Group items tagged

In the long run, we believe, teachers need to help students develop a critical awareness of generative machine models: how they work; why their content is often biased, false, or simplistic; and what their social, intellectual, and environmental implications might be. But that kind of preparation takes time, not least because journalism on this topic is often clickbait-driven, and “AI” discourse tends to be jargony, hype-laden, and conflated with science fiction.

Make explicit that the goal of writing is neither a product nor a grade but, rather, a process that empowers critical thinking

Students are more likely to misuse text generators if they trust them too much. The term “Artificial Intelligence” (“AI”) has become a marketing tool for hyping products. For all their impressiveness, these systems are not intelligent in the conventional sense of that term. They are elaborate statistical models that rely on mass troves of data—which has often been scraped indiscriminately from the web and used without knowledge or consent.

The race to build high-performance, AI-powered search engines is likely to require a dramatic rise in computing power, and with it a massive increase in the amount of energy that tech companies require and the amount of carbon they emit.

Every time we see a step change in online processing, we see significant increases in the power and cooling resources required by large processing centres

third-party analysis by researchers estimates that the training of GPT-3, which ChatGPT is partly based on, consumed 1,287 MWh, and led to emissions of more than 550 tons of carbon dioxide equivalent—the same amount as a single person taking 550 roundtrips between New York and San Francisco

When we say that someone has “learned” a subject, we typically mean that they have shown evidence of mastery not only of basic cognitive processes like factual recall and working mechanical exercises but also higher-level tasks like applying concepts to new problems and judging between two equivalent concepts. A student learning calculus, for instance, needs to demonstrate that s/he can do things like take derivatives of polynomials and use the Chain Rule. But if this is all they can demonstrate, then it’s stretching it to say that the student has “learned calculus”, because calculus is a lot more than just executing mechanical processes correctly and quickly.

Even if the student can solve optimization or related rates problems just like the ones in the book and in the lecture — but doesn’t know how to start if the optimization or related rates problem does not match their template — then the student hasn’t really learned calculus. At that point, those “applied” problems are just more mechanical processes. We may say the student has learned about calculus, but when it comes to the uses of the subject that really matter — applying calculus concepts to ambiguous and/or complex problems, choosing the best of equivalent methods or results, creating models to solve novel problems — this student’s calculus knowledge is not of much use.

Khan Academy is great for learning about lots of different subjects. But it’s not really adequate for learning those subjects on a level that really makes a difference in the world.

"Media literacy has long focused on personal responsibility, which can not only imbue individuals with a false sense of confidence in their skills, but also put the onus of monitoring media effects on the audience, rather than media creators, social media platforms, or regulators,"

the need to better understand the modern media environment, which is heavily driven by algorithm-based personalization on social-media platforms, and the need to be more systematic about evaluating the impact of various media-literacy strategies and interventions

In response, bills to promote media literacy in schools have been introduced or passed in more than a dozen states. A range of nonprofit, corporate, and media organizations have stepped up efforts to promote related curricula and programs. Such efforts should be applauded—but not viewed as a "panacea," the Data & Society researchers argue.

We can say "Black lives matter," but we must also demonstrate through our actions that Black lives matter, and that means we must radically alter many of our institutions and practices, recognizing their inhumanity and carcerality. And that includes, no doubt, ed-tech. How much of ed-tech is, to use Ruha Benjamin's phrase, "the new Jim Code"? How much of ed-tech is designed by those who imagine students as cheats or criminals, as deficient or negligent?

"Reimagining" is a verb that education reformers are quite fond of. And "reimagining" seems too often to mean simply defunding, privatizing, union-busting, dismantling, outsourcing.

if Betsy DeVos is out there "reimagining," then we best be resisting

In an online environment, seminars will work best if they occur asynchronously in the discussion boards in an LMS

The 4 key elements for a seminar that need to be replicated during remote instruction include: A prompt or text(s) that the student considers independently in advance Guiding questions that require analysis, synthesize and/or evaluation of ideas The opportunity to share personal thinking with a group Ideas being developed, rejected, and refined over time based on everyone’s contributions

Students need specific guidance and support for how to develop, reject, and refine ideas appropriately in your course.  If you want students to share well, consider requiring an initial post where you and students introduce yourselves and share a picture. Describe your expectations for norms in how everyone will behave online Provide a lot of initial feedback about the quality of posting.  Consider giving samples of good and bad posts, and remember to clarify your marking criteria. Focus your expectations on the quality of comments, and set maximums for the amount you expect to reduce your marking load and keep the discussions high quality. Someone will need to moderate the discussion. That includes posting the initial threads, reading what everyone posts all weeks and commenting to keep the discussion flowing.  Likely, the same person (you or a TA) will also be grading and providing private feedback to each student. Consider making the moderation of a discussion an assignment in your course. You can moderate the first few weeks to demonstrate what you want, and groups of students can moderate other weeks. It can increase engagement if done well, and definitely decreases your work load.

