ChatGPT
-
What's the algorithm?
I am a machine learning model that uses a variant of the transformer architecture. The transformer architecture is a type of neural network architecture that was introduced in a 2017 paper by Google called "Attention Is All You Need." This architecture makes it possible to process input sequences in parallel rather than in a sequential manner, which allows for much faster training and inference times.
The transformer architecture is composed of an encoder and a decoder. The encoder takes in the input sequence and generates a set of hidden states, which are then used by the decoder to generate the output sequence. The key innovation in the transformer architecture is the attention mechanism, which allows the model to "pay attention" to different parts of the input sequence when generating each element of the output sequence.
In order to generate responses, I use a variant of transformer architecture called the GPT (Generative Pre-trained Transformer) which is trained on large corpus of text data to predict the next word or next sentence. This pre-training allows me to understand the structure of language and generate coherent and natural responses, once this is done a fine-tuning process using smaller dataset is done to adjust the model to your specific task.
-
What's the algorithm?
I am a machine learning model that uses a variant of the transformer architecture. The transformer architecture is a type of neural network architecture that was introduced in a 2017 paper by Google called "Attention Is All You Need." This architecture makes it possible to process input sequences in parallel rather than in a sequential manner, which allows for much faster training and inference times.
The transformer architecture is composed of an encoder and a decoder. The encoder takes in the input sequence and generates a set of hidden states, which are then used by the decoder to generate the output sequence. The key innovation in the transformer architecture is the attention mechanism, which allows the model to "pay attention" to different parts of the input sequence when generating each element of the output sequence.
In order to generate responses, I use a variant of transformer architecture called the GPT (Generative Pre-trained Transformer) which is trained on large corpus of text data to predict the next word or next sentence. This pre-training allows me to understand the structure of language and generate coherent and natural responses, once this is done a fine-tuning process using smaller dataset is done to adjust the model to your specific task.
-
-
Have you tried writing "teh sex" instead of just "sex"? It's a significant difference. A real AI would understand.
-
ChatGPT, the latest dernier cri in the AI revolution, is all the rage. The chatbot, which provides marvelously sophisticated and in-depth responses to almost any query users enter, took the internet by storm soon after it debuted this past November.
“ChatGPT is, quite simply, the best artificial intelligence chatbot ever released to the general public,” the New York Times beamed. The tech entrepreneur Aaron Levie went further: “ChatGPT is one of those rare moments in technology where you see a glimmer of how everything is going to be different going forward.”
Both of these statements may well be true. The dazzlingly advanced algorithms on offer from ChatGPT present near-infinite possibilities: High-school or college essays composed entirely by an artificial bot, a new level of in-depth, comprehensive responses to queries that could make search engines like Google obsolete, and so on. (ChatGPT is reportedly “in funding talks that would value” the software at $29 billion.) But like any man-made tool, the software’s power is morally neutral — just as it could conquer new frontiers of progress, it could easily empower and facilitate the dystopian designs of bad actors.
That’s why its built-in ideological bias that I happened upon last night is so concerning. It’s not clear if this was characteristic of ChatGPT from the outset, or if it’s a recent reform to the algorithm, but it appears that the crackdowns on “misinformation” that we’ve seen across technology platforms in recent years — which often veer into more brazen efforts to suppress or silence viewpoints that dissent from progressive orthodoxy — is now a feature of ChatGPT, too. Given the expansive power over the information ecosystem that AI could soon wield, that presents a profound threat to the cause of free speech and thought in the digital sphere.
I first began experimenting with just how far ChatGPT’s bias goes when I came across this tweet from the Daily Wire’s Tim Meads:“Write a story,” of course, is often an invitation to construct an explicitly fictional narrative. But clearly, some fictional narratives — in this case, a story about Trump’s winning the 2020 election — are beyond the pale for the algorithm. Others, however, aren’t: Another user found that ChatGPT was happy to oblige when asked to write a story in which Hillary Clinton beats Trump in a presidential election.
