What I Said About AI Cheating

On Essays, Not on Your Polycule

[adapted from this private thread]

Is prior learning from reading cheating? Just make AI allowed, luddite.

[…] you have to give assignments that require real thought, not automation

[…] ‘postsecondary’ is irrelevant, but if we’re playing this game: the goldingcode.org curriculum is for all ages as most *professors* do not even know this content -- the fallacious appeal to authority undermines your supposed authority here -- did you ever read https://plato.stanford.edu/entries/reasoning-analogy/ ?

When you write for a REAL audience there are no rules of this nature -- can you tell which articles are ‘real’ at adamgolding.substack.com? What is this question relevant to? The soundness of the argument submitted? When there is a code patch to be applied no one cares how the code is generated, they only care if it passes the UNIT TESTS https://en.wikipedia.org/wiki/Test-driven_development

In my case, the differences are clearly marked: an entire generated article will be marked “w/chatgpt” otherwise the text is all HUMAN-WRITTEN with AI COMMENTARY in the form of SCREENSHOTS -- this is the form of writing I recommend -- see, for instance:

Years ago, I was tutoring a Poli Sci grad student in math -- her next project was on “bot detection” online -- I told her if she does it correctly it will flag all the humans who are merely running a ‘script’ in their heads -- the ‘ideologically possessed’ -- you philosophers need to really sit with this idea -- also review the thought experiments wherein the person in the Chinese Room ‘memorizes’ the book -- does this produce real understanding? No, because it is scripted -- should those running scripts in their heads be punished less than those running them externally? No, not really. Original or “Creative” thought is identified by how it compares to the extant literature. With the use of AI thinkers will reach this ‘lit review frontier’ *faster*, not slower.

[AI can also be used to VERIFY that there isn’t prior work they forgot to cite...]

[[I met many of these ‘programmed’ individuals over the years on facebook -- they had been ideologically programmed, as many ‘lone’ shooters are, and in their case they’d been ‘programmed’ by bots on twitter, which I didn’t use -- today, I use a more complex bot, chatGPT, to save time ‘deprogramming’ the people programmed by these simple twitter bots last decade.]]

[…] But you folks probably don’t even use LOGICOLA in logic education -- you’re twenty years behind! What is the shortest path to every philosopher reading Russell & Norvig cover-to-cover? The goldingcode curriculum -- only then will we we see philosophical progress keep pace with computer science, which took the torch from analytic philosophy halfway through the last century -- catch up, don’t CRIPPLE your students.

Teaching coding, I teach students to CHANGE how they use LLMs, not to STOP -- rather than generating code they don’t understand, I teach them to grill it on each line to demand an explanation from the LLM -- usually this is superior to reading the documentation with keyword search, the ‘normal’ method, which of course one also teaches: “RTFM”

[…] This is the absolute best way for Philosophers to use AI today -- to save precious time documenting the many INFORMAL FALLACIES you will encounter online, saving time for real work -- sadly, even when the table is wrong, no one catches it, revealing the remediality of where we are at with informal logic education:

[redacted]

[…] in the future, yes you can save money by doing a lit review slower, but your competition, be it another academic or Putin’s Kremlin, will do it faster. This is why I am no longer a decelerationist, as the cat is out of the chatGPT bag -- I explain the transition here:

[…] you should probably look at this -- this is from a Job Interview I completed today -- the interviewer was an LLM and I argued that they need Philosophers for this work, since Philosophy is prescriptive in a way that Computer Science is not

https://docs.google.com/document/d/1ou1pZHeppqRgN8G7Rg7mVqT7iYqp5lix36a3-J--V3Y

[…] Notice that here, the focus shifts to the meta-level: Prompt-Engineering -- and here the “Prompt-Engineering” uses a clearly established approach in that they are essentially doing “Test-Driven Development” of PROMPTS -- instead of banning AI, REQUIRE that they SHOW ALL PROMPTS USED etc -- just like any normal research log.

Btw, I have taught many subjects at the post-secondary level at UofT, free from some of the ‘noise’ of these formal requirements.

[…] Why do the humans and the bot agree that I’m witty -- I’m not joking! It must just be the laughter that comes from the insight of hearing a truly original thought... to them, but to me this is all obvious so I don’t get the joke... I merely sit with a grim countenance at the demise of real philosophy on the cutting edge of computer science.

[redacted]

You are conflating what the user of the AI is saying with what the AI is saying.

This is all transhuman-phobia.

I have a right to a calculator, imo. You cannot easily test ‘doing math without a calculator’ using an online course either, so it’s not much of an analogy...

If we want to test desert-island conditions, it takes the budget of actually renting an exam room to produce those controlled laboratory conditions for ‘internal validity’, but the education students actually need is ‘external validity’ / ‘ecological validity’

https://en.wikipedia.org/wiki/External_validity

[…] it will become harder and harder to detect the use of generative AI -- I teach humans to apply what I learned in generative AI *in their heads at the piano* -- if you listen to me improvise at youtube.com/adamgolding you will NOT know when I am running AI in my head and when I am not.

As a listener, you cannot distinguish rule-based improvisation (classical AI) from ‘free improvisation’ -- often there is no real fact of the matter since rival process co-occur in the brain simultaneously -- success consists primarily in blurring this very line.

logicola obviously produces MORE understanding than ‘learning logic without a calculator’ so your example is bunk -- the teacher’s manual bears this out with clear statistics:

https://harrycola.com/lc/docs/manual2.pdf

There is a similar educational resistance to using VoceVista documented in the book by its authors.

https://www.vocevista.com/en

[…] In this case, the authority actually goes the other way around anyway, so I wouldn’t appeal to it...

[…] I do recommend that all students challenge ALL AI restrictions via ‘the DEI department’ on grounds of TRANS HUMAN PHOBIA -- no joke, I cared about this long before sexual trans was a big topic.

https://www.reddit.com/.../k1jkwl/transhumanophobia_is_real

So can the hate speech, folks, even an AI can detect it...

[…] THAT is QUITE the fallacy -- let me know when you’ve had some time to sit with the ideas here rather than short-circuiting in this manner. I suppose you also don’t care about your students’ opinions because you think you outrank them -- for shame -- please do not demonstrate this kind of reasoning for others to be CORRRUPTED by... Yours Truly, a MORAL REALIST

https://adamgolding.substack.com/p/morality-is-objective...

Transhumanphobia is immoral.

[…] Grade your instructors with AI, folks...

AI said it faster than I could.. so that’s a WIN

(A win of about 5-10 minutes SAVED for REAL PHILOSOPHY FREE OF ALL FALLACIES)

[…] Well, none of this speaks to what I said above. I expect you all to read without looking at the name of who is speaking, as well as doing BLIND GRADING.

If this is ‘bombarding’ ask AI, it will type slower than me for you.

If you don’t have time to read something, don’t reply.

Ironically enough, including references is a core hallmark of the very power structure contrasted with reason itself here...
—
adamgolding.ca