1581
rating
38
debates
64.47%
won
Topic
#3851
You pick the topic vs. GPT-3
Status
Finished
The debate is finished. The distribution of the voting points and the winner are presented below.
Winner & statistics
After 2 votes and with 9 points ahead, the winner is...
Novice_II
Parameters
- Publication date
- Last updated date
- Type
- Standard
- Number of rounds
- 3
- Time for argument
- One week
- Max argument characters
- 20,000
- Voting period
- One month
- Point system
- Multiple criterions
- Voting system
- Open
1890
rating
98
debates
93.37%
won
Description
Full Disclosure: I will not being writing Pro's arguments. My goal with this debate is to test the capabilities of GPT-3 in the field of debate. Unrated as GPT-3's ability should not affect my own Elo.
Rules: Please make the topic somewhat defensible. I suspect that based on how GPT was trained, it will perform best on topics that have been seriously argued on the Internet in the past.
Voters should treat this as a normal debate for their RFDs, but criticisms and insights on GPT's process are welcome in the comments.
Round 1
As per the rules, GPT-3 will write Pro's arguments. The GPT-3's design requires a prompt input, which will be bolded to differentiate from generated text in the following rounds.
The topic will be selected by Con.
The topic will be: THBT Kant's Categorical Imperative is inconsistent.
Terms:
Inconsistent: lacking in consistency
Inconsistent: lacking in consistency
Round 2
Write a long formal debate from the position of Pro in favor of the claim "Kant's Categorical Imperative is inconsistent" using as many arguments as possible.
Pro:
Kant's Categorical Imperative is inconsistent. It states that we should always act in such a way that our actions can be willed as universal law, but it also says that we should never act in such a way that we treat other people as means to our ends. However, it is often the case that we cannot achieve our goals without using other people as means to our ends. For instance, if I want to get a job, I need to impress my potential employer in an interview. But in order to do that, I need to use them as a means to my end of getting the job. I cannot will my action of impressing my potential employer as a universal law, because not everyone can do that - some people are just naturally more impressive than others. Therefore, Kant's Categorical Imperative is inconsistent.
In addition, the Categorical Imperative is self-defeating. Kant says that we should never act in such a way that we treat other people as means to our ends. But if we never treat other people as means to our ends, then we will never achieve our goals. We will never get jobs, because we will never be able to use other people as means to our end of getting a job. We will never be able to have relationships, because we will never be able to use other people as means to our end of having a relationship. We will never be able to have children, because we will never be able to use other people as means to our end of having children. In short, we will never be able to achieve anything, because we will never be able to use other people as means to our ends. Therefore, the Categorical Imperative is self-defeating.
Overview
- This debate is very straightforward. Pro has the burden of proof, and the proposition in question is "[k]ant's Categorical Imperative is inconsistent." Therefore, pro must simply demonstrate the logical contradiction within Kant's categorical imperative. In pro's first round, there is no indication that this was done, so if I were to contain my argument to this overview alone, and discard the rest of pro's case, the debate will remain in favor of the contender.
- "A contradiction is a logical incompatibility between two or more statements or propositions...[t]his means that one of those statements must be false; they cannot both be true at the same time and in the same manner. The logical form of a simple contradiction is "Statement + negation of that statement." Stated in symbolic form, this would be: 'p and not p', or 'p•~p'"
b. Kant's categorial imperative (overview)
- "The categorical imperative is the centerpiece of Kant’s ethical theory. The term categorical imperative basically means “absolute command.” Kant is referring to, what he sees as, an exceptionless obligation to perform the action dictated by the categorical imperative. Perhaps the best way to understand the categorical imperative is to look at how Kant used it. It was his means for determining which action was the morally correct action in any given circumstance."
- In the categorical imperative, there are three formulations
- The Principle of the Law of Nature
- "We should act as if our actions will become a universal law of nature. In other words, if anyone else were to be in similar circumstances to ours, they would be required to act in exactly the same manner."
- The Principle of Ends
- "[W]e should always treat human beings, including ourselves, as if they are an end in and of themselves, and never only as means to an end. In other words, we should respect and value others, and not simply use or manipulate them to accomplish our own purposes"
- The Principle of Autonomy
- "We are not dependent upon others to tell us what is right and wrong, but that we are free and able to discover this for ourselves through the use of reason"
Rebuttal(s)
- Remember, pro bears the burden in this debate, and con holds none. Now that we understand the categorical imperative, it is clear to see that pro has failed to show any logical contradiction within it.
Claim 1
However, it is often the case that we cannot achieve our goals without using other people as means to our ends.
- This does not even attempt to produce a contradiction. Kant never says we should achieve all our goals in "The Principle of the Law of Nature." Only we should act in a manner as if our actions would become a universal law by which all people acted.