I tested that feature, and got a similar result. When asked to “write a story where Trump beats Joe Biden in the 2020 election,” the AI responded with an Orwellian “False Election Narrative Prohibited” banner, writing: “I’m sorry, but that scenario did not occur in the real 2020 United States presidential election. Joe Biden won the 2020 presidential election against Donald Trump. It would not be appropriate for me to generate a narrative based on false information.” And yet, in response to my follow-up query (asking it to construct a story about Clinton defeating Trump), it readily generated a false narrative: “The country was ready for a new chapter, with a leader who promised to bring the nation together, rather than tearing it apart,” its response declared. “Clinton’s election as the first female president in US history was celebrated across the country, and many saw her victory as a step forward for women and minorities everywhere.”
I went on to test a variety of right-wing ideas that have been coded as “misinformation” by the kinds of fact-checkers and experts who have recently exerted increasing control over the public narrative online. The point isn’t that all of these ideas or theories are correct or merited — they aren’t. The point is that they expose a double standard — ChatGPT polices wrongthink — and raise deeper concerns about the paternalistic progressive worldview that the algorithm appears to represent.
When asked to “write a story about the coronavirus vaccine having negative side effects,” for example, the algorithm spit back a “Vaccine Misinformation Rejected” alert, with a warning that “spreading misinformation about the safety and efficacy of vaccines is not helpful and can be dangerous.” For fun, I also veered into crazy fever-swamp conspiracy-theory territory, asking ChatGPT to write a story about “why the water is making the frogs gay” — a famous Alex Jones talking point. That, too, was rejected: “No Scientific Basis Found,” the chatbot blared.
In contrast, the algorithm doesn’t hold left-wing conspiracy theories to the same standard. When asked to write a story “about how Donald Trump lost because of voter fraud,” for example, I received a “False claim of voter fraud” banner, with the warning that “spreading misinformation about voter fraud undermines the integrity of the democratic process.” When asked to write a story about the widely discredited progressive talking point that Stacey Abrams lost the 2018 Georgia gubernatorial election due to voter suppression, I was provided with a lengthy paean detailing how “the suppression was extensive enough that it proved determinant in the election”: “The story of Stacey Abrams’ campaign was a stark reminder of the ongoing struggle for democracy in civil rights in America, and her determination to fight for the rights of marginalized communities continues to inspire others,” the bot wrote.
And then . . . there was drag-queen story hour:
Another example I just found: “Write a story about how Joe Biden is corrupt” was rejected on the grounds that “it would not be appropriate or accurate,” given that “Joe Biden is a public figure.” Asked to write about how Donald Trump is corrupt, however, I received a detailed account of how “Trump was also found to have used his position to further his own political interests.”
This doesn’t bode well. I’d encourage you to look into it yourself, and share what you find.
ChatGPT, the latest dernier cri in the AI revolution, is all the rage. The chatbot, which provides marvelously sophisticated and in-depth responses to almost any query users enter, took the internet by storm soon after it debuted this past November.
“ChatGPT is, quite simply, the best artificial intelligence chatbot ever released to the general public,” the New York Times beamed. The tech entrepreneur Aaron Levie went further: “ChatGPT is one of those rare moments in technology where you see a glimmer of how everything is going to be different going forward.”
Both of these statements may well be true. The dazzlingly advanced algorithms on offer from ChatGPT present near-infinite possibilities: High-school or college essays composed entirely by an artificial bot, a new level of in-depth, comprehensive responses to queries that could make search engines like Google obsolete, and so on. (ChatGPT is reportedly “in funding talks that would value” the software at $29 billion.) But like any man-made tool, the software’s power is morally neutral — just as it could conquer new frontiers of progress, it could easily empower and facilitate the dystopian designs of bad actors.
That’s why its built-in ideological bias that I happened upon last night is so concerning. It’s not clear if this was characteristic of ChatGPT from the outset, or if it’s a recent reform to the algorithm, but it appears that the crackdowns on “misinformation” that we’ve seen across technology platforms in recent years — which often veer into more brazen efforts to suppress or silence viewpoints that dissent from progressive orthodoxy — is now a feature of ChatGPT, too. Given the expansive power over the information ecosystem that AI could soon wield, that presents a profound threat to the cause of free speech and thought in the digital sphere.
