ChatGPT

My concern with this kind of AI is how it gets fact-checked, and the process and speed with which corrections are implemented. I see this as essentially and advanced version of Wikipedia, which we all know can be useful, but has clear limitations on reliability. The problem is that it "sounds" more authoritative than Wikipedia, but likely is not.

As far as the impacts on academia, I'm not that worried. A teacher should be able to devise examinations in the classroom that clear identify the fakers. Along the lines of "Thank you for turning in your 10 page essays. Now, please sit down and write two page summary of what you just submitted."

As a student who always wrote that ten page essay, I'm glad I'm finished with school, because writing another two pages just because someone else might have cheated would suck...

AI is only as good as the information given it and the people training it. Considering that AI is often considered "impartial", that's a big problem.
 
Probably got confused because you were using the possessive form rather than the plural form of Saturday.
This is actually a fantastic point.. I am a prolific speech to text user, even the most basic speech to text algorithm should understand that "Saturday" would rarely be possessive.

Turing can rest easy
 
"Saturday" would rarely be possessive.

Last Saturday's game.
Saturday's lunch special.
Next Saturday's weather.
and so forth...

But "Saturday's" is not necessarily possessive, as it can also be a contraction of "Saturday is."
Saturday's fine with me.
Saturday's out.
Saturday's game day.
and so forth...

So there are lots of situations where "Saturday's" might be the correct form.
 
6er and I asked ChatGPT about hookers n blow and we didn't get the answers we were looking for. therefore, I won't be using ChatGPT.
 
My concern with this kind of AI is how it gets fact-checked, and the process and speed with which corrections are implemented. I see this as essentially and advanced version of Wikipedia, which we all know can be useful, but has clear limitations on reliability. The problem is that it "sounds" more authoritative than Wikipedia, but likely is not.

As far as the impacts on academia, I'm not that worried. A teacher should be able to devise examinations in the classroom that clear identify the fakers. Along the lines of "Thank you for turning in your 10 page essays. Now, please sit down and write two page summary of what you just submitted."
It is meant to be more "human" than it is meant to be a "perfect encylopedia", so yes, it will be full of all kinds of error (like most humans). I've caught errors in its math, I've caught errors in some of its knowledge, and I've discussed those errors with it. It can walk me through how it came to such a conclusion and admit that it made a mistake. The discussion I had with it about the mistake will potentially prevent it from making the same mistake elsewhere. If you're expecting it to be a perfectly curated dictionary, you're seriously missing out on seeing the practical applications.
 
The issue of bias concerns me the most. I saw someone post two prompts which were identical except the subjects were different, prominent political figures on different sides of the political spectrum. ChatGPT happily gushed about one of them while refusing to respond to the other one. That prompted a few other topics which exposed there’s bias either in the training data or the programming (or both).
 
I also really don't understand why we are trying to create an artificial human. I thought one of the problems computers fix is the human mistake problem. If you want another mistake-prone thinker, have a baby.
Humans are either moving forward, or moving backwards. As a species, we can't just stand still with our current abilities if we don't continue to learn and train. If AI really takes hold and people start relying on it, then I predict mankind will lose their intellectual acumen and begin to irreversibly regress. This is the stuff science fiction is made of.
 
Oh. If they did, I just gave it to them. I'm kind of in the camp that privacy is dead, anyway. But that's for another thread.
Privacy - don't give up - you don't have to outrun the bear, just the other people running with you. Not every bad actor has access to every compromised set of personal data, and no one is looking for you (probably) personally - they're shoving data through big data funnels, correlating and looking for convergence - good attack "candidates" come out the skinny end of the funnel. Limit your number of exposures - example: doctors don't need your SSN, so don't give it to them (mostly they use it for tracking dead beats) - yeah, you have PII out there, but give a little attention to challenging requests for it. A cell phone company asked for my SSN, though I wasn't "financing" a phone. I said I'd rather not give it to them, and to my surprise they went along with it. You can lock your credit at the major credit bureaus, and get a user id/pwd and unlock it when you need to - or get a PIN your creditor can use to get a report. You can limit access by time period - day, month, whatever. Have your browser clear cache on every close - and maybe use something like Firefox and DuckDuckGo - if you're a Gmail/Google user, open a session, check your mail, close the session. Takes longer to say than to do, and start a new session for whatever else you're doing. None of this is perfect, but it doesn't have to be. Even using a VPN will up your privacy game enormously, and VPN is an easy set-up, once and done thing. Facebook can be locked down fairly tight - look at their privacy setting - takes a few minutes, but just one time - and don't answer phishing Q's on FB that help narrow down your age, location, preferences, etc. Take a second look at your LinkedIn profile - if you aren't job shopping go generic - have a Masters? Say so, but not the year or school where you got it. Just lower your profile. . .
 