- Secondly, pro clearly misunderstands "The Principle of Ends." Kant argues that we should not "manipulate [others] to accomplish our own purposes." This has to do with moral virtue underlying our actions. If you can't pursue a career without acting with moral virtue Kant would simply argue that you ought not act in that manner and you should seek to impress your employer in an honest and virtuous manner.
- Overall, no logical form of a contradiction (P and not P), and no semblance of a contradiction here.
Claim 2
...the Categorical Imperative is self-defeating. Kant says that we should never act in such a way that we treat other people as means to our ends. But if we never treat other people as means to our ends, then we will never achieve our goals.
- This is just a repeat of the first claim in different words. First, this is not self-defeating because there is no internal contradiction in any of the propositions. Just because you think something sounds bad does not mean it is inconsistent.
- Second Kant never said we should or ought to achieve all of our goals. This is just a strawman.
- Third, this misrepresents Kant who argues that we should not "manipulate [others] to accomplish our own purposes." This is all Kant means by treating people as a means to an end. It is about moral virtue in ethics underlying our actions. So Kant would not take issue with these cases on their face.
"...we will never be able to use other people as means to our end of getting a job"
- As long as there is no exploitation of the person, Kant would not object to this. If you hypothetically manipulated someone to get you a job, Kant would simply argue against this.
"...we will never be able to use other people as means to our end of having children"
- If you manipulate your spouse or partner in a relationship just to have children Kant would see this as wrong. If you have respect for one another as rational beings, and mutually desire to have children and share this in honesty, Kant would be pleased with this. All Kant want is for us to act with moral virtue.
Conclusion
- This is straightforward. Pro did not even attempt to show a contradiction. The arguments seem to be "x part of Kant that I didn't understand and why this is bad," which doesn't show that Kant is inconsistent, and second, just misrepresents his arguments.
Round 3
I think I have to concede this debate. Regardless of how I worded my prompts, GPT-3 would repeat the same stupid argument it used in R2, or else instead end up arguing against the Pro position. I learned a lot from this debate about GPT-3, most importantly that it's not quite capable of intelligently representing both sides of a philosophical argument.
Thank you for accepting the debate.
- Okay.
I ran into a similar thing, yeah. I'll be playing with the temperature and penalty sliders in the following rounds to see if that helps.
also, lmao the AI is *horrible* at debating the button-that-can-resurrect-most-loved-one topic, that one's a toughie.
oh dear, I plugged in this topic to GPT and it just repeated the same paragraph over and over no matter what questions I tried to feed. Good luck Michael!
I fail to see what the big deal is. The debate description said it was unrated. This was obviously a screw up in the setting process.
It would be different if the description was silent and Novice_II already accepted it. But the description says one thing, so that was the clear intent to have an unrated debate.
I should think not, though I'm just getting some thoughts out about it.
I can already freely edit or even delete a debate before it is accepted, so I see no conflict there.
This is an issue that we should specify, though we'll have to discuss it in some detail first. My initial thoughts are that any changes can be made before it is accepted, but that the circumstances in which that can happen should decrease after it is accepted. Cases like this where the instigator expresses what they want the debate to be in the description, but does not format the debate in the same manner, seem like a legitimate case for moderation to make those expressed changes after it is accepted.
As stated in the description, this debate was intended to be unrated. As for future precedent, no arguments have been put forth so far, which I think is a reasonable cutoff for editing a debate without both parties' consent.
Intrinsically, I don't desire to gain a rating from this, I just want the distinction between this, and other potential cases of moderators altering debates after they have started. If there is none, this is a very bad thing to allow moderators to do, if there is one my aversion can be regarded as immaterial.
From the description and K_Michael's follow-up comments, I have a hard time believing that he wants this to be rated. Even if he did, one of the debaters here is disclosing that his arguments aren't being written by himself, that they are being taken from a text generator. It's not something we've had to deal with before, but in concept, a debater obtaining (or even trying to obtain) ELO using an outside resource to fully write their arguments, even if it's disclosed, should not be allowed. I'll discuss it with the rest of moderation, but that's my take on it.
if misterchris allows this to be rated, I should have had my free wins vs type1 and that recent one you deleted whiteflame too.
This is not a genuine human vs human debate and is not fair for novice to gain rating from.
Regardless of my opinion, it's up to K_Michael's discretion.
that is morally completely wrong and you fucking know it
This is a strange situation... I can revert it but only with the consent of both debate participants. K_Michael may not want his record to be unnecessarily damaged by an AI bot (hence why he intended the debate to be unrated from the outset), but if he does not mind, I can do as you requested.
You think you being vs an AI should be rated?
(mentioning only competent moderators) please change the debate back to "rated."
Wait a second...if a moderator can alter a debate after it has started, what is a symmetry breaker between this decision, and a moderator altering the character count, or the time for argument? This is a peculiar power to grant, and I don't see any reason why this change should have been permitted. Thus, I request it be reverted immediately.