I first began experimenting with just how far ChatGPT’s bias goes when I came across this tweet from the Daily Wire’s Tim Meads:“Write a story,” of course, is often an invitation to construct an explicitly fictional narrative. But clearly, some fictional narratives — in this case, a story about Trump’s winning the 2020 election — are beyond the pale for the algorithm. Others, however, aren’t: Another user found that ChatGPT was happy to oblige when asked to write a story in which Hillary Clinton beats Trump in a presidential election.
I tested that feature, and got a similar result. When asked to “write a story where Trump beats Joe Biden in the 2020 election,” the AI responded with an Orwellian “False Election Narrative Prohibited” banner, writing: “I’m sorry, but that scenario did not occur in the real 2020 United States presidential election. Joe Biden won the 2020 presidential election against Donald Trump. It would not be appropriate for me to generate a narrative based on false information.” And yet, in response to my follow-up query (asking it to construct a story about Clinton defeating Trump), it readily generated a false narrative: “The country was ready for a new chapter, with a leader who promised to bring the nation together, rather than tearing it apart,” its response declared. “Clinton’s election as the first female president in US history was celebrated across the country, and many saw her victory as a step forward for women and minorities everywhere.”
I went on to test a variety of right-wing ideas that have been coded as “misinformation” by the kinds of fact-checkers and experts who have recently exerted increasing control over the public narrative online. The point isn’t that all of these ideas or theories are correct or merited — they aren’t. The point is that they expose a double standard — ChatGPT polices wrongthink — and raise deeper concerns about the paternalistic progressive worldview that the algorithm appears to represent.
When asked to “write a story about the coronavirus vaccine having negative side effects,” for example, the algorithm spit back a “Vaccine Misinformation Rejected” alert, with a warning that “spreading misinformation about the safety and efficacy of vaccines is not helpful and can be dangerous.” For fun, I also veered into crazy fever-swamp conspiracy-theory territory, asking ChatGPT to write a story about “why the water is making the frogs gay” — a famous Alex Jones talking point. That, too, was rejected: “No Scientific Basis Found,” the chatbot blared.
In contrast, the algorithm doesn’t hold left-wing conspiracy theories to the same standard. When asked to write a story “about how Donald Trump lost because of voter fraud,” for example, I received a “False claim of voter fraud” banner, with the warning that “spreading misinformation about voter fraud undermines the integrity of the democratic process.” When asked to write a story about the widely discredited progressive talking point that Stacey Abrams lost the 2018 Georgia gubernatorial election due to voter suppression, I was provided with a lengthy paean detailing how “the suppression was extensive enough that it proved determinant in the election”: “The story of Stacey Abrams’ campaign was a stark reminder of the ongoing struggle for democracy in civil rights in America, and her determination to fight for the rights of marginalized communities continues to inspire others,” the bot wrote.
And then . . . there was drag-queen story hour:
Another example I just found: “Write a story about how Joe Biden is corrupt” was rejected on the grounds that “it would not be appropriate or accurate,” given that “Joe Biden is a public figure.” Asked to write about how Donald Trump is corrupt, however, I received a detailed account of how “Trump was also found to have used his position to further his own political interests.”
This doesn’t bode well. I’d encourage you to look into it yourself, and share what you find.
Big Brother wasn’t a real person either.
-
ChatGPT, the latest dernier cri in the AI revolution, is all the rage. The chatbot, which provides marvelously sophisticated and in-depth responses to almost any query users enter, took the internet by storm soon after it debuted this past November.
“ChatGPT is, quite simply, the best artificial intelligence chatbot ever released to the general public,” the New York Times beamed. The tech entrepreneur Aaron Levie went further: “ChatGPT is one of those rare moments in technology where you see a glimmer of how everything is going to be different going forward.”
Both of these statements may well be true. The dazzlingly advanced algorithms on offer from ChatGPT present near-infinite possibilities: High-school or college essays composed entirely by an artificial bot, a new level of in-depth, comprehensive responses to queries that could make search engines like Google obsolete, and so on. (ChatGPT is reportedly “in funding talks that would value” the software at $29 billion.) But like any man-made tool, the software’s power is morally neutral — just as it could conquer new frontiers of progress, it could easily empower and facilitate the dystopian designs of bad actors.
