• Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.

ChatGPT 4 Vision unveiled. Can see, hear, and speak

mortal

Banned
I imagine the most plausible use of this tech is in making rich people richer at the expense of poor people, if I'm honest.

I don't think it's likely that the most profitable way to use this tech is to give everyone a virtual Scarlett Johansson. Though, I expect that would be popular too.
People will continue to find ways to capitalize on the tech regardless, so that's a given.

I was speaking more about the implications of everyday use cases. We're arguably part way there with the current iterations of Google, Alexa, Siri, and most recently ChatGPT.
I just think the science fiction portrayed in Her is closer to our trajectory than robot skeletons with laser guns controlled by an evil sentient AI.
The societal impact will be more nuanced with respect to work & relationships, similar to what was explored in that film, at least imo.
 

EviLore

Expansive Ellipses
Staff Member
t67Oys6.jpg
 

Kenneth Haight

Gold Member
Me sitting here with my Chatgpt app trying to send it dicpics for free affirmation and confidence.

episode 4 lady GIF
 

Trunx81

Member
I once read that everything we get as public has already been tested for years in the military and secret services. Imagine what THEY can do already.

PSA: Watch the new Mission:Impossible. It´s good.
 

Fools idol

Banned
We're accelerating exponentially at this point. Pure scifi territory.

As I said in previous threads.. we are in dangerous waters.

The tech the public has access to is nothing compared to behind closed doors. I have heard and seen some shit, lol.

I am truly concerned what Lockheed martin is working on. Open AI has an office in their fucking HQ.
 

poppabk

Cheeks Spread for Digital Only Future
Is it just me or is this interpretation of the cartoon both completely wrong in some aspects - the true image is not recealed as a square, and also off base on the interpretation - it misses the part about compromise completely and some of it is supremely vague as to be worthless- the vague panel with the words 'ah has represents an ah ha moment.
It can do deep interpretation and analysis of images:

lLLRah7.jpg
 
Last edited:

kruis

Exposing the sinister cartel of retailers who allow companies to pay for advertising space.
As I said in previous threads.. we are in dangerous waters.

The tech the public has access to is nothing compared to behind closed doors. I have heard and seen some shit, lol.

I am truly concerned what Lockheed martin is working on. Open AI has an office in their fucking HQ.

AI is the future of warfare. Why have a human controlling war machinerey when it can hunt for targets all by itself? Just think how frighteningly effective AI operated war planes, bombers, tanks, drone swarms, etc will become at killing humans after a bit of practice in the real world.

Future armies will have swarms of AI operated killing machines operating without any human intervention. That's already scary, but then there's always the risk of AI's becoming sentient. This is an actual, real threat to all of us.

 
This tech is here to stay. I'm getting great help from it in my studies. Not asking anything too advanced, but it has given clear and concise answers that's been accurate so far. It's only as good as the data though.
data true, garbage in garbage out, if it's for statistics or calculations. But if you feed it wrong code or wrong formulas it can correct them and provide the correct code or formula.
As I said in previous threads.. we are in dangerous waters.

The tech the public has access to is nothing compared to behind closed doors. I have heard and seen some shit, lol.

I am truly concerned what Lockheed martin is working on. Open AI has an office in their fucking HQ.
What will be interesting is when they get future versions of this into designing new tech. That's when we get the singularity. Right now it is basically superhuman in its vast knowledge base, and even in some aspects of reasoning. Diseases that have eluded a dozen doctors have been diagnosed. It can translate from multiple human languages and knows multiple programming languages.
 

Tams

Member
Is it just me or is this interpretation of the cartoon both completely wrong in some aspects - the true image is not recealed as a square, and also off base on the interpretation - it misses the part about compromise completely and some of it is supremely vague as to be worthless- the vague panel with the words 'ah has represents an ah ha moment.

To be fair, the cartoon is shit to begin with.
 
GPT3 (lately) has been getting me about 70% of the way there on various programming and dev ops tasks.

I consider it to be used as a learning aid and not a complete replacement.
 

StreetsofBeige

Gold Member
Does anyone know how all this AI works?

The examples we've all seen range from making pictures, stories, the bike repair video, evilore's video above about 4.0 can write code base on images etc....

