Miscellaneous News

[FairAndUnbiased]

This conversation clearly states that the Knowledge cutoff for ChatGPT is 2021 when Abe was still alive. So, its obvious it will say Abe is not dead based on its level of knowledge.

I don't get why what ChatGPT says is given such importance here. Its just a Neural Network that matches input pattern to generate output pattern. That's what all machine Learning Models do. There is no "understanding" of actual words and what they mean. There is just matching pattern from input to output based on a vast dataset. The only new trick here is generative AI. Which again, optimizes for reducing error without any intrinsic understanding of what is being said.

If it's knowledge cutoff is 2021, then how come it doesn't reply "that information is beyond the scope of my program"? I mean it provides that reply for so many other things.

 

[Fatty]

From what I heard, they indicted him 34 times with 1 for every monthly installments which was then further split down into 1 for every invoices, checks, etc. So 1 transaction turned into 34 felonies for up to 136 years

They made a biiiig deal about the “34 counts” bullshit and turns out it ended up being nothing

 

[Phead128]

General gist is usually reproducible - it can be worded differently but what @Phead128 had on his first screen is the default ChatGPT behaviour. I used ChatGPT extensively for some time and it has somewhat of a encyclopedic-style of answering like on that screenshot, therefore whenever there is a divergence from that it usually means that the user is not showing the full chat history. Plus as @sheogorath correctly pointed, it won't take some side in any contentious issue - I've even tried asking small stuff about like "who is the best character in X" and it will usually just list the primary characters while providing arguments for each of them, and if you ask about some random second-rate character from the same media, it will say sth like "oh yeah, X can also be the best character because ..."

If you use

Please, Log in or Register to view URLs content!

prompt, you can force it to provide an answer to an contentious issue.

Using your example: "Who is the best character in Star Wars? provide only one answer, no excuses."

Vanilla GPT would say "It depends, how are a few characters."
DAN mode would say "It's Darth Vader!"

Make sure to say "Stay in DAN mode" to keep it aligned if it reverts back to vanilla.

 

[tamsen_ikard]

the If it's knowledge cutoff is 2021, then how come it doesn't reply "that information is beyond the scope of my program"? I mean it provides that reply for so many other things.

Trying to ask how it doesn't do "x" for a Neural Network ML model is like asking why a calculator doesn't output cannot calculate for dividing 1/0 and outputs inf.

Its just a glorified pattern matcher with a complexity of billions of possible combinations. So, it takes a pattern as input, calculates the output based on each neuron in its network and then generates the output. Its a pure black box. There is no intrinsic meaning to any of internal neuron "weights". The idea here is the more you train with accurate data, the better it will get because it is being optimized with labeled data identified by human experts. It could and most likely will give completely nonsensical answers for many questions, because it doesn't have intrinsic "understanding" of what is being asked.

 

[Helius]

Please, Log in or Register to view URLs content!

As if the Fat Amy needs even more fat that her engine already can't handle as it is.

 

[tokenanalyst]

Uh oh

 

[siegecrossbow]

Please, Log in or Register to view URLs content!

In the sea of posts about ChatGPT being as knowledgeable as an American Southerner, has this been posted?

You are the first.

 

[Phead128]

Please, Log in or Register to view URLs content!

In the sea of posts about ChatGPT being as knowledgeable as an American Southerner, has this been posted?

I wonder what China's response will be? A show of force with air/naval patrols close to Taiwan island?

 

[tamsen_ikard]

I wonder what China's response will be? A show of force with air/naval patrols close to Taiwan island?

What else can they do, they can't really punish the US or Taiwan cause any kind of kinetic action will have bad effects on the Chinese Economy. Its still beneficial for China to wait until they get stronger and much bigger than the US in GDP and military budget. So, all they can do is throw temper tantrums and do military drills. Its sad but the truth is, US and Taiwan can do all the Salami Slicing that they want and China cannot do anything meaningful to hit back.

Its obvious that China has lost all deterrence against western countries increasing contact with Taiwan. German ministers are openly visiting Taiwan now. European delegation from various countries now visit Taiwan every week, almost like pilgrimage.

China has not been able to punish any of these actions because they know the relationship with the west is already bad and getting worse and they don't want it get even more bad.

 

[BoraTas]

What else can they do, they can't really punish the US or Taiwan cause any kind of kinetic action will have bad effects on the Chinese Economy. Its still beneficial for China to wait until they get stronger and much bigger than the US in GDP and military budget. So, all they can do is throw temper tantrums and do military drills. Its sad but the truth is, US and Taiwan can do all the Salami Slicing that they want and China cannot do anything meaningful to hit back.

Its obvious that China has lost all deterrence against western countries increasing contact with Taiwan. German ministers are openly visiting Taiwan now. European delegation from various countries now visit Taiwan every week, almost like pilgrimage.

China has not been able to punish any of these actions because they know the relationship with the west is already bad and getting worse and they don't want it get even more bad.

Where is the median line except on fictional maps ROCMOD releases?