How and why do well-educated, well-intentioned scientists produce biased AI systems? Sociological theories of privilege provide one useful lens.

Scientists also face a nasty subconscious dilemma when incorporating diversity into machine learning models: Diverse, inclusive models perform worse than narrow models.

fairness can still be the victim of competitive pressures in academia and industry. The flawed Bard and Bing chatbots from Google and Microsoft are recent evidence of this grim reality. The commercial necessity of building market share led to the premature release of these systems.

The distinction between a corporation and an algorithm is fading. Does that make an algorithm a person? Here we have this interesting confluence between two totally different worlds. We have the world of money and politics and the so-called conservative Supreme Court, with this other world of what we can call artificial intelligence, which is a movement within the technical culture to find an equivalence between computers and people. In both cases, there's an intellectual tradition that goes back many decades. Previously they'd been separated; they'd been worlds apart. Now, suddenly they've been intertwined.

Since our economy has shifted to what I call a surveillance economy, but let's say an economy where algorithms guide people a lot, we have this very odd situation where you have these algorithms that rely on big data in order to figure out who you should date, who you should sleep with, what music you should listen to, what books you should read, and on and on and on. And people often accept that because there's no empirical alternative to compare it to, there's no baseline. It's bad personal science. It's bad self-understanding.

there's no way to tell where the border is between measurement and manipulation in these systems

The problem waiting round the corner for universities is essays generated by AI, which will leave a textual pattern-spotter like Turnitin in the dust. (Earlier this year, I came across one essay that felt deeply odd in some not quite human way, but I had no tangible evidence that anything untoward had occurred, so that was that.)

To accuse someone of plagiarism is to make a moral charge regarding intentions. But establishing intent isn’t straightforward. More often than not, the hearings bleed into discussions of issues that could be gathered under the heading of student ‘wellbeing’, which all universities have been struggling to come to terms with in recent years.

I have heard plenty of dubious excuses for acts of plagiarism during these hearings. But there is one recurring explanation which, it seems to me, deserves more thoughtful consideration: ‘I took too many notes.’ It isn’t just students who are familiar with information overload, one of whose effects is to morph authorship into a desperate form of curatorial management, organising chunks of text on a screen. The discerning scholarly self on which the humanities depend was conceived as the product of transitions between spaces – library, lecture hall, seminar room, study – linked together by work with pen and paper. When all this is replaced by the interface with screen and keyboard, and everything dissolves into a unitary flow of ‘content’, the identity of the author – as distinct from the texts they have read – becomes harder to delineate.

AI tags photos of women in everyday situations as sexually suggestive. They also rate pictures of women as more “racy” or sexually suggestive than comparable pictures of men.

“You cannot have one single uncontested definition of raciness.”

“Objectification of women seems deeply embedded in the system.”

Think of ChatGPT as a blurry JPEG of all the text on the Web. It retains much of the information on the Web, in the same way that a JPEG retains much of the information of a higher-resolution image, but, if you’re looking for an exact sequence of bits, you won’t find it; all you will ever get is an approximation. But, because the approximation is presented in the form of grammatical text, which ChatGPT excels at creating, it’s usually acceptable. You’re still looking at a blurry JPEG, but the blurriness occurs in a way that doesn’t make the picture as a whole look less sharp.

a way to understand the “hallucinations,” or nonsensical answers to factual questions, to which large-language models such as ChatGPT are all too prone. These hallucinations are compression artifacts, but—like the incorrect labels generated by the Xerox photocopier—they are plausible enough that identifying them requires comparing them against the originals, which in this case means either the Web or our own knowledge of the world. When we think about them this way, such hallucinations are anything but surprising; if a compression algorithm is designed to reconstruct text after ninety-nine per cent of the original has been discarded, we should expect that significant portions of what it generates will be entirely fabricated.

ChatGPT is so good at this form of interpolation that people find it entertaining: they’ve discovered a “blur” tool for paragraphs instead of photos, and are having a blast playing with it.