Ethan Mollick's tweet - "Look, if everyone is worried about students cheating on essays for AI, instructors can just cheat right back. I asked OpenAI to give me an essay question & make a rubric for grading. I had GPT-3 actually write the essay. I then had the OpenAI grade the essay & give comments. ✅ "

@emollick | 100,213 followers

Look, if everyone is worried about students cheating on essays for AI, instructors can just cheat right back.

I asked OpenAI to give me an essay question & make a rubric for grading. I had GPT-3 actually write the essay.

I then had the OpenAI grade the essay & give comments. ✅

Tweet Engagement Stats

Stats are based upon replies and quotes of this tweet

Top Words	%
students	7.7%
essay	6.6%
thread	5.5%
future	5.5%
looks	4.4%
gpt	4.4%
writing	3.3%
anyone	3.3%
any	3.3%
interesting	3.3%
generated	3.3%
questions	3.3%
actually	3.3%
ways	3.3%
More
Less

Top Emojis	%
👏 E0.6 Clapping Hands	2.2%
😂 E0.6 Face With Tears Of Joy	2.2%

Top Applications	%
mobile.twitter.com Twitter Web App · mobile.twitter.com	33.0%
twitter.com Twitter for iPhone · twitter.com/download/iphone	33.0%
twitter.com Twitter for Android · twitter.com/download/android	27.5%
twitter.com Twitter for iPad · twitter.com/#!/download/ipad	3.3%
about.twitter.com TweetDeck · about.twitter.com/products/tweetdeck	2.2%
twitter-thread.com Bot For Thread · twitter-thread.com	1.1%
More
Less

Top Content Types	%
URL	72.5%
Photo	2.2%
GIF	1.1%

Engagement Map

Quotes

Replies

Retweets

Replies and Quotes

Total of 46 replies and 85 quotes found

In reply to @emollick

Internet Ethics

@IEthics

Article from 2019: vice.com/en/article/pa7…

Lars "Land is a Big Deal" Doucet

@larsiusprime

My dad tells me a college story about how all the students pooled together and had only one guy regularly go to class and tape record the lecture while everyone else skipped, and then one day the professor brought in his own tape recorder, hit play, and left.

JP Pardo Guerra also on Mastodon

@pardoguerra

Look, there is a strike at the UC and there is also potential to automate 85% of what graders do. I'm gonna leave the math to you but I'll tell you: supporting higher education also means not giving out mindless homework that is easily automatable on either end.

In reply to @emollick

Antoine Vernet 🇺🇦 (post.news/antoinevernet)

@Antoine__V

There is some work on marking though, it seems to give either full marks or 0. That's not going to fly.

Najwa Alghamdi

@NajwaGhamdi

أحد أمثلة ChatGPT التي تستحق التأمل
1⃣ الاستاذ:يستخدم ChatGPT لكتابة سؤال المقالة التي يجب على الطلاب تسليمها وتحديد معايير التقييم
2⃣ الطالب:يستخدم ChatGPT للإجابة لكتابة المقالة المطلوبة في 1⃣
3⃣ الاستاذ:يستخدم ChatGPT لتصحيح مقالة الطالب في 2⃣ وتقديم النقد

واكتملت الدائرة..

thee j rozwell

@dj_rozwell

homebrew bicameral mind

Bruce Clark

@bruceclarkprof

The "AI cheating apocalypse" as an infinite loop. A thread simultaneously hilarious and chastening.

이은석 / Eunseok Yi

@paparanga

요거 보고 한국어로 좀 따라해봤는데 재밌네요. pic.twitter.com/P6DmGBvjwS

In reply to @emollick

Dr Eliot Attridge

@Delta_Cephei

save thread

In reply to @emollick

Steve Tadelis

@steve_tadelis

holy shit - this is really remarkable (and--maybe it's the luddite in me--absolutely terrifying!)

In reply to @emollick

Some Guy

@JustHereToShare

I could see this impacting a ton of legal writing from law schools to @LegalZoom @donotpay to lots of standard legal filings, even orders & date anyone say, opinions. Have you seen it getting tested out any? If @courtlistener’s archive was fed it, what might result?

Mike Cameron

@mikercameron

So this is where we are in the early days of AI.

If this is what a generalised AI system can do now imagine what a task specific AI will be doing in 5 years time, having been fully trained in you subject area.

The future is here. It’s just lying in wait for you.

BenneHolwerda

@BenneHolwerda

This is going to hit singularity pretty quick.

In reply to @kl4sp

Aslak Raanes

@aslakr

Spar tid, og la greia skrive en vurdering av stud-/elevenes innlevering. Det kan sikkert òg brukes til å skrive svar på en eventuelt klage

shardcore ⧖

@erocdrahs

interesting as this is to us AI enthusiasts, don’t underestimate the coming storm in IRL teaching.