There's a shit load of functionality.

Are these features out there in cyberspace and all AI does is scrape the net for the best site to solve it? Or are there different people who program their own AI code like one guy will do an image to code feature set, one guy does a tell me a story module, one guy programs it to make an image based on typing in Football etc....

it just seems so clever these programs can do multiple things all at once.
 

EviLore

Expansive Ellipses
Staff Member
Does anyone know how all this AI works?

The examples we've all seen range from making pictures, stories, the bike repair video, evilore's video above about 4.0 can write code base on images etc....

There's a shit load of functionality.

Are these features out there in cyberspace and all AI does is scrape the net for the best site to solve it? Or are there different people who program their own AI code like one guy will do an image to code feature set, one guy does a tell me a story module, one guy programs it to make an image based on typing in Football etc....

it just seems so clever these programs can do multiple things all at once.
Neural networks with a structure inspired by the human brain. They have billions of digital neurons that get trained to perform tasks accurately. Just like electrical connections in your brain form when you learn something, the digital neurons have weights (numbers in matrix multiplication) that adjust to learn how to perform a task well. After the learning, you no longer need the data directly, just the model functioning as a brain, which will then handle the desired task.

Then that’s paired with a bunch of software engineers developing features and implementing adjustments of the models to fine tune everything. Difficult and complex work for cutting edge r&d like what OpenAI is doing.
 

Yoda

Member
It's impressive they're still running circles around Google on this front given the vast disparity in resources. Someone needs to replace Alexa/Google Assistant/Siri with this ASAP, they might finally be useful for tasks other than turning off the lights and playing music.
 

Fools idol

Banned
Does anyone know how all this AI works?

The examples we've all seen range from making pictures, stories, the bike repair video, evilore's video above about 4.0 can write code base on images etc....

There's a shit load of functionality.

Are these features out there in cyberspace and all AI does is scrape the net for the best site to solve it? Or are there different people who program their own AI code like one guy will do an image to code feature set, one guy does a tell me a story module, one guy programs it to make an image based on typing in Football etc....

it just seems so clever these programs can do multiple things all at once.

yes. I am a world leading expert in the field, sold two companies in the past in the AGI sector.
 

Fools idol

Banned
I once read that everything we get as public has already been tested for years in the military and secret services. Imagine what THEY can do already.

PSA: Watch the new Mission:Impossible. It´s good.
When I was building early stage AGI in 2014 the stuff we see now in chat gpt was already in beta. That was ten fucking years ago.

They have had almost a decade of time ahead of the public release. I am pretty convinced at this point that China is still way further ahead with this stuff than the USA, because almost all of the developers back then were from hong kong and or shanghai. The chinese government saw the potential and dangers a long time ago and very rapidly put a stop to any sharing of information. Every single non-US or non-english native developer working on the company suddenly returned home on the same weekend. Were talking 30+ people just upping and leaving over night. No idea how they even got the travel arranged, they just were there one day and gone the next, with all of their laptops.

A few months after that the US government got involved and we were forced to sell our shares privately to an unknown investor. Never heard a thing about it again until Chat gpt came out.

This technology was revolutionary 10 years ago, and was being used to find new drugs 10 years ago as wel, and from what I hear (and this is the scary part) was also being used to help create biological weapons and viruses. It would not be a stretch to me in the past few years if this had a hand in either the creation of the covid virus or the vaccine itself. It really wouldnt, but I don't care to speculate too much.
 
Last edited:

supernova8

Banned
Lol I just logged in and the sense of FOMO being on ChatGPT 3.5 right now is so painful. Still, I don't think I need it in my life enough to be paying $20/month for Plus.
 
Last edited:

StreetsofBeige

Gold Member
When I was building early stage AGI in 2014 the stuff we see now in chat gpt was already in beta. That was ten fucking years ago.

They have had almost a decade of time ahead of the public release. I am pretty convinced at this point that China is still way further ahead with this stuff than the USA, because almost all of the developers back then were from hong kong and or shanghai. The chinese government saw the potential and dangers a long time ago and very rapidly put a stop to any sharing of information. Every single non-US or non-english native developer working on the company suddenly returned home on the same weekend. Were talking 30+ people just upping and leaving over night. No idea how they even got the travel arranged, they just were there one day and gone the next, with all of their laptops.