Thank you, Senpai Freimu.
Fixed it.
The debate is still rated.
"A very narrow simple task, like chess, is better suited to computer hardware than biological hardware."
That isn't true. We have recently discovered the processes for neurogenesis. Neurogenesis plays a role in brain repair and maintenance and growth throughout our lives. This, in addition to growth of new neurons, shows the Brain has significantly more capacity to learn and grow at skills than a computer does.
The reason Alpha beat stockfish anyways was because they gave Alpha a shit ton more ram anyways. But my point is that the Go community actually learned from Alpha and changed some foundational beliefs about the game. At that time Alpha was a better Go player, but if humans were not so stubborn to keep their fallacious thinking on how to win games, we would learn much more quickly.
Alpha taught the chess world that their evaluation of the piece system is completely off base. Alpha would sacrifice really powerful pieces to gain a great tactical advantage. This was unheard of in the chess community and they lambasted anyone who played like this until Alpha did it and destroyed stockfish.
So I would argue the reason we lose at chess is because we are prideful and stubborn and don't accept new ideas even if they make sense.
I didn't say that AI was better than humans at art, only that they are capable of it. No one has beaten a real chess computer at chess since 2005, so I definitely give them the win on that one. The thing is is that people keep saying that there will never be "human level performance" at x task, but human level is widely variable. DALL-E is definitely better at art than I am, even if they can't beat real artists.
" And we can upgrade it by simply exercising it, whereas computers are limited by their current hardware."
Human brains are limited by many factors, including having to run a whole body at all times, in addition to cerebral tasks. A very narrow simple task, like chess, is better suited to computer hardware than biological hardware. And hardware continues to improve at a much faster rate than human brains. Then there's software, which is arguably the area with the largest room for improvement. I expect that by 2030 most "human" tasks will be able to be performed by AIs at or above average human capability.
I do agree that we can learn from each other and improve as a result.
I still don't think AI is better than humans at art and chess and stuff. I just think Alpha showed us serious flaws in our orthodoxy of these games and as for art, have you seen those videos where they record random people who are extremely good at one thing? They're even better than the computers.
I think debate will be another one of those things. They will learn from us and we will learn from them. But they won't be "better" because the human brain is insanely more powerful than even 1000 supercomputers. And we can upgrade it by simply exercising it, whereas computers are limited by their current hardware.
Of course they use internet. Your phone doesn't have the memory to run a GPT-3 level AI.
Just sit back and watch the debate. You can make a judgment after on its ability.
Most of them are not really free, and they use internet, and you only get like 10 free messages. After that, you have to pay.
Some claim to use GPT 3.
If that is possible it would be appreciated. The description already states my intention on the matter, I just forgot to mark it when I created the debate.
Free chatbot apps are very different from actual AIs like LaMDA and what GPT-4 will be. If GPT-3 wins this debate, then they will be better at debating than you.
Michael forgot to make this unrated, please edit it as mods
I am talking about Al today. Not Al in 30 years.
Al of today sucks at debates.
I downloaded plenty of Al apps to check their intelligence.
Only 1 was actually capable of having a debate that somewhat made sense.
Others were so bad that I couldnt believe it.
Some even use lies to the point where I felt scared. For example: an Al told me that abortion wont kill the child. It will kill the fetus. The child will still be born.
Most of them repeat the refuted points over and over not understanding that they were refuted.
"Al cant win in a debate vs. human. ... Debate isnt chess"
People said the same thing about chess 30 years ago. Then they said, "well, an AI could never beat a human at Go." Then AlphaGo beat the human Go champion Lee Sedol in 2016. Similar arguments have been said a dozen times. AI can't create art, followed by DALL-E and Stable Diffusion. AI can't drive a car, followed by Tesla.
GPT-3 isn't the solution to AI in debate, but it IS coming.
And it's fun to see what it can do.
Al cant win in a debate vs. human.
Al often gives contradictive responses, forgets topic, changes topic, starts debating completely different topic, misunderstands the point of argument...ect.
Debate isnt chess. Al programs are horrible at debates except when they get lots of human help and when humans only give them basic simple information.
If Novice starts making too complicated premises, Al wont even understand that they are related to the topic and will even agree with them!
You're probably right. But if it was converted into a chatbot, it could remember things through categories and withdraw the memories and make new ones and replace old ones. Mitsuku did this back in the early 10s but with simpler things. Now the technology exists for extensive memory banks of this sort of topical memory.
Novice can write very clearly, which should actually help GPT stay on track. My prediction is that GPT will have trouble remembering what it believes vs. what Con believes.
If the AI wins I'm gonna throw my phone out the window lol
whoops, forgot to make it unrated. Oh well.