May 29, 2023

ChatGPT Hobbyists

ChatGPT has understandably garnered a huge amount of attention from all corners of academia, from philosophy to economics. One of the more quixotic examples I’ve encountered recently is Robert W. McGee, and his many papers on this topic, such as “Is Chat Gpt Biased Against Conservatives? An Empirical Study”.

A professor of accounting at Fayetteville University, McGee’s biography reads as something like a Marvel Cinematic Universe version of a nerdy academic supervillain. In addition to having published 59 non-fiction books, McGee apparently holds 23 academic degrees, including 13 doctorates, as well as being a world champion in various martial arts, such as Taekwondo and Tai Chi.

Tags:  large-language-models ,  chatgpt ,  sociotechnical-systems ,  measurement ,  evaluation

Apr 13, 2023

Samsung's Encounter with ChatGPT

As ChatGPT continues to ricochet through the news cycle, media outlets are surely on the hunt for new angles they can present to the public in order to keep this story in motion. Among other threads, one that has gained some traction is the question of risks to privacy and security presented by these new systems.

Last week, a number of US outlets reported on data leaks at Samsung, in which three employees (in separate incidents) apparently entered confidential company information into ChatGPT. According to reports, in one case, an employee tried using ChatGPT to help debug code, another to optimize code, and a third to have it produce a summary of meeting notes.

Tags:  machine-learning ,  large-language-models ,  chatgpt ,  data ,  privacy ,  security

Feb 19, 2023

ChatGPT and Sociotechnical Instability

I’ve written about this before, but it’s worth remembering that almost nothing in sociotechnical systems is guaranteed to remain stable for very long. We’ve recently had two great examples of this, with the first being the changes to Twitter, and the second being ChatGPT (and, by extension, the new Bing).

In the first case, a platform which had long seemed relatively static, (especially compared to all the rest), rather suddenly changed hands, which led to major changes in what it delivered. On some level, many of the technical changes to the actual functionality of the site were relatively minor. Much bigger, however, was the impact of many people abandoning the platform for alternatives like Mastodon. Although it seems to me like people have gradually been filtering back to Twitter, most people seem to have the sense that the experience has changed. Obviously more dramatic infrastructural changes, like prioritizing tweets from paying users, could produce even more dramatic shifts. Regardless, it’s a good reminder that what we think of as “Twitter” is the product of a combination of people and software, either or both of which can shift dramatically in a short period of time.

Tags:  large-language-models ,  twitter ,  sociotechnical-systems ,  chatgpt ,  instability ,  archives

Jan 21, 2023

ChatGPT Dominance

I expect that almost anyone reading this will have heard of ChatGPT by now. Released about a month ago, ChatGPT is a system developed by OpenAI which provides text responses to text input. Although details are scarce, under the hood ChatGPT is basically a large language model, trained with some additional tricks (see Yoav Goldberg’s write up for a good summary). In other words, it is a model which maps from the text input (treated as a sequence of tokens), to a distribution over possible next tokens, and generates text by making repeated calls to this function, and sampling tokens from the predicted distributions.

Tags:  machine-learning ,  large-language-models ,  chatgpt ,  open-science ,  evaluation

Dec 12, 2022

AI, software, and governance

In a recent article covering the FTX collapse, the New York Times described large language models (LLMs) as “an increasingly powerful breed of A.I. that can write tweets, emails and blog posts and even generate computer programs.” There is a lot that we could pick apart in this definition (e.g., what makes LLMs part of a “breed”, what distinguishes the ability to write an email as opposed to a blog post, etc.), but for the moment I’d like to focus on the term “A.I.” (henceforth “AI”). Referring to LLMs as an example of AI is certainly not atypical. Indeed, it increasingly seems like LLMs have become one of the modern canonical examples of this concept. But why is it that we think of these systems as members of this category? And how much rhetorical work is being done by referring to LLMs as a type of “AI”, as opposed to “models”, “programs”, “systems”, or other similar categories?

Tags:  artificial-intelligence ,  machine-learning ,  algorithmic-decision-making ,  governance ,  software ,  large-language-models ,  sociotechnical-systems ,  power

Sep 23, 2022

Hacking LLM bots

For anyone who missed it, a Twitter account named @mkualquiera recently deployed what seems like a kind of adversarial attack in the wild on a large language model (LLM)-based Twitter bot. I’ll link to the key post below, but it’s worth providing a bit of context, as it wasn’t immediately clear to me what was going on when I first saw the tweet.

Tags:  machine-learning ,  large-language-models ,  twitter ,  bots ,  sociotechnical-systems

Jul 3, 2022

Report from FAccT 2022

The fifth iteration of FAccT (the ACM Conference on Fairness, Accountability, and Transparency) was held earlier this month (June 21–24) in Seoul, South Korea. More than just a hybrid conference, this was actually a full in-person conference, combined with a full on-line conference. These happened in parallel, with virtual sessions starting before and continuing after the in-person component each day. Around 500 people attended in person, with another 500 participating remotely.

Tags:  accountability ,  algorithmic-decision-making ,  FAccT ,  face-recognition ,  large-language-models ,  machine-learning ,  science ,  sociotechnical-systems

May 14, 2022

Counting Deaths

Although morbid, it’s fascinating to read a recent article in the NYT about efforts in Sierra Leone to use “electronic autopsies” in a large scale attempt at counting deaths. According to the article, this undertaking is part of a broader effort at data collection, including questions on age, religion, marital status, etc. The novelty, it seems, is in trying to be thorough with respect to what people have died of (including extensive questions about symptoms), even though this information is being collected potentially long after the fact.

Tags:  data ,  statistics ,  quantification ,  mortality ,  archives ,  demography ,  biopolitics ,  governmentality ,  Michel-Foucault ,  John-Graunt ,  Ian-Hacking

Apr 25, 2022

Modular Domain Adaptation

Despite their limitations, off-the-shelf models are still quite widely used by computational social science researchers for measuring various properties of text, including both lexicons, like LIWC, and cloud-based APIs, like Perspective API. The approach of using an off-the-shelf model has some definite advantages, including standardization and reproducibility, but such models may not be reliable when applied to a domain that differs from the ones on which they were developed…

Tags:  machine-learning ,  domain-adaptation ,  measurement ,  text-classification ,  lexicons

Mar 12, 2022

Stability and Change

One of the biggest frustrations with software is that things are constantly changing. From operating systems to apps to web interfaces, things rarely remain the same for very long, especially for users of Windows or MacOS.

There are many reasons for this of course. For decades, hardware has continued to improve at a steady rate, and so software is constantly being rewritten to take advantage of the latest capabilities. Moreover, the incredibly sloppy standard for software quality and reliability (compared to traditional engineering disciplines) means that even the most professional software is shipped with massive numbers of bugs and vulnerabilities, which constantly need to be patched. This is a particularly large problem in institutional settings which are not set up for this pace of updates; some of the worst effects of the WannaCry ransomware attack, for example, were on hospitals that were still using hopelessly out of date Windows machines.

Tags:  archives ,  software ,  wayback-machine ,  stability ,  change