A few months after that the US government got involved and we were forced to sell our shares privately to an unknown investor. Never heard a thing about it again until Chat gpt came out.

This technology was revolutionary 10 years ago, and was being used to find new drugs 10 years ago as wel, and from what I hear (and this is the scary part) was also being used to help create biological weapons and viruses. It would not be a stretch to me in the past few years if this had a hand in either the creation of the covid virus or the vaccine itself. It really wouldnt, but I don't care to speculate too much.
Maybe I'm out of the loop, but I think a lot of people just caught on to AI the past year with Chat3.0. What happened to 1.0 ad 2.0? Were these even available for the public to download and test out or were sites already using earlier forms? Or they were but shitty and nobody cared at the time?
 

Toons

Member
Fortunately most humans are too stupid to use this for anything effective.

Government regulation will become a necessity eventually.

On the bright side this has positive implications for data security.
 

Fools idol

Banned
Maybe I'm out of the loop, but I think a lot of people just caught on to AI the past year with Chat3.0. What happened to 1.0 ad 2.0? Were these even available for the public to download and test out or were sites already using earlier forms? Or they were but shitty and nobody cared at the time?

The main difference between GPT-2 and ChatGPT is the training data really. And they were created for different things.... GPT-2 was designed for what we called NLP's (natural language process).

GPT2 was for tasks like text completion, understanding, summarisation, and translation. The data for it was fed on was different, mostly libraries worth of digital literature, and some cherry picked text from the internet mixed in.

ChatGPT was designed for conversational AI and chatbots, and the key difference is that it was trained on a dataset of human-to-human conversations. ChatGPT is optimised to generate human-like responses in a conversational context, whereas GPT-2 is more suited for general-purpose language tasks.

I can explain the next levels briefly, and what you can expect going forward ;

Gpt 4 combines image processing capability and generation with the various input / output recognitions like voice and visual sign language comms.

GPT5 is essentially the 'brain upgrade' which will improve accuracy, speed and capacity of data input output. The training models will be fully self governed at that point relying far less on human intervention during the training input stage.

the furthest out I was aware of was GPT8, which is what we called special operations back then, which is capable of building civilisation level models and projections of future events based on vast human history and behavioural data. it was being used at the time by developers who had access to accurately predict stock market moves. There is a pioneering company in this space which uses a very famous algorithm in the financial markets called 'Perelman fields medal' and it was used by blackrock and the us government to control financial markets and manipulate assets. It was created by a man named Alex Vieira, and all of the richest investors in the world would meet with him annually to discuss their control of markets and currencies.

It can also be used to suggest media campaigns to sway elections, opinions and sentiment in any given market in real time. They would select and test it in third world countries to manipulate markets and public opinion of various subjects and monitor results. It was shockingly capable and accurate. Again I iterate this was available ~6-7 years ago. The algorithms of social media now are heavily entwined by this technology, no doubt about it, which is why I'm so concerned. People are being subconsciously led by this technology already.

This is a rabbit hole that goes very deep, but it is for the most part readily available information. I believe the company goes by the name of Intuitive Code, is based in portugal. This is bordering on illuminati like stuff.
 
Last edited:

Tams

Member
The main difference between GPT-2 and ChatGPT is the training data really. And they were created for different things.... GPT-2 was designed for what we called NLP's (natural language process).

GPT2 was for tasks like text completion, understanding, summarisation, and translation. The data for it was fed on was different, mostly libraries worth of digital literature, and some cherry picked text from the internet mixed in.

ChatGPT was designed for conversational AI and chatbots, and the key difference is that it was trained on a dataset of human-to-human conversations. ChatGPT is optimised to generate human-like responses in a conversational context, whereas GPT-2 is more suited for general-purpose language tasks.

I can explain the next levels briefly, and what you can expect going forward ;

Gpt 4 combines image processing capability and generation with the various input / output recognitions like voice and visual sign language comms.

