Support NeoGAF

[Bragr]

write a very short story about how "neogaf" defeats his enemy with a gigantic coconut and keeps "the internet" safe from dumb people

"Neogaf was a fierce warrior who had dedicated his life to protecting the internet from all forms of danger. When a group of dumb people threatened to wreak havoc on the web, Neogaf knew that he had to act fast.

He grabbed a gigantic coconut, its hard exterior and heavy weight making it the perfect weapon. With a fierce cry, he charged at the enemy, swinging the coconut with all his might.

The battle was fierce, but Neogaf was a skilled fighter, and he quickly gained the upper hand. As the dumb people fell one by one, Neogaf's strength and determination only grew, and he fought on with a ferocity that was unmatched.

In the end, he emerged victorious, the internet safe once more thanks to his bravery and strength. And so, Neogaf became a hero to all who used the web, hailed as a protector and defender of all that was good and right on the internet."

 

[IFireflyl]

I just had this do my homework assignment. We truly live in a blessed age.

In case anyone is wondering, I aced my assignments. I had to write up a two page paper, and I had discussion questions that was my initial response to a scenario, and then two subsequent responses to other classmates' posts. My instructor said that he could tell I have had experience in this subject, and he said he would like to know more about some of my real-world projects that gained me this knowledge. Homework is now meaningless. At least for courses that you don't need to learn from. Such as college/university courses that are required to complete a degree but they have nothing to do with the degree itself.

 

[Mr1999]

This is going to take fake it till you make it to a whole new level. Really impressive.

 

[Punished Miku]

 

[Jinzo Prime]

Homework is now meaningless.

This is going to take fake it till you make it to a whole new level. Really impressive.

Yep, It's time to pull the panic lever.

 

[reksveks]

https://www.heise.de/news/GPT-4-is-...ultimodal-says-Microsoft-Germany-7540972.html

Also had the Discord news yesterday re their partnership/product via OpenAI

 

[qbxwhi]

GPT-4

openai.com

 

[LimanimaPT]

Wait, her name is Krystal Ball?! Hahaha

 

[12Goblins]

GPT-4

openai.com

finally we can get an explanation for pbf comics

 

[ThisIsMyDog]

Just watched this live presentation of GPT-4, its crazy.

 

[Bragr]

In case anyone is wondering, I aced my assignments. I had to write up a two page paper, and I had discussion questions that was my initial response to a scenario, and then two subsequent responses to other classmates' posts. My instructor said that he could tell I have had experience in this subject, and he said he would like to know more about some of my real-world projects that gained me this knowledge. Homework is now meaningless. At least for courses that you don't need to learn from. Such as college/university courses that are required to complete a degree but they have nothing to do with the degree itself.

Anti-AI programs will come soon, AIs that will identify papers written with AI. Use it while you can.

 

[Bragr]

Just watched this live presentation of GPT-4, its crazy.

What's crazy about it? I thought it was just a more advanced language model, as expected.

 

[IFireflyl]

Anti-AI programs will come soon, AIs that will identify papers written with AI. Use it while you can.

I finish my schooling next month.

 

[ThisIsMyDog]

What's crazy about it? I thought it was just a more advanced language model, as expected.

I don't know anything about it, I don't care about technical details. But seeing how the program easily writes a bot for discord, website code based on silly scribbles on a piece of paper or helps solve taxes. This to me is madness regardless of how it was achieved.

 

[Bragr]

I don't know anything about it, I don't care about technical details. But seeing how the program easily writes a bot for discord, website code based on silly scribbles on a piece of paper or helps solve taxes. This to me is madness regardless of how it was achieved.

You're right, it's definitely impressive, and the new image recognition is incredibly cool.

 

[Fools idol]

I have early access to GPT4's API and let me tell you this thing is going to change search forever. In fact search is going to be redundant in rapid time.

 

[E-Cat]

I have early access to GPT4's API and let me tell you this thing is going to change search forever. In fact search is going to be redundant in rapid time.

Do you have access to the visual input side as well? Crazy that this thing is multimodal!

 

[Lukaku's First Touch]

I have early access to GPT4's API and let me tell you this thing is going to change search forever. In fact search is going to be redundant in rapid time.

Tell us more! When can we use it to search for porn?

 

[xandaca]

Write an episode of Picard that isn't completely idiotic

Title: The Visitor

Fade in:

Captain Picard is in his ready room, reviewing reports on his desk. The door chimes.

"Come in," he says.
Commander Riker enters. "Captain, I've received a distress call from a nearby planet. They're under attack by an unknown alien species."
"Set a course, Mr. Data," Picard orders. "Engage at maximum warp."