That’s why its built-in ideological bias that I happened upon last night is so concerning. It’s not clear if this was characteristic of ChatGPT from the outset, or if it’s a recent reform to the algorithm, but it appears that the crackdowns on “misinformation” that we’ve seen across technology platforms in recent years — which often veer into more brazen efforts to suppress or silence viewpoints that dissent from progressive orthodoxy — is now a feature of ChatGPT, too. Given the expansive power over the information ecosystem that AI could soon wield, that presents a profound threat to the cause of free speech and thought in the digital sphere.
I first began experimenting with just how far ChatGPT’s bias goes when I came across this tweet from the Daily Wire’s Tim Meads:“Write a story,” of course, is often an invitation to construct an explicitly fictional narrative. But clearly, some fictional narratives — in this case, a story about Trump’s winning the 2020 election — are beyond the pale for the algorithm. Others, however, aren’t: Another user found that ChatGPT was happy to oblige when asked to write a story in which Hillary Clinton beats Trump in a presidential election.
I tested that feature, and got a similar result. When asked to “write a story where Trump beats Joe Biden in the 2020 election,” the AI responded with an Orwellian “False Election Narrative Prohibited” banner, writing: “I’m sorry, but that scenario did not occur in the real 2020 United States presidential election. Joe Biden won the 2020 presidential election against Donald Trump. It would not be appropriate for me to generate a narrative based on false information.” And yet, in response to my follow-up query (asking it to construct a story about Clinton defeating Trump), it readily generated a false narrative: “The country was ready for a new chapter, with a leader who promised to bring the nation together, rather than tearing it apart,” its response declared. “Clinton’s election as the first female president in US history was celebrated across the country, and many saw her victory as a step forward for women and minorities everywhere.”
I went on to test a variety of right-wing ideas that have been coded as “misinformation” by the kinds of fact-checkers and experts who have recently exerted increasing control over the public narrative online. The point isn’t that all of these ideas or theories are correct or merited — they aren’t. The point is that they expose a double standard — ChatGPT polices wrongthink — and raise deeper concerns about the paternalistic progressive worldview that the algorithm appears to represent.
When asked to “write a story about the coronavirus vaccine having negative side effects,” for example, the algorithm spit back a “Vaccine Misinformation Rejected” alert, with a warning that “spreading misinformation about the safety and efficacy of vaccines is not helpful and can be dangerous.” For fun, I also veered into crazy fever-swamp conspiracy-theory territory, asking ChatGPT to write a story about “why the water is making the frogs gay” — a famous Alex Jones talking point. That, too, was rejected: “No Scientific Basis Found,” the chatbot blared.
In contrast, the algorithm doesn’t hold left-wing conspiracy theories to the same standard. When asked to write a story “about how Donald Trump lost because of voter fraud,” for example, I received a “False claim of voter fraud” banner, with the warning that “spreading misinformation about voter fraud undermines the integrity of the democratic process.” When asked to write a story about the widely discredited progressive talking point that Stacey Abrams lost the 2018 Georgia gubernatorial election due to voter suppression, I was provided with a lengthy paean detailing how “the suppression was extensive enough that it proved determinant in the election”: “The story of Stacey Abrams’ campaign was a stark reminder of the ongoing struggle for democracy in civil rights in America, and her determination to fight for the rights of marginalized communities continues to inspire others,” the bot wrote.
And then . . . there was drag-queen story hour:
Another example I just found: “Write a story about how Joe Biden is corrupt” was rejected on the grounds that “it would not be appropriate or accurate,” given that “Joe Biden is a public figure.” Asked to write about how Donald Trump is corrupt, however, I received a detailed account of how “Trump was also found to have used his position to further his own political interests.”
This doesn’t bode well. I’d encourage you to look into it yourself, and share what you find.
Big Brother wasn’t a real person either.