GPT5 is essentially the 'brain upgrade' which will improve accuracy, speed and capacity of data input output. The training models will be fully self governed at that point relying far less on human intervention during the training input stage.

the furthest out I was aware of was GPT8, which is what we called special operations back then, which is capable of building civilisation level models and projections of future events based on vast human history and behavioural data. it was being used at the time by developers who had access to accurately predict stock market moves. There is a pioneering company in this space which uses a very famous algorithm in the financial markets called 'Perelman fields medal' and it was used by blackrock and the us government to control financial markets and manipulate assets. It was created by a man named Alex Vieira, and all of the richest investors in the world would meet with him annually to discuss their control of markets and currencies.

It can also be used to suggest media campaigns to sway elections, opinions and sentiment in any given market in real time. They would select and test it in third world countries to manipulate markets and public opinion of various subjects and monitor results. It was shockingly capable and accurate. Again I iterate this was available ~6-7 years ago. The algorithms of social media now are heavily entwined by this technology, no doubt about it, which is why I'm so concerned. People are being subconsciously led by this technology already.

This is a rabbit hole that goes very deep, but it is for the most part readily available information. I believe the company goes by the name of Intuitive Code, is based in portugal. This is bordering on illuminati like stuff.

How do we know you aren't an LLM/A.I.?
 

Fools idol

Banned
How do we know you aren't an LLM/A.I.?
haha 🤣

well all that I will say is that a lot of my work, and my companies work, went into the early stage development of this technology and many other algorithms used by big tech.

I am not entirely proud of it, and strongly disagree how it is being used by many government and corporate entities.

These kind of scumbags even had a youtube channel back in 2016 with hundreds of videos bragging about market manipulation and crashing markets.



the SEC and gov did nothing, can do nothing. These guys have so much money, influence and power thanks to this technology.

What if I told you that this guy and his algorithm was solely responsible for the entire AMC stock fiasco, and that the story that was given on the news was entirely bullshit? this has been in plain sight for a long time.



TLDR, these guys were instructed to 'create' the frenzy in AMC stock and announced the plans to do so months before it happened. It was brazen and proves how powerful and dangerous this technology is. The behavioural models are so strong and accurate. Imagine how many financial lives were wrecked by this.

they were also heavily involved with the rise of crypto. Able to use the models to predict well in advance of moves, regulatory changes and everything.

its scary and gross. But it is reality. These videos are in full view for 7+ years
 
Last edited:

Ikutachi

Member
It's cool what it can do for gaming. It could make a SM64 romhack so imagine more refinements in the future.

 

EviLore

Expansive Ellipses
Staff Member
yes. I am a world leading expert in the field, sold two companies in the past in the AGI sector.
When I was building early stage AGI in 2014 the stuff we see now in chat gpt was already in beta. That was ten fucking years ago.

They have had almost a decade of time ahead of the public release. I am pretty convinced at this point that China is still way further ahead with this stuff than the USA, because almost all of the developers back then were from hong kong and or shanghai. The chinese government saw the potential and dangers a long time ago and very rapidly put a stop to any sharing of information. Every single non-US or non-english native developer working on the company suddenly returned home on the same weekend. Were talking 30+ people just upping and leaving over night. No idea how they even got the travel arranged, they just were there one day and gone the next, with all of their laptops.

A few months after that the US government got involved and we were forced to sell our shares privately to an unknown investor. Never heard a thing about it again until Chat gpt came out.

This technology was revolutionary 10 years ago, and was being used to find new drugs 10 years ago as wel, and from what I hear (and this is the scary part) was also being used to help create biological weapons and viruses. It would not be a stretch to me in the past few years if this had a hand in either the creation of the covid virus or the vaccine itself. It really wouldnt, but I don't care to speculate too much.
The main difference between GPT-2 and ChatGPT is the training data really. And they were created for different things.... GPT-2 was designed for what we called NLP's (natural language process).

GPT2 was for tasks like text completion, understanding, summarisation, and translation. The data for it was fed on was different, mostly libraries worth of digital literature, and some cherry picked text from the internet mixed in.

ChatGPT was designed for conversational AI and chatbots, and the key difference is that it was trained on a dataset of human-to-human conversations. ChatGPT is optimised to generate human-like responses in a conversational context, whereas GPT-2 is more suited for general-purpose language tasks.