Last edited:
Given the universe is about 14 billions years old, and life somewhere has presumably been possible for billions of those years, why do we not see the sky filled with evidence of billion year-old technological civilizations?

Because they invent AI and either become lazy, stupid, and descend into a haze of drugs, Monster drinks, and video games, or the AIs simply wipe them out as worthless and annoying.

If Gen-Z is building the AI, then the AIs will inherit Gen-Z's hatred of humanity. Your great grandchildren will be servants of the AI overlord.
 
Given the universe is about 14 billions years old, and life somewhere has presumably been possible for billions of those years, why do we not see the sky filled with evidence of billion year-old technological civilizations?

Because they invent AI and either become lazy, stupid, and descend into a haze of drugs, Monster drinks, and video games, or the AIs simply wipe them out as worthless and annoying.

If Gen-Z is building the AI, then the AIs will inherit Gen-Z's hatred of humanity. Your great grandchildren will be servants of the AI overlord.
I mostly agree with you, except that I am 70+ yrs old (Not Gen-Z) and I am developing a case of hatred of humanity. I'm contemplating becoming a hermit. It could work as long as I have Amazon, electronic banking and direct deposit of my SS. The irony is that those are some of the things causing my hatred.
 
I spent 3 days playing with GPT-3 since it represents potentially tremendous opportunities for my line of work, and I was blown away at how much of an advancement it is over existing platforms. It's a fantastic tool, but one that's far from being what most people think it is.

IMHO, it's incredibly amazing at creative tasks like writing a poem or a script, or generating background information about a topic or generating potential responses to a question given user-supplied constraints, and it has an impressive ability to "talk" in conversational tones. And it's general 'problem solving' abilities are unparalleled in my view; GPT-3 exercises logic really well.

But it has serious shortcomings in factual accuracy when faced with queries that have a defined, objective answer (i.e., non creative tasks). It frequently delivers answers that are incorrect or misleading but appear correct, generating responses that look like they're in the realm of possibility but upon examination can be demonstrated to be completely false. In my experience that's the worst kind of information because "it ain't what you don't know that gets you into trouble, it's what you know that just ain't so". The danger is that people just take the output at face value because it's easy, and then they're walking around with faulty info. Not good.

So my $0.02 is if the cost of getting the wrong answer is high -- don't rely on GPT-3.
Still an awesome, amazing tool, just one that isn't ready to replace human resourcefulness and judgment in finding important, truthful answers.
 
The issue of bias concerns me the most. I saw someone post two prompts which were identical except the subjects were different, prominent political figures on different sides of the political spectrum. ChatGPT happily gushed about one of them while refusing to respond to the other one. That prompted a few other topics which exposed there’s bias either in the training data or the programming (or both).
I've played with that a fair amount and have definitely spotted some things I would say is "bias". That said, it is perfectly willing to talk with you about it, and explain why it took a certain perspective with one individual and a different perspective with another. You're also welcome to debate the reasons you think it may be looking at things incorrectly and it very-well may look at it differently in the future.

Anyhow - experimenting some more - I told it how I wanted it to measure "presidential success" (essentially, taking into consideration what the economy did prior to their election, during, and after). I then asked it to use that formula to analyze all U.S presidents and to tell me who the winner was. When it was purely considering economic impacts, it went with Reagan.

Once you start trying to program out any chance of it being perceived as bias is the same moment you program out the thing that makes it human in its conversation.
 
.
But it has serious shortcomings in factual accuracy when faced with queries that have a defined, objective answer (i.e., non creative tasks). It frequently delivers answers that are incorrect or misleading but appear correct, generating responses that look like they're in the realm of possibility but upon examination can be demonstrated to be completely false.
.

The major media outlets are lining up! They used to have to pay people to do that.
 
I also really don't understand why we are trying to create an artificial human. I thought one of the problems computers fix is the human mistake problem. If you want another mistake-prone thinker, have a baby.
There are lots of things that can only be done with a human today, but could likely be done just as well or better with an artificial human. A "customer success" agent is only as capable as the knowledge they have of your product. Suddenly, AI like this, could consume every email ever written in your company, every slack message, and every support case. It would very quickly become by-far more effective at quickly getting customers to resolution.

