Professors panic over AI
Professors are panicking over artificial intelligence rather than getting ahead of things and teaching students how to use AI tools in their work. It’s the ones who learn to mobilize that kind of technology who will be out ahead. These students will always have technological help within reach. What’s the point of making them memorize things like they’re sitting in a 19th‑century classroom with a slate to write on?
Blork 09:50 on 2023-05-15 Permalink
Kate, I don’t think that’s a very good reading of the immediate situation.
Imagine a history teacher. You know history. You teach history. It’s not your job to teach students how to use AI to write their history papers.
Similarly, you’re a psychology teacher. It’s not your job to teach students how to use AI to write their psychology research papers.
In the meantime, your students are not studying history and psychology; they’re just getting the bots to do the research and write the papers for them. What are you, as a history or psychology teacher supposed to do about that now? Today? This week? Knowing that many of your students are not doing the reading and research because they know they don’t have to?
What if you don’t even understand AI or don’t care about it? You’re a history/psychology teacher, not an AI-usage instructor.
Kate 10:08 on 2023-05-15 Permalink
Then the colleges and universities have to get out ahead of it, because this genie is not going to be crammed back into the bottle.
AI text is useful. It also has to be recast and edited often, to get the information into the right order. I know, because I’ve done some of this when adding footnotes to a book about to be published. (I’ve never used it on the blog.)
I don’t see how you could write an extended essay using AI, not yet. But you can send it to get pieces of information that you need to support your subject. This isn’t something that profs should be trying to lock down, but the schools should instead be mobilizing their information studies people to help both teachers and students use the tools effectively.
jeather 10:26 on 2023-05-15 Permalink
I would love to see what AI would post on this blog, fwiw.
But I remember the concerns when online translations became available, and yet teachers have managed that. Wikipedia, same. Also, frankly, the teachers I have spoken to see two results: one, they edit the AI so much it’s honestly more work than doing it yourself; two, they leave the senences in about “I’m an AI and can’t write your homework”.
Spi 10:39 on 2023-05-15 Permalink
This is the same narrative that was advanced when search engines were a new thing. We don’t need to know everything just how to use Google, but much of education isn’t about simply using tools in order to achieve a result. There’s inherent value behind the process of learning/struggling and developing a deeper understanding of the subject matter rather than just a bunch of loosely associated facts. Most people understand that as it relates to basic/primary education, but it holds true for higher-ed.
Where as search engines are a more effective tool to access information for you to sort/ingest and make use of, AI on its current trajectory it entirely skips that and just outputs results. If you as a person/student don’t have the sufficient classical training/knowledge on a subject to correct/edit the AI and have just become a person asking it a question then what’s your value in the whole process?
Blork 10:40 on 2023-05-15 Permalink
AI is developing 10 times faster than anything colleges and universities can stay ahead of. That’s part of the problem; it’s rolling out so fast, with no checks or balances.
Colleges and universities are notoriously slow on things like this. It would probably take five years to put together a plan for dealing with AI and another two years to create any kind of training for students, and that training would probably not even be for all students.
Remember that the problem isn’t just the writing. College/university papers involve reading and research, and then the writing of a paper about that research. AI lets students skip the research and most of the writing. In other words, they skip 90% of the learning.
Kate 10:46 on 2023-05-15 Permalink
jeather, I hadn’t blogged this piece about a Montreal North bike demonstration of the need for more bike paths in the area, on the weekend.
I asked Chat-GPT to “write a very brief summary of this article in English.” This is the result:
The article discusses the importance of active transportation infrastructure, particularly for cyclists, in promoting safety and encouraging more people to choose biking as a mode of transportation. It emphasizes the need for improved infrastructure, such as protected bike lanes and safer road designs, to create a conducive environment for active transportation and reduce accidents. The article highlights the benefits of investing in cycling infrastructure and calls for increased efforts to prioritize active transportation and enhance road safety measures.
You’ll notice Chat-GPT makes general statements about cycling and infrastructure but overlooks the elements most people here would lead with: where the ride happened and political promises that have been made. So it’s not up to blog standard, not yet.
Blork 10:49 on 2023-05-15 Permalink
One of the key differences (I think) between the resistance to AI and the resistance to other innovations in the past (the internet, Wikipedia, Google, etc.) is that in the past it was mostly people who didn’t know much about the new technologies who were complaining about it.
With AI it’s people who actually know and understand the technology and the sociological implications who are raising the concerns.
The people who resist the resistance tend to be (in my opinion) people who don’t think long-term; they’re just looking at the immediate benefits. Faster letter writing! Faster first drafts of documents! (Etc.)
What they’re not thinking about is the long-term effect of losing a generation of literacy and apprenticeship.