I can explain the next levels briefly, and what you can expect going forward ;

Gpt 4 combines image processing capability and generation with the various input / output recognitions like voice and visual sign language comms.

GPT5 is essentially the 'brain upgrade' which will improve accuracy, speed and capacity of data input output. The training models will be fully self governed at that point relying far less on human intervention during the training input stage.

the furthest out I was aware of was GPT8, which is what we called special operations back then, which is capable of building civilisation level models and projections of future events based on vast human history and behavioural data. it was being used at the time by developers who had access to accurately predict stock market moves. There is a pioneering company in this space which uses a very famous algorithm in the financial markets called 'Perelman fields medal' and it was used by blackrock and the us government to control financial markets and manipulate assets. It was created by a man named Alex Vieira, and all of the richest investors in the world would meet with him annually to discuss their control of markets and currencies.

It can also be used to suggest media campaigns to sway elections, opinions and sentiment in any given market in real time. They would select and test it in third world countries to manipulate markets and public opinion of various subjects and monitor results. It was shockingly capable and accurate. Again I iterate this was available ~6-7 years ago. The algorithms of social media now are heavily entwined by this technology, no doubt about it, which is why I'm so concerned. People are being subconsciously led by this technology already.

This is a rabbit hole that goes very deep, but it is for the most part readily available information. I believe the company goes by the name of Intuitive Code, is based in portugal. This is bordering on illuminati like stuff.
haha 🤣

well all that I will say is that a lot of my work, and my companies work, went into the early stage development of this technology and many other algorithms used by big tech.

I am not entirely proud of it, and strongly disagree how it is being used by many government and corporate entities.

These kind of scumbags even had a youtube channel back in 2016 with hundreds of videos bragging about market manipulation and crashing markets.



the SEC and gov did nothing, can do nothing. These guys have so much money, influence and power thanks to this technology.

What if I told you that this guy and his algorithm was solely responsible for the entire AMC stock fiasco, and that the story that was given on the news was entirely bullshit? this has been in plain sight for a long time.



TLDR, these guys were instructed to 'create' the frenzy in AMC stock and announced the plans to do so months before it happened. It was brazen and proves how powerful and dangerous this technology is. The behavioural models are so strong and accurate. Imagine how many financial lives were wrecked by this.

they were also heavily involved with the rise of crypto. Able to use the models to predict well in advance of moves, regulatory changes and everything.

its scary and gross. But it is reality. These videos are in full view for 7+ years


Mm. Maybe try running this con at the local pub instead. You'll have better luck convincing your audience.
 

Cyberpunkd

Member
This tech is here to stay. I'm getting great help from it in my studies. Not asking anything too advanced, but it has given clear and concise answers that's been accurate so far. It's only as good as the data though.
This is what I realized. In the past in school you couldn't just give an answer, but also provide an explanation. ChatGPT does that all for you e.g. with a math problem.
 

GymWolf

Member
Will? Many have already been using ChatGPT for their schoolwork. Kind of a no-brainer really.
If i was a professor i would just use more oral tests to see if they are the shit without chatgpt.

When i copied the shit out of math homeworks from the class nerds the prof always fucked us up by simply telling us to do the same on the board, easy way to catch people who actually studied from people who copied.

Are professors so stupid nowadays to not even test this shit out?
 

Tams

Member
If i was a professor i would just use more oral tests to see if they are the shit without chatgpt.

When i copied the shit out of math homeworks from the class nerds the prof always fucked us up by simply telling us to do the same on the board, easy way to catch people who actually studied from people who copied.

Are professors so stupid nowadays to not even test this shit out?

Some countries (Nordic ones in particular) scrapped homework. And yet they do very well in education league tables.

So really, most homework has always been a waste of time.
 
Last edited:

Faust

Perpetually Tired
When I was building early stage AGI in 2014 the stuff we see now in chat gpt was already in beta. That was ten fucking years ago.

They have had almost a decade of time ahead of the public release. I am pretty convinced at this point that China is still way further ahead with this stuff than the USA, because almost all of the developers back then were from hong kong and or shanghai. The chinese government saw the potential and dangers a long time ago and very rapidly put a stop to any sharing of information. Every single non-US or non-english native developer working on the company suddenly returned home on the same weekend. Were talking 30+ people just upping and leaving over night. No idea how they even got the travel arranged, they just were there one day and gone the next, with all of their laptops.

