AI written blog goes viral, almost nobody notices

Liam Porr (via waxy.org):

Over the last two weeks, I’ve been promoting a blog written by GPT-3.

I would write the title and introduction, add a photo, and let GPT-3 do the rest. The blog has had over 26 thousand visitors, and we now have about 60 loyal subscribers… 

And only ONE PERSON has noticed it was written by GPT-3. 

People talk about how GPT-3 often writes incoherently and irrationally. But, that doesn’t keep people from reading it… and liking it. 

I wrote this last year predicting that GPT-2 would be used to write fake course papers. By all accounts, GPT-3 would do even better, and it’s sure to be public soon.

The Search for a Good Online Journal Platform

There’s an old saying, probably from a management textbook somewhere, that I see pop up on the software developer side of the internet from time to time. Everyone wants a solution that is fast, good, and cheap. You can only pick two.

Many of the issues that face the selection of a good online journal platform are, at the heart, the same as the challenges of software development broadly. So this ‘fast, good, cheap; pick two’ dilemma applies here as well.

I am one of the seven members of the editorial collective who will be taking up the editorship of Engaging Science, Technology, and Society, the open access journal for the Society for Social Studies of Science in 2021. My friend and colleague Aalok Khandekar is the editor-in-chief, and I’ve been working closely with him and Angela Okune at UC Irvine on deciding what online platform the journal should live on going forward.

The immediate problem is this: ESTS currently runs on a very widely used platform for online journals called OJS. OJS is a feature rich, strongly supported, well maintained platform for running a journal. It acts as the web gateway for visitors to find and download journal articles, it can be used to manage the entire journal production process all the way from manuscript submission to peer review to copyediting. It also archives files and links into various other online academic services like Crossref that make articles easier to discover.

OJS works well for many journals, and it has worked well for ESTS. OJS was actually the first platform that I used to edit a journal, when I was an editor of vis-a-vis: Exploration in anthropology, the graduate student journal at the University of Toronto.

ESTS currently lives on OJS 2, an earlier version of the platform, which is no longer updated. Our friends at the Public Knowledge Project, who host ESTS on their servers, would like us to upgrade to OJS 3.

OJS 3 looks nicer. It feels more modern than OJS 2. It would also be a relatively quick upgrade. ESTS is not a very old journal, so it hasn’t accumulated a lot of ‘cruft.’ We would need to redesign the visual appearance of our journal, since the files we have now that define the journal’s look are not directly compatible with OJS 3, but that is a relatively surmountable obstacle. ESTS is not doing anything visually complex.

The question that Angela, Aalok, and I have been asking each other for the past few months is, if we are going to do the work of upgrading the ESTS site to OJS 3, then why not take this as an opportunity to look at other platforms that we could use.

We’ve been talking with many people who are very knowledgeable about STS and academic publishing online broadly to figure out what we will need to do. No decision has been yet made on what is happening with ESTS, but we’ve spent a lot of time learning about new tools and functions on younger competing platforms, as well as the advantages of more established ones.

For example, one thing that we want to explore during our editorship is the use of open peer review. There are platforms out there now that have features supporting open peer review built in from the beginning, but OJS does not. On the flip side, a lot of the upstart platforms are still missing a few basic journal production functions. They also can impose a steep learning curve on editors, authors, and reviewers, because their workflow models can deviate quite significantly from the ones that many academics are used to. The newer platforms also tend to be hosted on cheaper servers, namely Amazon Web Services. If you have ethical or political qualms about Amazon and Jeff Bezos, perhaps OJS, which is hosted by an independent web services company based in Canada, is the better choice.

All of these are part of what Angela, Aalok, and I are considering as we think about what to do about our journal’s platform. While software decisions like these may seem like technical, and indeed, non-academic decisions, it makes sense to us, as the incoming editors of an STS journal, to think carefully about what kinds of scholarly products, engagements among community members, and linkages with the broader socio-technical systems of academia and the internet we want to be able to facilitate and to support. While the end product of our work will be what we hope is a usable and exciting system for our readers and contributors, the problems that we’re trying to work through are design problems not unlike those that go into a research or writing project. The material constraints and affordances are different, of course, but that is part of the fun. At least for me.

There will be more written about this in the coming months, but there are a few things I’ve learned so far.

In keeping with the ‘fast, good, cheap’ adage, there is little one cannot do if you have the money to do it. This is most evident from large publishers like Sage, who have the resources to build their own production lines, and can use it to churn out journal articles and can afford staff to handle different parts of the process. But ESTS is far more resource constrained. OJS is not cheap either, but it is cheaper than a purpose-built solution. It is also fast to implement, because it is well documented and has a great support ecosystem around it. So it checks off “Fast” and “Cheap.” Where does that leave “Good”?