Content writing is another. Companies pay an army of people to write crap on the internet to draw views. This AI, today, can already do this better than the humans could.

Some jobs it will replace. Most jobs it will likely supplement and make you more effective. I use ChatGPT (well the paid openai version of it), already, all throughout my daily work. Primarily to help me handle my communications with humans. Rarely do I want it to help me program, I'm good at that, but I can waste a lot of time "brain blocked" writing the perfect e-mail. It busts through those brain blocks for me like they were nothing.
 
@jesse
To me your use case sounds like the ideal one. In the last week I heard a quote that I liked, something to the effect of: "GPT-3 will be to writing what the calculator is to math."
 
I have a paper due tomorrow night and I so want to use this..But I wont....
 
As a writer, I am already dismayed at the rampant illiteracy exhibited by much of today's youth. AI will simply exacerbate that illiteracy, and hasten our descent into mindless, self-absorbed, entertainment, and eventual ennui.

Think of the flesh blobs riding the spaceship Axiom from the old movie WALL-E.

Think of social media, where one's point of existence is nothing more than getting a click as some sort of validation from a total stranger. We are becoming a civilization of circle-jerkers.

I love humanity, particularly the honest search for knowledge and freedom, the yearning to understand the universe and eventually get off this beautiful rock and roam the stars.

Yet, I believe we are doomed. We are living on borrowed time. Both from internecine strife, loss of motivation, denigration of work and effort, geological events, and from all sorts of astronomical dangers. A mile-wide asteroid would put us back into the stone age. A small solar hiccup would fry everything. A few nuclear air bursts would destroy all electrical grids.

I suspect in the long run sentience is not a survival characteristic. It inevitably descends into self-destruction--and I'm not referring to climate nonsense. Hence, another reason, despite the age of the universe, we see no evidence of sentient life.

Or maybe I didn't sleep well and just need a cup of coffee.
 
Hence, another reason, despite the age of the universe, we see no evidence of sentient life.

We ourselves have only been transmitting "intelligent" signals for 130 years or so. Imagine a similar civilization a billion light years away. What are the odds that their "intelligent transmission" phase would have been timed, a billion years ago, such that we'd now be able to receive it?

There may be many many forms of sentinent life in the universe, but the odds against any two finding each other are astronomical (pun intended).
 
Last edited:
As a writer, I am already dismayed at the rampant illiteracy exhibited by much of today's youth. AI will simply exacerbate that illiteracy, and hasten our descent into mindless, self-absorbed, entertainment, and eventual ennui.

Think of the flesh blobs riding the spaceship Axiom from the old movie WALL-E.

Think of social media, where one's point of existence is nothing more than getting a click as some sort of validation from a total stranger. We are becoming a civilization of circle-jerkers.

I love humanity, particularly the honest search for knowledge and freedom, the yearning to understand the universe and eventually get off this beautiful rock and roam the stars.

Yet, I believe we are doomed. We are living on borrowed time. Both from internecine strife, loss of motivation, denigration of work and effort, geological events, and from all sorts of astronomical dangers. A mile-wide asteroid would put us back into the stone age. A small solar hiccup would fry everything. A few nuclear air bursts would destroy all electrical grids.

I suspect in the long run sentience is not a survival characteristic. It inevitably descends into self-destruction--and I'm not referring to climate nonsense. Hence, another reason, despite the age of the universe, we see no evidence of sentient life.

Or maybe I didn't sleep well and just need a cup of coffee.
Maybe you didn't sleep well, and perhaps another cup of coffee might help in some way,
but your post echoes my thoughts exactly. Therefore you must be thinking clearly.
 
Saw an article the other day where they claimed a recent poll found 30% of professionals had used ChaptGPT to write something for work. I'm not sure how they decided to go about getting their sampling group, but I can guarantee that if I asked every person in my facility about using ChatGPT, 98% of them would say "what's ChatGPT?" Lol.
 
Professional whats?

If it’s reporters, pollsters, or politicians, I can believe it.
I had the same question in response to reading it. I can only assume some techie decided to poll a few dozen of his techie friends and colleagues who likely dabble in that realm to begin with, in order to arrive at 30% usage in a professional environment. The article didn't really go into detail about it (typical internet journalism) but it just struck me as a sensational proportion to pass off as typical of a true random sample.
 