For example, one of the issues with the writer’s strike in the US is the use of AI to create early drafts of scripts, and pitch documents. That benefits the people paying the bills right now. But traditionally that work was done by junior writers as part of their apprenticeship. Junior writers would often work in a writers room for a couple of years before they even touched a script, but in the meantime they were learning from their more experienced colleagues.
But now, AI and other cost-saving changes to writers rooms mean no apprenticeships. What will the result of that be a generation from now if there are hardly any writers that have had the chance to develop their skills and talents by coming up through the system? Plenty of AI bots writing your scripts, but no actual talent there to make them good and original.
Joey 11:04 on 2023-05-15 Permalink
Easy to say that teachers/schools need to ‘get ahead’ of technology that is developing in real-time. The only option I can realistically imagine is to make most/all graded assignments in-person written exams, perhaps using only pen and paper.
@jeather presumably your teacher friends are also considering the possibility that they are being submitted lots of AI-produced work that they can’t detect…
jeather 11:18 on 2023-05-15 Permalink
TBH I am talking to high school teachers, they have pretty good senses of how their students write. I assume this would differ in first year undergrad. But again, turning an AI essay into undetectable AI-ness does require a fair bit of active work (have you ever fixed something translated by a website? ugh).
Kevin 11:42 on 2023-05-15 Permalink
The CEGEP student in my household had classes this semester where they looked at AI text and essays, and they were pretty dismissive of it.
They said that, given the correct prompts, ChatGPT eventually turned out something equivalent to what a high school student would write if they just learned about the topic in the past week. It certainly wasn’t college-level writing and insight.
Ephraim 12:20 on 2023-05-15 Permalink
Chat-GPT and the like also make up things when they don’t know. For example, when asked to write about the owners of a particular item, it just grabbed two random (CIS gendered) names as those of the owners and kept on referring to them.
We, as humans can do certain suppositions from standpoints that AI can’t. For example, we can argue an entire dystopian version of the world if France has won on the Plains of Abraham, whereupon Quebec remains a French colony. Then, would it have been sold by Napoleon as part of the Louisiana Purchase? Then, you need to argue it with citations, citations that you won’t be getting from Chat-GPT and arguments with no citations are just conjecture at University level.
But just to show how silly it can be. Here is some of what it said about “Craig Street” in Montreal. “Today, Craig Street is a vibrant and dynamic commercial district that is home to a variety of shops, restaurants, and businesses. From high-end fashion boutiques to cozy cafes and specialty food stores, Craig Street has something to offer for everyone. The street is also conveniently located near many of Montreal’s top tourist attractions, including the Montreal Museum of Fine Arts, McGill University, and Mount Royal Park.”
Of course, the problem is that Craig street no longer exists and hasn’t since 1976.
How about this article about Mordechai Richler street in Mile-End?
“Mordechai Richler Street is a charming and eclectic street located in the heart of Montreal’s Mile-End neighborhood. Named after the famed Canadian author and Montreal native, Mordechai Richler, the street is a testament to the enduring legacy of this literary icon.”
Garbage in… Garbage out
Meezly 12:34 on 2023-05-15 Permalink
Given how even the ‘godfather of AI’ had recently quit Google to warn humanity of the dangers of AI technology, it may not be so simple just getting ahead of things.
Kate 12:40 on 2023-05-15 Permalink
Ephraim, I asked Chat-GPT about Mordecai Richler Street in the Mile End and it said “I apologize, but I couldn’t find any information regarding a specific “Mordecai Richler street” in the Mile End neighborhood of Montreal. It’s possible that the street you’re referring to may not exist or may have a different name.”
It went on with a couple of paragraphs about Richler.
However! When I asked about Craig Street: “Craig Street, or Rue Craig in French, is a notable street located in the downtown area of Montreal, Quebec, Canada. It runs parallel to the west side of Crescent Street, one of Montreal’s renowned entertainment and dining districts. […] Craig Street is known for its vibrant atmosphere and its association with nightlife and entertainment. It is home to a variety of restaurants, bars, clubs, and shops, making it a popular destination for locals and visitors alike, especially during the evenings and weekends. The street offers a diverse range of cuisines, from international flavors to local specialties, catering to different tastes and preferences” – and several more paragraphs of nonsense.
OK, a final sample, I asked about a nonexistent street and a nonexistent church, and it said in both cases that they didn’t exist. Then I asked about Saint-Enfant-Jésus church, and it says:
The church is known for its architectural and historical significance. It was constructed in the Gothic Revival style and was designed by architect Joseph-Égilde-Césaire Daoust. The construction of the church began in 1906 and was completed in 1908.
The exterior of the Church of Saint-Enfant-Jésus features prominent pointed arches, intricate stone detailing, and a tall bell tower with a spire. The interior of the church is adorned with beautiful stained glass windows, elaborate woodwork, and decorative artwork, creating a serene and reverent atmosphere.