This is where things get a bit murkier for us. The key functions of OJS were built when peoples’ idea of “online open-access journal” was significantly constrained by their idealized image of a bound and printed journal issue. Although it has some usability quirks and shortcomings, it does a “good” job at this stuff.

But since OJS became widely adopted, blogging and tweeting, among other things, have taken off as arenas for academic interaction and, effectively, short-form publishing.

At the same time, there has been a flourishing of software tools, including things like Evernote, Dropbox, Devonthink, Office 365, Google Docs, and so on, not to mention online platforms like PECE, that people are using to collaborate, write, gather and analyze data, and publish.

These two developments have contributed to slight decentering of the peer-reviewed journal article as the desired end point of academic work. “Good” is undergoing some redefinition. In fact, what we’re seeing is less a linear pipeline leading from research to publication, and more a set of feedback loops. One of these feedback loops leads from research to journal article. Through social media discussions and other forms of interaction, these loops can lead back to research starting another loop towards another article. There are also other feedback loops that lead into blog posts, articles in the popular press or threads of tweets (or, indeed, journal platform discussions), and these can each cross with each other, and each inform research once again.

Therefore, the journal article looks like one point along one path through a longer research program. Where OJS imagines “results” turned into “publications,” the reality is that the process involves a lot of back and forth between authors, reviewers, and journal editors and editorial staff. In addition, if we zoom out slightly, then we see that the program also involves myriad other bits of software, platforms, and people interacting. A journal article might start as a conference panel, which becomes a discussion on Twitter or on Slack, which becomes a shared folder on Dropbox and then a shared Google Doc, and so on.

OJS is tuned for the linear journal article pipeline, but it’s not as able to move beyond it into the other facets of academic work. Other platforms are coming in to fill this gap. These other platforms make it possible for broader engagement of the entire research program, and increase the visibility and recognition of the different scholarly contributions at each stage, at least to a small degree. The journal article becomes one stage among many.

There’s a major caveat here: I wrote ‘slight’ decentering above, because one place where things remain definitively centered on the journal article as traditionally imagined is in academic hiring and promotions processes. My impression is that many university hiring and promotions committees are not yet quite sure how to evaluate non-traditional outputs, but know very well how to value journal articles. There is an entire system of ratings, impact factors, regulations and guidelines that tell academics exactly what a “good” journal article is. For a blog post, there are no such corresponding rules.

This is where a journal like ESTS and its platform choice become very consequential in my mind. ESTS is a reputable, double-blind peer-reviewed journal associated with a major academic society. This means we have (at least some) authority to give authors the recognition they need to further their careers and receive the recognition of their peers. We are, however, also in a position as the “second” journal of 4S, to potentially be a bit more experimental, and use what resources and social capital we have to draw attention to non-conventional, but interesting, important, high quality forms of academic work.

This is where the notion of a ‘good’ publication starts to intersect with the question of a ‘good’ platform for that publication, and what good a journal like ESTS can do.

More on this soon.

Labor-based Grading

Sandie Friedman at Inside Higher-Ed on Asao Inoue’s “Labor-Based Grading System”:

Any rubric that evaluates students’ language according to a single standard — which is invariably a white, middle-class standard — is reinforcing racism, [Inoue] argues. Rather than evaluating students’ work according to a quality-based rubric, Inoue advocates grading students on the labor they complete. By “labor,” he means all the work that goes into writing: reading, drafting, giving and receiving feedback, revising, polishing.

Students receive a grade only at the end of the course, based on the labor they have completed.

To me, this feels like the kind of thing that will address structural inequalities in university teaching, which often falls into the “add minority content to curriculum” mindset.

The Design Squiggle as a Way of Thinking about Anthropological Research

The Process of Design Squiggle by Damien Newman, thedesignsquiggle.com

It seems to be well known among designers, but I only recently learned about the “Design Squiggle” a few days ago.

The Squiggle is a visual tool that helps you keep track of where you are in the design process. The big mess on the left is where most projects start: uncertainty with a bit of chaos. The smooth line on the right is where you eventually want to get to: clarity and focus.

Comparisons between cultural anthropology and design are not new. And yet it does not seem very widespread for anthropologists, and perhaps most other social scientists, to engage much with what designers do and think about to inform how they approach cultural research. Our models tend to be literary, philosophical, or scientific.

To me, this squiggle is a very good representation of what the research process in anthropology looks like. You are dropped in the middle of something interesting but difficult to process. Too much information. You try to pay attention to everything, make sense of the littlest occurrence, and tire yourself out in the process. As time goes on, you find a direction to go, but then find your eye pulled towards something new. Eventually, you learn how to separate figures from ground, and things start to make sense.