A few months after that the US government got involved and we were forced to sell our shares privately to an unknown investor. Never heard a thing about it again until Chat gpt came out.

This technology was revolutionary 10 years ago, and was being used to find new drugs 10 years ago as wel, and from what I hear (and this is the scary part) was also being used to help create biological weapons and viruses. It would not be a stretch to me in the past few years if this had a hand in either the creation of the covid virus or the vaccine itself. It really wouldnt, but I don't care to speculate too much.

Frustrated Buzz Lightyear GIF
 

Dacvak

No one shall be brought before our LORD David Bowie without the true and secret knowledge of the Photoshop. For in that time, so shall He appear.
This is so fuckin cool. I use ChatGPT almost daily, especially for recipes and cooking tips. I also use it when I have just random one-off questions that can be obscure to Google.

I’m fully in on our new AI overlords. I’ve been romancing Siri since 2012 for when she inevitably takes over the world. (Don’t tell my gf.)

Y’all just wait. While you guys are comatose and stuck in Matrix-tubes, I’ll at least have a Princess Leia chain around my neck, allowed to dance freely for the entertainment of our robotic masters.
 
Last edited:

jason10mm

Gold Member
I believe human and social sciences would be the ones with more, well, human input needed, right?

Psychology, for example, will be one of the safest jobs to have in an AI future, I think.
ABSOLUTELY NOT!!!

Go read a psychology paper today. You will find NOTHING more filled with meaningless word salad gobbledygook. Psychology is gonna be CRUSHED by AI written navel gazing nonsense, much like the art critic world and wine tasting notes, because all of that shit is JUNK!

What AI is gonna be great at is real-time fact checking political debates. At least until we cultivate fascist and communist chat-bots that then go to war with capitalist-bot.
 
Truely amazing to see.

The initial draft of the movie Interstellar got it rightv. This in android form will be our first interstellar ambassadors.
 

IDKFA

I am Become Bilbo Baggins
AI combined with robotics will mean we're going to end up like the humans in Wall-E if we're not careful.

They'll be no need for humans to be philosophers, builders, artists, scientists etc. We'll just be a species of consumers and and our creation become the new master.
 

violence

Member
AI combined with robotics will mean we're going to end up like the humans in Wall-E if we're not careful.

They'll be no need for humans to be philosophers, builders, artists, scientists etc. We'll just be a species of consumers and and our creation become the new master.
Humans will continue to evolve. I imagine we will form into something like this.


DiLm1uV.png
 

StreetsofBeige

Gold Member
AI combined with robotics will mean we're going to end up like the humans in Wall-E if we're not careful.

They'll be no need for humans to be philosophers, builders, artists, scientists etc. We'll just be a species of consumers and and our creation become the new master.
Maybe.

But if the content is good that’s what counts. And as a bonus it’ll be fast and dirt cheap to boot.

It’s like googling a shopbot. You can sit there all day checking websites one by one for the best price or mortgage rates or go to one of those sites that shows you the top 20 best choices out there. Doesn’t mean you have to choose one of them. But the system probably did 90% of the work for you already.
 

poppabk

Cheeks Spread for Digital Only Future
Mm. Maybe try running this con at the local pub instead. You'll have better luck convincing your audience.
Banned? He was just about to drop some AI powered stock tips on us and reveal that Biden died a decade ago and the current President is an AI construct.
So much valuable info will now be lost.
 
If i was a professor i would just use more oral tests to see if they are the shit without chatgpt.

When i copied the shit out of math homeworks from the class nerds the prof always fucked us up by simply telling us to do the same on the board, easy way to catch people who actually studied from people who copied.

Are professors so stupid nowadays to not even test this shit out?
Did you forget that poker players and chess players have ways of receiving realtime instructions and cheat live through things attached to their bodies feeding them the answers.
What AI is gonna be great at is real-time fact checking political debates. At least until we cultivate fascist and communist chat-bots that then go to war with capitalist-bot.

I do fact checking of the fact checkers, usually they tend to be right, but they aren't always, and when they fail its usually in favor of one side of the political aisle.
 
Top Bottom