So you see, you do get some useful stuff, but you have to use your judgment.
thomas 13:07 on 2023-05-15 Permalink
@kate In case you are interested here are the results from Google Bard (released last week and I find superior to GPT-4 for summarization) using your prompt.
Here is a brief summary of the article:
Dozens of cyclists in Montreal-Nord protested on Saturday to demand new bike lanes. The Association for Safe Mobility in Montreal-Nord has been advocating for better cycling infrastructure in the borough for years. The city has promised to build two REVs (Réseau express vélo) and a major street bike path in Montreal-Nord by 2027, but cyclists are concerned that these projects will not materialize soon enough. Some elected officials attended the protest, but they were unable to provide any concrete timelines for the construction of new bike lanes. Cyclists are hopeful that their continued advocacy will eventually lead to the creation of safer and more accessible cycling infrastructure in Montreal-Nord.
Here are some additional details from the article:
The Association for Safe Mobility in Montreal-Nord has been holding a “protest ride” every year since 2021 to raise awareness about the need for better cycling infrastructure in the borough.
The group has also been meeting with elected officials and city planners to discuss their proposals for new bike lanes.
The city has said that it is committed to building new bike lanes in Montreal-Nord, but it has not provided any specific timelines for the projects.
Cyclists are concerned that the city is not doing enough to address the need for safer and more accessible cycling infrastructure in Montreal-Nord.
Note that Bard does not currently handle French so I provided it with a translation from GPT-4.
Blork 14:01 on 2023-05-15 Permalink
What a lot of you are missing is that it’s not about the writing; it’s about the research and the learning.
If ChatGPT can write up a reasonable account of some event or thing, and then you spend a few minutes fixing the language so it doesn’t sound like AI, what you have NOT done is the research and reading on the original topic.
You have not LEARNED anything. You haven’t learned HOW to get information, how to SORT information, how to DISTILL information into an idea and a coherent message, and how to WRITE that message.
Before AI ever existed, those were the goals of assigning papers to students. It was not about just having an end product (the paper). It was about CREATING that end product and all the things you would learn in the process. That’s what’s at risk here.
Tee Owe 14:22 on 2023-05-15 Permalink
Totally agree with Blork – the whole point of being a student is to learn, and to learn how to learn. AI is to learning what GPS is to geography – useful adjuncts maybe but in the absence of core knowledge, potentially dangerous.
Tim S. 15:16 on 2023-05-15 Permalink
The irony is that as technology advances, more students will be asked to work with pen and paper and oral exams might even make a comeback. I’ll just take a moment to pat myself on the back for using the lockdown to teach cursive writing to my eldest.
And as Ephraim and Blork are saying, garbage in garbage out. AI is great at digesting what humans have already created, but I don’t see how it will create new knowledge, except through massive output of the thousand monkeys variety.
Blork 15:58 on 2023-05-15 Permalink
Garbage in, garbage out. If we don’t train people how to create non-garbage, we will become more and more reliant on the garbage producers. Result: a generation from now all we will have is garbage and nobody will be smart enough to know the difference between garbage and non-garbage.
Kate 16:16 on 2023-05-15 Permalink
thomas, Google Bard says it doesn’t work in Canada yet! (Yes, I know – use a VPN.)
thomas 16:22 on 2023-05-15 Permalink
Google Bard doesn’t work in Canada because it is only in English, if you sign up with a VPN US server you can use it.
Ephraim 19:02 on 2023-05-15 Permalink
@Kate – Depends on how you ask ChatGPT… I emailed you a copy of the output for the non-existent street
The worry isn’t simply about universities, because with universities, you require citations and well, you have to do it yourself. But with discrimination and not understanding why AI makes certain choices. For example, if AI handles loan applications, we may not understand why it’s turning down an application and if that relates to discrimination. Does it relate to what time I filling in the application? Does it relate to the postal code, neighbourhood, housing, the birth date, their birth day of the week? We don’t know. AI might notice a pattern with people who fill in applications on Wednesday between 9:15AM and 9:30AM that we just don’t really see. But where someone lives might be considered and that might indicate something like the neighbourhood has a higher percentage of PoC. It’s one of the reasons that Europe is talking about a law that will require a way to explain such decisions, so they can be manually overridden if they are based on discrimination.
Ephraim 22:14 on 2023-05-15 Permalink
I also asked Chat GPT about my uncle, a well known holocaust survivor who has written a book about his experience. There are many articles on the web about him in different languages as well. The book is not available on the web, just to purchase, so Chat GPT couldn’t reference the book. But it also didn’t get the right person and mixed his history with someone else’s, mixing up where he met my aunt and said he had passed, when he is definitely still alive.
It also can’t really give good citations behind an argument of supposition, like what France would have done with Quebec if they hadn’t lost in the Seven-Year’s war.