You pick the topic vs. GPT-3

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.