Having a machine that can turn out a “pretty good” essay, that isn’t detectably plagiarised, is going to cause a lot of headaches…

Sam Rye

@sam__rye

The dystopian future of academia - an AI-based tête-à-tête mediated by humans and our quaint submission and marking systems.

Who’s working on a better way of encouraging, nurturing and scaffolding transformative learning experiences?

In reply to @emollick

Ben

@BMunise

Everyone is responding "handwritten essays" but look, when I have 100 students in a single class I barely have time to answer their *emails*.

Veikko Isotalo

@veikko_isotalo

Opettamisen tulevaisuus AI chatbottien aikana pic.twitter.com/2dguE3O1F8

Terri Grassby

@DrGrassby

Take a look at this AI generated assignment, rubric, essay and feedback comments. This will be very hard to detect. Glad we have vivas for our projects. @UniSurreyBioSci @ItsLongLi @TeachingBioche1 @RalphManders @SurreyDietetics @fueledbybananas

Grant

@Brummy

very soon all coursework will need to be done in exam conditions, as any homework just cannot be trusted as authentic - be that art or essay or poetry

In reply to @emollick

DJM

@zeninguem67

Profs could solicit in-class writing samples, train AI on them (one model per student), then use another AI to figure out whether a student’s essay is their own or was written by someone or something else. Or, you know, just tell the students that's what you're going to do.

Jonathan D. Riddle

@jonathandriddle

Like I said, @anthonympetro and @bbhendrickson, robots grading robots.

Gabriel Mathy

@gabriel_mathy

Nice

Lizo

@CryptoCompass2

I was playing around with the new AI. It is powerful. Most human roles devoid of empathy and wisdom will likely be easily replaced sooner than we might think.

Mark Gubrud

@mgubrud

AI shock is here.

And everybody is having a good chuckle, but bear in mind that this is just getting started.

LLMs encode general knowledge the world wide and an inch deep. Now let them also use the internet.

Human-like AGI and machine consciousness are coming. In this decade.

Marc Porter Magee 🎓

@marcportermagee

This is crazy. Just have the bots replace the teachers and the students.

Dana Turjeman

@DanaTurjeman

I guess my work as a Prof is done.

Now just gotta write my articles.

Oh, wait.

David Clinch

@DavidClinchNews

Sending this to my wife who does college essay consulting ->

Steven Van de Walle

@StevenVdWalle

Binnenkort kan ik al mijn werk outsourcen. Studenten ook.

dan

@BierIncognito

If you just feed a student's A-level and GCSE work into an AI to write the essays based on the AI generated questions, and then mark them using AI you can actually cut out the "going to university" part of going to university and save a lot of money

🇺🇦 Ben Kunz

@benkunz

If I were an AI plotting to take over the world, this is exactly what I'd do:

Have students and teachers of knowledge fight over who could use me the quickest in their stead...

Jazz Tehara

@DrJazzTehara

So basically, it’s just not worth going to Uni anymore.

Zhan Li 🌎🌍🌏

@thezhanly

👏👏👏👏👏👏👏👏👏👏👏👏👏

In reply to @emollick

Eric Lawton

@Eric0Lawton

I spent a lot of time responding to government RFPs. They are very structured and it looks like responses are ripe for AI takeover, though I wouldn't bet $millions on them developing an accurate cost and risk assessment.

In reply to @emollick

ｊ𝐀ｍ𝔦є Ⓐ

@Userisjamie

I’m more concerned about the answer… this is MBA material? We just ignoring foreign exchange, subsidies, corp bureaucracy, fear, shareholder demands, monopolies, resources, supply chains, labour law, domestic politics … 🫠

Dr. Tanusree Jain

@TanusreeJain

Smart. Alternatively, chuck assignments all together. There are other more interesting ways to assess students!

In reply to @kalkylert

David Olgart

@DavidOlgart

Bot For Thread

@bot4thread

Thread🧵 unrolled: https://t.co/Toiuhl5V68

Matias Kiviniemi

@PragmaticZone

AI will be to studying what medicine is to sports.. There will be legal and illegal ways but it'll be really hard to succeed without it.

In reply to @emollick

Antwerp

@antewarp2021

Danny Wong (黄永年)

@dannyjnwong

GPT3 gonna change a lot of our industries in drastic ways.

In reply to @emollick

Marco Narajos

@MNarajos

problem solved.

In reply to @emollick

Gavin Jaunky

@GavinJaunky

Is this real?? Looks like a dream for HR/recruitment teams?!

Holland March

@hollandmarching

Uhh, guys?