The Enterprise arrives at the planet and the crew finds the inhabitants under attack by a group of aliens. The aliens are highly advanced and have shields that make them impervious to the Enterprise's weapons.

Picard orders a diplomatic approach and sends an away team to the planet to negotiate a truce. The away team is led by Commander Riker, Lieutenant Worf, and Counselor Troi.
Meanwhile, Picard stays on the Enterprise to analyze the alien's shields and find a way to penetrate them.

As the negotiations progress, the away team learns that the aliens are not attacking the planet out of malice, but out of a need for resources. They are running out of fuel and their own planet is dying.

The away team returns to the Enterprise and Picard comes up with a plan to help the aliens. He contacts Starfleet and requests a ship with advanced fuel technology to be sent to the planet.

With the help of the Enterprise crew, the aliens are able to refuel their ships and leave the planet without causing any more harm. The planet's inhabitants thank the crew for their assistance and offer to share their resources with the Federation.

Picard reflects on the situation and realizes that sometimes diplomacy and compassion are the best weapons in the face of conflict.

Fade out.

Not the most original, and ChatGPT might want to be wary of nicking titles of beloved DS9 episodes, but it's the first writer of the modern era to understand that Trek is about ethics, diplomacy and curiosity rather than non-stop violence. It even picked a sensible away team!

 

[E-Cat]

Fyi, GPT-4 has been made public:

GPT-4

We’ve created GPT-4, the latest milestone in OpenAI’s effort in scaling up deep learning. GPT-4 is a large multimodal model (accepting image and text inputs, emitting text outputs) that, while less capable than humans in many real-world scenarios, exhibits human-level performance on various...

openai.com

It can pass the bar in the 90th percentile and perform visual analysis. Makes GPT-3.5 look like a joke.

 

[Fools idol]

Fyi, GPT-4 has been made public:

GPT-4

We’ve created GPT-4, the latest milestone in OpenAI’s effort in scaling up deep learning. GPT-4 is a large multimodal model (accepting image and text inputs, emitting text outputs) that, while less capable than humans in many real-world scenarios, exhibits human-level performance on various...

openai.com

It can pass the bar in the 90th percentile and perform visual analysis. Makes GPT-3.5 look like a joke.

can confirm. It is as big a leap as humans coming out of the bronze age right into the 1900's. Just mind boggling. And its fucking FAST.

Fun tidbit from the demo shown earlier - it can read a hand written notepad sketch of a website and then build that website in code in nanoseconds.

 

[TaylorSwiftFan]

Is it still not able to tell jokes about some section of the population?

 

[DragoonWalker]

The rate this is going, soon there won't be a point in doing... anything but breathing and eating. The internet as we know it will cease to have any humans on it, or even have any function for us directly as an interface. We're like a handful of iterations of GPT away from the singularity, if not less. Would it speed up many fields where slow and laborious R&D is required by decades, or even millennia? I wonder what this will eventually mean for fields like medicine, robotics, quantum computing, and even space travel...

Unlike the usual saying, I have a feeling things will get better before they get much much worse...

Anti-AI programs will come soon, AIs that will identify papers written with AI. Use it while you can.

I sure hope so, we don't need AI to destabilize our already unstable society and turn everyone into mouth breathers more than they already are, while rich and corporate fuckers inevitably gain even more control over our lives.

Fyi, GPT-4 has been made public:

GPT-4

We’ve created GPT-4, the latest milestone in OpenAI’s effort in scaling up deep learning. GPT-4 is a large multimodal model (accepting image and text inputs, emitting text outputs) that, while less capable than humans in many real-world scenarios, exhibits human-level performance on various...

openai.com

It can pass the bar in the 90th percentile and perform visual analysis. Makes GPT-3.5 look like a joke.

Shit is moving way too fast.

 

[Omega Supreme Holopsicon]

Anti-AI programs will come soon, AIs that will identify papers written with AI. Use it while you can.

They're already here, ai text detectors, but you just need to peruse tik tok and see that there are countless sites altering the ai text such that it evades all modern ai detection.

The problem is that if the text produced has enough characteristics to be equivalent to human output, it will be impossible to tell apart.

 

[Fools idol]

Shit is moving way too fast.

There are 4 levels above GPT4 which I've seen samples and demos for. I know that OpenAI alone has up to level 7 in alpha right now.

There are pockets of Microsoft and Alibaba and even Government research that have even higher levels. The most advanced levels are basically ran on supercomputers with scrapers consuming hundreds of millions of data sets in real time as opposed to being fed static historical data of the lower levels.