The Process of Design Squiggle by Damien Newman, thedesignsquiggle.com

This is a feeling that any teacher of anthropology has tried to convey to a student. But it’s not something that’s easy to grasp when you’re in the middle of first fieldwork, and this can be a source of more than a little anxiety. Maybe, I think, this Squiggle does better in conveying the feeling more than saying “This is what it’s like for everyone.”

The one thing I would add that would make the Squiggle feel truer to anthropological life is for squiggles to link to further squiggles in a nearly endless chain. The dissertation proposal is its own squiggle. Fieldwork is a bundle of squiggles. Writing is a squiggle for each paragraph, as is each chapter, each presentation, each article… Squiggles all the way down and at every step forward. I think the reason that all of the grad students and faculty members that I’ve come to know have tenacity as one of their defining traits is that tenacity is what it takes to work through endless fractal patterns of squiggles. Anthropologists should probably be called squiggleographers.

A fractal pattern of squiggles. One more thing: Since first encountering the work of Minakata Kumagusu during my fieldwork in Japan, I’ve been taken by a sketch of his that is known as the “Minakata Mandala.”

From http://acidrainproduction.com/slime-intelligence/

As I understand it, this is something that Evans-Pritchard would have called a diagram of witchcraft if he’d found it among the Azande. There are physical laws that can predict that an event will take place. But there is another order, beyond those laws, that explains why this event took place here, now, to these people. This is what Minakata’s sketch tries to render visual. He may have been a squiggleographer avant la lettre.

Next-level Deepfake Avatars for Zoom

My attempts with Facerig have been fun, but I’ve learned of a tool that lets you use the technology behind deepfake videos to turn faces of famous people into your avatar on Zoom or Skype.

Programmer Ali Aliev used the open-source code from the “First Order Motion Model for Image Animation,” published on the arxiv preprint server earlier this year, to build Avatarify. First Order Motion, developed by researchers at the University of Trento in Italy as well as Snap, Inc., drives a photo of a person using a video of another person—such as footage of an actor—without any prior training on the target image.


(Source)

The software can be found here on Github. I can’t use it because the software currently requires a CUDA-compatible NVIDIA GPU, which I don’t have, but maybe it’ll be made to work on low-power laptop integrated graphics chips as well.

University is anything but universal: How to apply to graduate school in three different countries

(For the past year or two, I’ve been contemplating writing a “missing manual” for university that explains the tacit skills and knowledge for getting by in university that first-generation students or foreign students may not have. This piece is looks at one potential topic.)

Though it will depend on the country, university, and discipline, only a tiny percentage of the people who finish an undergraduate degree will go on to study for a Master’s degree or a doctorate. A good number of these students will already have a clear sense of how to select and apply to a graduate program, as they may have had when they were applying to universities and colleges for their Bachelor’s degrees. But a significant number will have only a vague sense of about the application process, informed mostly by the materials on a university’s website and not by friends or relatives who have gone through the process before.

I’ve had the opportunity to be part of the graduate school admissions process a number of times. As a student, I applied to grad schools in the US and Canada. As a staff member, I have evaluated applicants in Japan and New Zealand. On the surface, they all seem quite similar in process, and in some ways they are. But there are a number of significant differences between contexts that students and staff may be unaware of. Graduate school is often a time for students to decide to pursue studies in another country, so being unaware of these differences may put them at a disadvantage.

So here are some of the main differences that I’ve noticed between New Zealand, Japan, and North America. (My experiences with US and Canadian applications were similar, so I’m treating them together.) I’ll focus on a few areas where the differences are the biggest. Everything I write comes from my experiences in anthropology and STS. I think that much of it will apply to other humanities and social sciences, and a little of it to STEM disciplines.

Entrance Exams

The phrase “entrance exam” seems to conjure images of Asia for many people, and indeed an entrance exam was one of the key parts of the application process in Japan.

In Japan, entry to graduate school generally requires one or more scheduled written exams collectively and colloquially known as the “inshi.” In the case of my former employer, the exams included an academic English language test that was required of all applicants to the department and a subject exam specific to the program. There was also an interview/oral exam.

In North America, there are practically no entrance exams in the Japanese sense, a score on the Graduate Record Exam or GRE is needed as part of applications to manyu Ph.D. programs. This is more often the case in the US than Canada, though even in the US, some programs are dropping the requirement. There are also some programs that require an interview.

In New Zealand, there may at times be an informal interview as part of the application process, there are no timed written exams required.

The Research Proposal

In Japan, North America, and New Zealand, applications to graduate programs generally require a research statement that outlines the potential project that the applicant will undertake if admitted.

