Has anyone tried the OpenAI ChatGPT?

Faith & Science Conversation
1

Really fascinating to see it work. Russell Moore has an article about it at Christianity Today.

I gave it try by asking:

What is the cosmological argument?

The cosmological argument is a philosophical argument for the existence of a first cause (or first mover) as the explanation for the universe and its origins. It states that everything that begins to exist has a cause, and the universe, being something that began to exist, must have a cause. This cause must be outside of time and space and uncaused, which is commonly referred to as God.

I then asked if an uncaused cause is observable and got an error message.

Edit: Please post any interesting results you come across

2 Likes
2

Interesting, and sounds pretty good. One place I read that professors have caught students using AI and one of the ways they recognize answers generated by AI is that they are “Too good” and beyond the understanding of the students they are teaching.

1 Like
4

It’s a shame as I had some real insights while working through a paper. There were the blue books too where I discovered that I had a way with words if put on the spot.

Apparently even realtors are using the program, as I saw a headline saying that it’s changing how they are doing their job.

6

A math professor that we know tried 3^5 on it, and it said 27. It also couldn’t accurately prove that an odd number times an odd number was odd. My father tried it on questions comparable to those for his class, and said that it looked comparable to a competent student plagiarizing Wikipedia.

1 Like
7

That it appears competent is a noteworthy accomplishment. Its answer on the cosmological argument was decent, and I’m assuming it’s not merely copying and pasting but genuinely composing the sentences in real time.

8

Can the real and natural numbers be put in a one to one correspondence?

Yes, the real numbers can be put into a one-to-one correspondence with the natural numbers. This can be achieved using a method called Cantor’s diagonal argument or construction.

9

Not bad. If someone replied to me with this in a forum, I’d think they knew what they were talking about.

What was Kant’s objection to the ontological argument?

Immanuel Kant’s objection to the ontological argument was that existence is not a predicate, and so cannot be used to affirm the existence of God. He argued that the concept of God as a being with all perfections, including existence, only refers to a possible object of experience, but does not guarantee its actual existence. According to Kant, existence cannot be deduced from an analysis of the concept of God, and can only be known through empirical experience.

1 Like
10

This is of course false. The real numbers are uncountable. Cantor’s diagonal argument is used to show that the real numbers cannot be put into one-to-one correspondence with the natural numbers.

See Cantor Diagonal Method -- from Wolfram MathWorld or

Having said that, although ChatGPT doesn’t always get its facts straight, it seems to be quite clever at language processing. It can frame its answers in the language of the King James Bible, for example. That can be quite entertaining :slightly_smiling_face:

2 Likes
11

I thought it was funny as it’s well known by almost everyone in math, science and philosophy majors.

I have suspicion that the Kant response is not entirely correct, but it’s pretty well informed.

Its answer to a question in political philosophy about what theories try to reconcile fairness and desert was impressive:

Luck Egalitarianism: combines the principles of fairness and desert, by acknowledging that individuals should be rewarded for their efforts and abilities, but also recognizes the role of luck in determining outcomes and seeks to address inequalities that arise due to luck.

Rawlsian Justice: Rawls’s theory of justice as fairness also incorporates both fairness and desert. Rawls argues that basic liberties and opportunities should be distributed equally, while income and wealth should be distributed according to the difference principle, which holds that inequalities should be arranged to benefit the least advantaged members of society.

Capabilities Approach: developed by Amartya Sen and Martha Nussbaum, it emphasizes the importance of expanding individuals’ capabilities and opportunities, rather than simply distributing resources equally. This approach recognizes that inequalities may arise from differences in abilities and efforts, but also considers the impact of luck, social circumstances, and discrimination.

These theories attempt to balance fairness and desert by taking into account both the role of individual effort and ability, as well as the impact of luck and societal factors, in determining the distribution of resources and opportunities.

12

Although grammatically far better than the old “grab sentences online with this keyword” approach of some older cheating software, ChatGPT answers for the science questions I tried had a tendency to stuff in the facts rather than being quite on target for writing style.

Directly asking ChatGPT questions with “you” get “I am just an AI unit”, but I imagine that few students couldn’t figure out how to reword the question.

13

There was an update yesterday to improve its mathematical answers and factuality.

It now answers 3^5 correctly, and I got an answer to a “what is your view” question with the following:

“As a language model, I don’t have personal opinions or views. However…”

14

@heymike3 I was only referring specifically to the math question. I edited my post to reflect that. I have no comment on the points of philosophy, as I don’t have the expertise for that.

15

That is kind of the point of AI. It keeps learning.

The only question I have is, how does it know when it gets things wrong? Does it only depend on the responses it gets. Is somebody watching these interactions? Or is it watching the internet for people talking about it?

1 Like
16

I’d be curious how it’s growth is measured. Would data usage or processing demand be a measurable indicator of learning?

17

No. The quality of information is critical. The more relevant the information is regarding its goals then the faster it will achieve those goals.

When the goals are simple, such as winning a game, then this part is easy. In that case the AI can teach itself. But the goals of this chat AI is nowhere near that easy.

You know… one way it can do this is to have the AI sign up for lots of internet forums and then the length of time it can go without being detected would be a good measure of success or failure. hmmm… makes one wonder how many people on the forums are real people… LOL

I once suggested a modification to the Turing test. That the length of time testing it must exceed the length of time programing it. With the new AI, you would need an additional modification since these are somewhat self-programming. That is getting rather close to the nature of life itself, and it makes me wonder if a modification including learning time would be fair since it seems that would apply to people also.

3 Likes
18

As it learns, there must be some measurable change in data or processing. It might be that the smarter it gets, the less processing demand it has.

19

I’m having a fascinating and informative chat right now:

Can ChatGPT make choices?

It cannot pick a number randomly?

Is this how humans choose a random number?

Can humans make random choices?

If a human cannot choose, does that mean they are mindless?

If the person is not acting, then that means they didn’t do it.

All it would take is a for a single person to act in a single instance for determinism to be false

(as with the following response, its previous responses were similarly cogent)

This is a philosophical argument that is often made in support of the idea of indeterminacy in human behavior, and specifically, the idea of free will. According to this argument, even if human behavior is generally determined by various factors, there could be instances in which a person acts in a way that is not completely determined by their past states and the laws of nature.

However, this argument is controversial and has been widely debated by philosophers and scientists. While some argue that indeterminacy is necessary for free will, others argue that determinism and free will are compatible and that the experience of making choices and having control over one’s actions can still exist even in a deterministic world.

Ultimately, the question of whether determinism is true or whether free will exists is a deeply philosophical one, and there is no widely accepted answer. Different people may have different views on these questions, and the debates surrounding them are ongoing.

20

It’s also pretty neat to see how the program handles follow up questions and comments which assume the program understood the text it just created.

21

I have wondered about this, too. But the bot-ty sounding names don’t read much. This is the list of users sorted assendingly by number of posts read. Sorting by time read, I think would be useful, but that option seems not to be available.

image
image984Ă—906 147 KB

1 Like
22

I tend to be a bit more pragmatic. If we can’t tell if the human mind is entirely deterministic, does it really matter?

Bending back to the topic, these self training AI’s can result in programs and hardware configurations that humans just don’t understand. They are a black box. The comparison to the human brain gets a bit eerie at that point.

1 Like