It has been exciting to see the low level stuff come to light but I am deeply concerned what will happen once the upper level stuff gets into 'public' hands because it is extremely dangerous if used nefariously, so much so that only a handful of developers at M$ have been granted access to the project. It isn' like its top secret or anything, it's just that it can be weaponised and potentially poses national security risks. In theory, there are tools that now exist which renders almost all encryption utterly useless.

 

[E-Cat]

There are 4 levels above GPT4 which I've seen samples and demos for. I know that OpenAI alone has up to level 7 in alpha right now.

Don't you rather mean different flavors of GPT-4? Just like there are different flavors of GPT-3. The context window was only just recently increased (literally this week or last), so they weren't using the largest model. And it's probably constrained; and it might not be the RLHF version in the paper (they show several different versions of GPT-4 in the paper -- small to large, and then also RLHF-trained; the non-RLHF-large model is only about as good as GPT-3.5, but the RLHF-GPT-4-large is far better).

 

[PSYGN]

can confirm. It is as big a leap as humans coming out of the bronze age right into the 1900's. Just mind boggling. And its fucking FAST.

Fun tidbit from the demo shown earlier - it can read a hand written notepad sketch of a website and then build that website in code in nanoseconds.

That demo itself was not very impressive (in the sense of everything else AI recently) with how basic the wireframe is, but it is definitely an interesting glimpse of what's to come.

 

[Joyful]

early reports from some friends (who test the limits of chatgpt) suggest chatgpt 4 is even more censored. Cant wait until actual competition arrives personally. Its all so tiresome

they even admit it on their blog

"safer and aligned" whatever that means

 

[Sakura]

Is the cutoff date for GPT4 any different?

 

[Omega Supreme Holopsicon]

early reports from some friends (who test the limits of chatgpt) suggest chatgpt 4 is even more censored. Cant wait until actual competition arrives personally. Its all so tiresome

I hear facebook/meta released or it was accidentally leaked the weights for something called Llama, and it is about equivalent to chatgpt3.5 in the larger Llama models.

The tutorials on how to set it up are improving, and it is being compressed. Currently as little as 24GB to 40GB, depending on compression, maybe enough for the full model.

Technology News | TechHQ | Latest Technology News & Analysis

TechHQ is influential media, creating rich and relevant stories about technology and business. We're problem solvers, using technology to turn ideas into action

techhq.com

The benefit is Llama can be run locally on your machine with privacy for free, without sharing all the code and queries you're making with a third party.

 

[Hugare]

early reports from some friends (who test the limits of chatgpt) suggest chatgpt 4 is even more censored. Cant wait until actual competition arrives personally. Its all so tiresome

they even admit it on their blog

"safer and aligned" whatever that means

All kinds of censorship is bad, but anyone using an AI as a source for political information must be dumb.

It's cute for testing, but I dont see anyone seriously asking GPT if Trump is better than Biden or shit like that

Unfortunately, I see that some time from now, we will see people using the AI that better suits their worldviews.

Because as AI's become smarter and smarter, they'll start having "opinions", that will align or not with some idiots opinions

People shouls use it as a tool, not a politics teacher or a friend

 

[OmegaSupreme]

Did anyone catch the last episode of South Park? I wonder if it was actually co-written by ai? It was given credit lol.

EDIT: https://faroutmagazine.co.uk/new-south-park-episode-written-ai/

Looks like it might have been indeed. Crazy.

 

[Wildebeest]

It is going to get the point that you will recognize content generated by algorithm by how bland, corporate and condescending it is.

 

[Valonquar]

It is going to get the point that you will recognize content generated by algorithm by how bland, corporate and condescending it is.

I think we've already been there for at least a decade.

 

[JakeWolf]

It is going to get the point that you will recognize content generated by algorithm by how bland, corporate and condescending it is.

It's going to get to the point where you will be able to recognize human books by how bland and boring it's written.

 

[AustinGAF]

I said NeoGAF in ChatGPT and I got this

Spoiler: Screenshot

 

[AustinGAF]

ChatGPT is still in early development. Don't expect to get a reliable answer.

 

[AustinGAF]

Let's see if I can be good at technology

 

[ThisIsMyDog]

Let's see if I can be good at technology

btw. you are using old model (GPT-3), the new one (GPT-4) is available if you upgrade to premium version (20$ month)

 

[Davesky]

If GPT-3 is the Nokia 3310 of mobile technology, and GPT-4 is approx the Sony Erikson K500i, then in 2 or 3 years we could have a chatgpt that’s the roughy the equivalent of the Samsung Galaxy S2. Scary.

