Theracrete, I know it’s time for the magic bullet, and who has had enough, who is full of iron and rage and certainty and life for it to matter what others think of her, for her to ever fear the slings and arrows of the unkind and the indifferent and the capricious, for her to be bold as the sun in spring and the wind in summer and the ocean in fall, for her to be unafraid, because it’s all in the flow, she’s in the center of the divine circle, here to stay, she belongs in the center of the divine circle, here to stay, she doesn’t give a damn what anyone thinks of her, because as we learned a long time ago, the flow is the divine circle, there is no place to run, no escape.
Last decade introduced a pretty massive problem that our society has never had to grapple with at current scale before, and I don’t think anyone at all really has a solution to it: what happens when a large portion of a population voluntarily abandons news sources which make a good-faith effort at factuality in favor of news media designed to stroke their brains the right way?
This has always been a problem to some degree (tabloids are a good pre-internet example). But:
- The number of outlets reporting (and claiming to report) the news has exploded in recent decades, meaning that people are no longer seeing the same reports, creating a major diversification in information access that has educational benefits but also weakens a shared understanding of reality.
- I know everyone has a biased tendency to view their ideological opponents especially negatively, but the growth of Fox News really did play a big role here. Fox News was the first ever 24-hour news source in the US intentionally dedicated to serving the interests of a political party and its associated ideological movements, and over time it has both grown in influence and continuously moved to the right without stop.
- The internet took all of this and made a million times worse. Anyone at all can claim to be reporting news and get it massive attention in hours, people have unprecedented power to self-select what they see and who they see it from, and there’s so many sources that very few people are ever receiving the same collection of information.
The CHAZ in Seattle has been a good example of this. All the reports I’ve seen from sources I consider reliable, both in and out of the left, describe CHAZ as a somewhat disorganized but nonetheless peaceful collection of various protesters who have occupied a city park. But every time I’ve sought to see how it’s reported on the right, I’ve seen manipulated and unrelated photos being shared of it that look like a war zone, discussions of widespread extortion and vigilante violence under the authoritarian reign of anarchists and gangsters, and so on. There are tens of millions of Americans who believe that, right now, Seattle is under siege by organized terrorists bankrolled by George Soros that are conquering territory in hopes of destroying America. Thousands of right-wing media organizations are reporting on what can only be described as an alternative universe like it were our own.
I’m not one of those guys who frets over divisiveness, but this in particular is a really existential problem for any society that even makes the claim to democracy. What are the long-term effects of everyone living in an individualized epistemological bubble? How do you get out of this with anything that could be considered a single polity?
(via morlock-holmes)
important announcement
i am cute
@eightyonekilograms replied to your post“I am going to demonstrate a tumblr bug. This is the original post. I…”
Ah I think I ran into this inadvertently once. What’s the repro?
It works when
I don’t know exactly what posts count as “susceptible,” but my guess is that “susceptible” = “was created as an NPF post.”
NPF is tumblr’s “v2″ implementation of how they store posts on their backend, see here and here.
When it was released, you could only create NPF posts using mobile clients. I’m not sure this is still true, but the trick only worked when I created the OP on mobile and not on web, so if we assume “mobile = NPF” (plausible) and “NPF means susceptible” (plausible), it all adds up.
Why do I think this might be an NPF issue, besides the observed connection to creating on mobile?
In the “legacy” storage format before NPF, posts were stored as HTML, and the structure of reblogs was represented in nested HTML blockquotes. Additionally, tumblr has not moved everything to NPF – as I implied above, some posts are “created as” legacy and some are “created as” NPF. And they do some translation between the two formats in various places (exactly which places, I’m still hazy on) as posts are requested, modified, and reblogged across different clients.
The bottom line is, somehow this “trick” is hitting some part of their code that (rightly or wrongly) thinks it’s dealing with the old format where blockquotes encoded structure, and it’s misunderstanding a content blockquote as a structure blockquote.
—-
The only mitigation I know of is to go into the Markdown editor and save the post “as” Markdown.
(I have no idea what this does behind the scenes. If I try to retrieve the post in NPF through the API it looks the same as before (the NPF looks “correct” in both cases), if I try to retrieve it in legacy/HTML format it’s fixed, and its display in the browser is fixed.)
Unfortunately, you don’t seem to be able to do the Markdown thing through the API, so my bot can’t do this.