In reply to @emollick

@V3ry_Offensiv3

I just finished something and then checked my work. pic.twitter.com/54YPdvnCuF

John Burrows

@JohnUChicago

This is my plan too!

In reply to @Obduliez

Vicente L Ruiz 🍏

@VicenteLRuiz

Con permiso, ejemplo extremo (vía @fabianplaza ): twitter.com/emollick/statu…

Toby Walsh

@TobyWalsh

Cheating the cheaters .....

mcc

@0xmcc

Teach 👏 students 👏 how 👏 to 👏 make 👏 technology 👏 work 👏 for 👏 them

design essay questions that assume students will use AI

Rod11

@rod11

It's easy to "hear" the robotic tone in the AI language. But is also very clear that AI tools are going to make it really important for our society to be able determine when a human is involved in an interaction.

Sardines For Dinner

@ForSardines

This shit fucking sucks. Ugh! I hate it!!

Patrick Myers

@pomyers

The AI utopia we are going towards in higher Ed and academia.

Now, who can help me use OpenAI or GPT-3 to answer #Reviewer2 questions and comments?

rory cussll

@rorycussll

Big day for AI!

Sigurður Ingi Ricardo

@Ziggi92

Bara ef við værum svona íslenskt AI

In reply to @emollick

Ayush Kumar

@_Ayu5h

80 points for 10 sec of work, seems more than fair xD

catur

@caturaperkasa

krezi🤯

samchi

@_Buluchi

Google need to respond quicklyyyy

Sam

@samuelglaspel

It makes me wonder if people would still have integrity in the future of AI.

In reply to @MrTomRad

Stephen 🌏 Taylor

@sjtylr

Here's an example:

MBAS

@mbas2709

Ngeriiiii, tp bisa lah dicoba hahaha

Tama

@hiyogaaditama

Dulu mikir,

"Ah gak mungkin sih AI, bisa gantikan para pekerja kreatif, masa iya AI bisa buat gambar, atau nulis."

Sekarang, pemikiran ku udah gak valid.

It scared me, but make me curious as well to wonder what kind of role for me in the future.

Daniel 🛸

@iMicromantic

Mierda.

Mark Schauer

@mtschauer

Shits getting wild.

Gionata

@GionataTweets

The future of education is already here

Ignacio Sepulveda

@Ignacio_S2112

This looks interesting

manny

@pocofeo__

Actually 🥜

In reply to @emollick

MnM

@Mamallra

It won't be free though

ANIA!!

@gurogenrr

humanos: não, não podemos deixar máquinas continuarem aprendendo como nos substituir melhor

também humanos: tá mas só mais um pouquinho

In reply to @emollick

Adam Locklin

@AdamLocklin

This is meta

Fabian Denhoff

@DenhoffFabian

Die Anwendungsmöglichkeiten für KI im Deutschunterricht sind… mannigfaltig 😬

In reply to @emollick

visarga

@visarga

Run this in a loop for 1 billion topics, iterating on the model and creating new training data. That's how you get GPT-4. Just kidding.

In reply to @emollick

James Wan

@James__Wan

Looks quite robust for this task.

waldenmode with an ide

@grokcamp

first time i've ever worried that paperclips might actually be real

In reply to @halftes6

ceci n’est pas une parodie (🐀)

@halftes6

logical extremes

Rizky Banyualam Permana

@rbpermana

beating em in their own game

Divas Parashar

@ParasharDivas

This is straight up sorcery 🤔

@Winzeey

Noted

In reply to @emollick

Burd | druB

@dbac_dynamite

If anyone does this they are a piece of shit and should turn in their degree

In reply to @emollick

Jeff Godun

@JGodun

save thread

Sigmund Aas

@Winter_arrived

Open AI og GPT3 kommer til å redusere bruken av små språk som norsk ganske dramatisk tror jeg

Fifi

@fifi13

Professors can be replaced by machines, good to know.

In reply to @emollick

Maxwell Kronen Von Berg

@mkvb0

To be honest, a lot of this sounds boilerplate. The suggestions for improvement sound like they could apply to any attempt.

Fredo

@fredgrenier12

This has to be the wittiest use 😂👌🏻

In reply to @emollick

mental blanking interval

@vsync

ok I'm angry because I thought we could just use adversarial learning to make the prompt harder and in playing around with concepts I stumbled upon an actual connection I forgot I knew, and now I don't know whether that's why I mashed them up

In reply to @emollick

Victor Petrescu

@VictorPetrescu

The arms race is on

In reply to @emollick

MarcoPolo

@Marco777Polo

Take care, it also just invents stuff that sounds true! When you then ask for the sources you see quite often they don't even exist, but were generated to look realistic.