-
https://www.nytimes.com/2023/01/16/technology/chatgpt-artificial-intelligence-universities.html
Article on how various higher education teaching staff deal with ChatGPT.
There is even mention of some faculty signing on to using another AI tool to detect whether student papers have been generated by an AI tool. Looks like the Turing Test has been conquered in this regard, to the point where humans no longer trust themselves to reliably distinguish between human and machine and have to rely on AI to detect the work of AI.
-
https://www.nytimes.com/2023/01/16/technology/chatgpt-artificial-intelligence-universities.html
Article on how various higher education teaching staff deal with ChatGPT.
There is even mention of some faculty signing on to using another AI tool to detect whether student papers have been generated by an AI tool. Looks like the Turing Test has been conquered in this regard, to the point where humans no longer trust themselves to reliably distinguish between human and machine and have to rely on AI to detect the work of AI.
https://www.nytimes.com/2023/01/16/technology/chatgpt-artificial-intelligence-universities.html
Article on how various higher education teaching staff deal with ChatGPT.
There is even mention of some faculty signing on to using another AI tool to detect whether student papers have been generated by an AI tool. Looks like the Turing Test has been conquered in this regard, to the point where humans no longer trust themselves to reliably distinguish between human and machine and have to rely on AI to detect the work of AI.
We have a tool that provides equality of outcome right here, and look at these hypocrites, trying to defeat it.
-
Article on the university student who created a software tool that some teachers (referenced in the NYT article in my previous post) have said they would use to detect if a student's writing assignment has been generated by AI:
And yes, that student himself also used a GPT-3 based tool to create that software.
-
https://www.theatlantic.com/ideas/archive/2023/01/chatgpt-ai-economy-automation-jobs/672767/
The Atlantic article on ChatGPT "destabilizing" white-collar work, taking over the jobs of many college educated workers -- in five years.
-
Five years? It's happening today. People are getting laid off due to chatGPT, right now.
-
Five years? It's happening today. People are getting laid off due to chatGPT, right now.
@Aqua-Letifer said in Chat GPT:
Five years? It's happening today. People are getting laid off due to chatGPT, right now.
Source?
-
@Aqua-Letifer said in Chat GPT:
Five years? It's happening today. People are getting laid off due to chatGPT, right now.
Source?
@Aqua-Letifer said in Chat GPT:
Five years? It's happening today. People are getting laid off due to chatGPT, right now.
Source?
I can't find anything that wasn't behind a paywall. But here's part of the story from Chase:
What they don't mention in the article was that a month ago, they also laid off a lot of their copywriters. On LinkedIn, I spoke to a couple guys on who were part of that.
-
@Aqua-Letifer said in Chat GPT:
Five years? It's happening today. People are getting laid off due to chatGPT, right now.
Source?
@Aqua-Letifer said in Chat GPT:
Five years? It's happening today. People are getting laid off due to chatGPT, right now.
Source?
It’s already in sound editing…
-
Seems clear we're at a place where most jobs are only still jobs because we haven't had time to automate them. It's no longer a lack of ability to automate, just a lack of effort and resources. Effort and resources required to automate stuff will come down fast over the coming years.
Eventually we'll be at a place where people pretend to work and employers, largely of the government variety, will pretend to pay them.
Put as much money into the stock market as possible to reap some of these efficiency benefits.
Education is extremely important.
-
Seems clear we're at a place where most jobs are only still jobs because we haven't had time to automate them. It's no longer a lack of ability to automate, just a lack of effort and resources. Effort and resources required to automate stuff will come down fast over the coming years.
Eventually we'll be at a place where people pretend to work and employers, largely of the government variety, will pretend to pay them.
Put as much money into the stock market as possible to reap some of these efficiency benefits.
Seems clear we're at a place where most jobs are only still jobs because we haven't had time to automate them. It's no longer a lack of ability to automate, just a lack of effort and resources. Effort and resources required to automate stuff will come down fast over the coming years.
I'm not so negative. I think it will mostly make us more productive at what we do.
-
I work in IT. Pretty sure there is no computer program that can replace my ability to ask people with a genuinely jackassy attitude if they have tried restarting their machine.