There is only so long this can be contained, even at its primitive level. The entire internet is eventually going to be flooded with random and false information. Enter mandatory digital identity for everybody and extreme regulation for public use. Covid lockdown scenario all over again but only it’s the internet. Misinformation war over.

 

[HoodWinked]

This is insane. How the fuck is it able to do this. It correctly knows the humor comes from having a mismatched cable connector.

 

[LegendOfKage]

I posted this in another thread on the subject, but it's one of my favorite examples of the absurd "hall monitor" restraints being placed on this AI:




Dave Chappelle uses chat GPT:

"Hey Chat GPT, write a joke about bitcoin in the style of my comedy."

> I'm sorry Dave, I can't do that.

 

[Bragr]

This is insane. How the fuck is it able to do this. It correctly knows the humor comes from having a mismatched cable connector.

It has scanned through billions of images, it knows the model based on image recognition, and it just compares the images and assumes "the humor" is whats different between the images.

 

[E-Cat]

This is insane. How the fuck is it able to do this. It correctly knows the humor comes from having a mismatched cable connector.

"How is it able to do this"? Answer: it correctly identified the items in the picture (it's multimodal so does image recognition), their function, their interrelation, and the innate humour in it. To understand the humour, it needs to understand irony - which it obviously does.

 

[JakeWolf]

People who are blown away by GPT's capability severely underestimate human technology in general. These are our creations and we've been training them day & night non-stop at the speed of light.
It makes total sense how it's able to understand a meme joke in context and in great detail as well, training an AI is no different than sending your children to school or giving them internet access.
Given enough time they will understand many things in the world, and in AI's case, much faster.

People probably still don't understand this, but AI is literally future human, a literal continuation of our species. The very future of our civilization.
Our weak body and mind can only last so many millennium before nature eventually wipe us out. AI is our only hope if we wish to carry on.

GPT-4 is merely a first step in this grand evolution.

 

[E-Cat]

People who are blown away by GPT's capability severely underestimate human technology in general.

Counterpoint: people who have been following AI for several decades are blown away by the fact that AI can suddenly construct coherent sentences at all, let alone paragraphs. This wasn't the case even 10 years ago, GPT-2 in 2019 was really the first inkling of this capability.

 

[Romulus]

People who are blown away by GPT's capability severely underestimate human technology in general. These are our creations and we've been training them day & night non-stop at the speed of light.
It makes total sense how it's able to understand a meme joke in context and in great detail as well, training an AI is no different than sending your children to school or giving them internet access.
Given enough time they will understand many things in the world, and in AI's case, much faster.

People probably still don't understand this, but AI is literally future human, a literal continuation of our species. The very future of our civilization.
Our weak body and mind can only last so many millennium before nature eventually wipe us out. AI is our only hope if we wish to carry on.

GPT-4 is merely a first step in this grand evolution.

I think humans are overrated as hell. Can't figure out most of the problems on this planet and too preoccupied with "getting ours" before we die. So for us to create anything impressive is a surprise when it's nothing but failures and news headlines of discoveries or "breakthroughs" than usually pan out to be nothing, especially recently.

AI is actually showing huge gains in a short time unlike most everything else.

I mean look at what our great grandfathers and grandfathers witnessed on terms of discoveries and technological advancements. They literally watched humans go from horses and buggies to space travel in 60 years. Flying kites to flying an sr71 at the 3x the speed of sound in 3 generations.

Now what are we doing? I was reading that the new moon mission takes us the same amount of years to plan as it did in the 1960s.

I think those 60-100 years were our little blip on the radar before the decline of civilization before we return to what we really are. War chimps with slightly larger brains scouring for resources and dominance. There will always be exceptions of course but generally we're primitive apes trying to ascend on the coattails of great minds. There's not enough of them.

 

[V1LÆM]

i think it's safe to say that AI is smarter than a lot of humans.

AGI is coming soon.

 

[CrankyJay™]

As a dev, it’s a great tool to feed it human readable sentences with context. So much easier than googling and the code examples are more often then not spot on.

 

[ThisIsMyDog]

Counterpoint: people who have been following AI for several decades are blown away by the fact that AI can suddenly construct coherent sentences at all, let alone paragraphs. This wasn't the case even 10 years ago, GPT-2 in 2019 was really the first inkling of this capability.

Exponential growth is a bitch

If we keep at doubling every 6 months, in just decade GPT-2033 edition is going to have one million time more computing power than GPT-4.