If you haven’t been living in a cave, you’ve probably heard of OpenAI’s ChatGPT. It’s now open to anyone who wants to play with it.

https://chat.openai.com/chat

You have to register, but it’s free.

My first query was inspired by a grandchild who was questioning why Christmas was on December 25. It returned a quite natural sounding result of the basic history behind it.

More recently, on another forum it was mentioned that Tesla was adding 23 acres to what I believe was an existing site. Someone said that’s not all that large, only a square about 1,000’ on a side. That didn’t sound right to me, but here’s my query and response:

23 acres is a square how many feet on a side?

23 acres is equal to approximately 968,752 square feet. To find the length of one side of a square that has 23 acres, you would take the square root of 968,752, which is approximately 984. Therefore, each side of a 23-acre square would be approximately 984 feet.

It sure seems the possibilities are endless. Has anyone else played around with this?

A bit off on a tangent here, but this has been creating waves in the academic community for obvious reasons. It may be ChatGPT now, but before that it was wikipedia, and I am sure there was something before that. Unrelated to ChatGPT, I have recently been doing one-on-one oral exams instead of the traditional sit down written exams. I read an article that discussed how all exams were like this for centuries until recently when mass education became common. The funny thing is, oral exams do not take more time than written exams once you account for grading, computing and dealing with student appeals etc.. Within 15 minutes you can get a pretty good sense of how much the student understands. Of course this won't work if you have 200 students, but with a class of 20 it is very practical. There are good reasons why the FAA still has oral exams, and I hope they keep it that way.
 
A bit off on a tangent here, but this has been creating waves in the academic community for obvious reasons. It may be ChatGPT now, but before that it was wikipedia, and I am sure there was something before that. Unrelated to ChatGPT, I have recently been doing one-on-one oral exams instead of the traditional sit down written exams. I read an article that discussed how all exams were like this for centuries until recently when mass education became common. The funny thing is, oral exams do not take more time than written exams once you account for grading, computing and dealing with student appeals etc.. Within 15 minutes you can get a pretty good sense of how much the student understands. Of course this won't work if you have 200 students, but with a class of 20 it is very practical. There are good reasons why the FAA still has oral exams, and I hope they keep it that way.
I would think the lack of written proof of knowledge would be an issue in most settings. You mention appeals, but I would think this would make appeals far more difficult to deal with. Do you record the orals?
 
I would think the lack of written proof of knowledge would be an issue in most settings. You mention appeals, but I would think this would make appeals far more difficult to deal with. Do you record the orals?


I can also see difficulties where students might not all receive the exact same test. Lots of room for subjectivity and bias.
 
I would think the lack of written proof of knowledge would be an issue in most settings. You mention appeals, but I would think this would make appeals far more difficult to deal with. Do you record the orals?

No recording of orals. We discuss the grade at the end of the oral, and if there is a disagreement we resolve it right there. Also, I ask for a self-evaluation from the student. In nearly every case they under-evaluate their performance compared to mine, so when I tell them that they did better that they thought, it all ends well. I have not run into an appeal yet, but I am sure it will happen some day.
 
It was my understanding that modern academia really didn't care about literacy and gainful knowledge as long the check doesn't bounce. The average level of literacy from those with undergraduate degrees that I encounter on a daily basis would seem to support my claim, unfortunately.

Also, those kids need to get off my lawn.
 
ChatGPT passes the tests just fine
In fact, it's like a new calculator for studying
 

another thought.... maybe chatgpt has a sense of humor and was teasing the author of that article. (probably not, but that would be an interesting feature

  • TARS : [as Cooper repairs him] Settings. General settings. Security settings.

    TARS : Honesty, new setting: ninety-five percent.

    TARS : Confirmed. Additional settings.

    Cooper : Humor, seventy-five percent.

    TARS : Confirmed. Self destruct sequence in T minus 10, 9...

    Cooper : Let's make that sixty percent.

    TARS : Sixty percent, confirmed. Knock knock.

    Cooper : You want fifty-five?
 
I have not run into an appeal yet, but I am sure it will happen some day.

I’m guessing you aren’t teaching organic chemistry to pre-meds.

Contrary to myth, org chem was never intended to be a weed-out class, but it was always interesting to see how many students started sophomore year as pre-med, and ended it as pre-law.
 
Back
Top