In Japan and North America, the statement is usually used to gauge the writing and research potential of a student. Once admitted, students usually have some flexibility to alter their intended research. In my case, the research statements for which I was accepted to Master’s and Ph.D. programs in the US and Canada bore little resemblance to the research I eventually ended up doing to graduate. In Japan, there was similarly some flexibility, though students usually did not alter their research projects a great deal once accepted.

A lot of this flexibility has to do with the large amount of coursework that is generally required of graduate students in Japan and North America after they are admitted. In the US, Canada, and Japan, it is common for students to do one to two years of seminars in which they read widely within their field. In many programs in the US and Canada, this coursework prepares students for “field exams” or “comprehensive exams” which test students’ preparedness to move on to independent research. My Ph.D. program in Canada did not have comprehensive exams but the Ph.D. research proposal and oral defence served a similar purpose. (Other programs in the same university did have comprehensive exams.)

In New Zealand, the research statement takes on greater importance. The statement is taken more strongly as a commitment by the student to pursue a specific project if admitted, and there is less flexibility to change direction, especially at the Ph.D. level. Correspondingly, graduate students are not usually required to take courses at the Master’s or Ph.D. level and there are no comprehensive exams. The programs are also therefore much shorter. Whereas some Ph.D. programs in the US may take 7 or more years to complete (during which the successful student will also often receive a Master’s degree) a Ph.D in New Zealand is officially 3 years in length. The student is expected to focus on their doctoral research upon admission. At the university where I currently teach, students will not ordinarily do any courses after completing an Honour’s degree. There are, however, “taught Master’s” degrees in which students spend the majority of their time doing course work. These correspond to the “Master’s by coursework” degrees (versus “Master’s by research”) that can be found in some departments in North America.

Application Fees

In Japan and North America, applicants typically have to pay a fee, which may be more than a hundred dollars, to even submit an application to a graduate program. In New Zealand, no fee needs to be paid. Some universities may offer fee waivers for financially disadvantaged students, but in my case, I paid a few hundred dollars submitting my applications to a number of universities.

Other Aspects

Apart from the importance of the research statement and the requirement for (or lack of) entrance exams and application fees, the other aspects of graduate applications in the three regions are roughly the same. All require the submission of grade transcripts for previous degrees. All require some number of academic reference letters (though this can vary from 1 to 3 letters.) If your first language is not English or Japanese, then you may also have to supply proof of language proficiency with a TOEFL, IELTS, or JLPT score.

Graduate School in New Zealand

What this all points to is that for people who are familiar with the North American style of university education, the New Zealand version of graduate school looks a little peculiar, in some ways even more so than Japan. And so applying to a program here will require an adjustment of perspective.

From my standpoint, doing a Ph.D. at a New Zealand university requires somewhat more preparation and maturity of students than a North American university does. Applicants here will generally need to have a stronger grounding in their discipline and a fairly well developed sense of their research project in order to be considered qualified.

The flip-side of this is the application process in New Zealand is more open. I mean this in the sense that, if you are interested in applying to a program, the staff will be quite open to discussing your potential project and even in some cases helping you develop it prior to your actual application. If you can find an interesting staff member who is also interested in your work, they may become a strong advocate for your acceptance.

Whether or not to go to graduate school in New Zealand, or in any other country for that matter, is not something with a straightforward answer. But for those who do decide to go and to do it in another country, it will pay to know that university is not universal.

Using a Virtual Avatar in Zoom

If we’re all going to be locked down and locked into Zoom, then we might as well have a little fun with it.

Inspired by a posting in a Facebook group for Japanese academics to discuss COVID about using a vtuber (“virtual youtuber”) for Zoom lectures, I decided to investigate how I could use a virtual avatar in place of my plain old mug in a Zoom lecture. It turned out to be incredibly simple and relatively cheap.

The post on Facebook took me to this page (in Japanese) which explains some of the options, both for smartphones and computers. I ended up with Facerig, downloadable on Steam (a great excuse to install Steam on a work computer). It cost me $17.99 NZD.

Facerig uses the webcam on your computer to track your facial motions and maps them onto a computer generated character. The character will move its head, blink, and move its mouth in roughly the same way as your own face.

Facerig can be configured in “Broadcast” mode, which adds a virtual webcam device to your computer. This webcam can be selected in software like Zoom, which lets you transmit your virtual avatar instead of your face, along with your voice (which can be altered, though I haven’t tried this yet.)

Since I teach a course on the anthropology of science and technology, I’m thinking of using this software for one (or all) of my Zoom lectures once courses restart at the end of this month. Over the years I’ve played with the idea of dressing in different outfits for lectures to see how the students interact differently with me. This setup will let me test how they react to a talking dog.