3 ms·

Clickable links: https://github.com/svix/svix-webhooks/issues/453 https://github.com/svix/svix-webhooks/issues/453 https://github.com/svix/svix-webhooks/issue

Clickable links: https://github.com/svix/svix-webhooks/issues/453 https://github.com/svix/svix-webhooks/issues/453 https://github.com/svix/svix-webhooks/issues/714 https://github.com/svix/svix-webhooks/issues/714 How do you know the issues were generated with ChatGPT? It's not immediately obvious to me, but this is the first time I've crossed paths with svix. Edit: Looks like this was also discussed yesterday~ https://news.ycombinator.com/item?id=34150261 https://news.ycombinator.com/item?id=34150261

KenRuf 4y ago
There are a few potential clues that might indicate that a piece of text was generated by ChatGPT or a similar language model: Lack of context or coherence: Language models like ChatGPT are designed to generate text based on a given prompt or input, but they do not have a deep understanding of the context in which the text is being used. As a result, the text they generate may lack coherence or may not make sense in the context in which it is being used. Repetitive or unnatural language: Language models can sometimes produce repetitive or unnatural language patterns, particularly if they are not provided with sufficient context or training data. Inconsistencies or errors: Language models may produce errors or inconsistencies in the text they generate, especially if they are not trained on a diverse or comprehensive dataset. It is important to note that these clues are not definitive indicators that a piece of text was generated by ChatGPT or any other specific language model. There are many other factors that could contribute to these characteristics in text, and it is not always possible to determine with certainty whether a piece of text was generated by a specific model. FYI this is written by ChatGPT so you get an idea of its writing style. I asked it "what are some clues that might indicate something is written by chatgpt".
- yccs27 4y ago
  Thank you for the last sentence information. I definitely got that eerie "AI text" vibe, but as ChatGPT puts it, "it is not always possible to determine with certainty whether a piece of text was generated by [ChatGPT]".
- dandelany 4y ago
  A few more clues which are demonstrated, but not explained, by ^this response: * Begins with a slight rewording of your question, rephrased as the introduction to the answer * Ends with at least a sentence or two of CYA-type caveats, like "ultimately it will depend on", "there are many factors", etc. * Writes in full sentences even when not really necessary - eg. this bullet point would be phrased as "Text generated by a language model would be more likely than average to be written in full sentences..." * Avoids pronouns when any ambiguity is present, even if it would be obvious to a human. eg. If the first sentence is: "Robert and Joe were playing football when Joe injured his hand", a human might continue: "He tried to keep playing but it hurt too badly." ChatGPT would be more like: "Joe tried to keep playing but his hand hurt too badly". (this was NOT written by ChatGPT :)
  - muttled 4y ago
    The regularity that it builds answers in this form makes me think they must have examples of it embedded in the prompt
ffssffss 4y ago
GPT-2 output detector flags it bright red (which is interesting since this was presumably done with GPT-3): https://huggingface.co/openai-detector/ https://huggingface.co/openai-detector/ I have a feeling we're going to have to stick this kind of thing in every piece of input processing middleware over the next couple years.
- elashri 4y ago
  This is an interested tool, thanks for sharing. It was accurate, I tried various ChatGPT generated responses. it gets very good results with all of them. Except when I went in depth about scientific topics. I wrote before this paragarah > "Interestingly, despite neutrinos being otherwise ideal as a dark matter candidate, this limit told us they could not be dark matter. We know that neutrinos are much lighter than 1-3 keV. So you literally cannot cram enough neutrinos into a galaxy for them to be the dark matter. Amusingly, this Pauli Exclusion Limit for fermions is also what supports a white dwarf or neutron star from gravitational collapse. If the star got smaller, there wouldn’t be enough room for the electrons (dwarfs) or neutrons (neutron stars) without doubling up on quantum states. If dark matter is a boson, you can’t use the Tremaine-Gunn bound, since bosons are okay with all sitting in the same quantum state. So in principle, you can fit an infinite number inside a galaxy. You can use another property of quantum fields to constrain bosonic dark matter, though. If the particle is too light, the quantum wavefunction is so dispersed it cannot be localized inside the galaxy. This puts a lower bound of ~10^-22 eV on bosonic dark matter. Rather less than the bound for fermionic dark matter. But at least this is even a “thing.” Of course, there are other reasons to eliminate neutrinos as a dark matter candidate, but yet we know more about what couldn’t be a dark matter, not what is the dark matter" I asked ChatGPT to `rewrite this in a more professional way` and got the following: > Neutrinos have been considered as a potential candidate for dark matter due to their neutral charge and low mass. However, the Tremaine-Gunn bound, which sets a limit on the maximum mass of fermionic dark matter particles based on their contribution to the gravitational potential of a galaxy, has shown that neutrinos cannot be the dominant form of dark matter. This is because their mass is significantly lower than the 1-3 keV range set by the bound. Additionally, the Pauli Exclusion Principle, which states that no two fermions can occupy the same quantum state simultaneously, also prevents neutrinos from being concentrated in sufficient quantities to serve as dark matter. On the other hand, bosonic dark matter particles, which can occupy the same quantum state, may be able to satisfy the constraints set by the Tremaine-Gunn bound. However, there are other considerations, such as the lower bound of approximately 10^-22 eV on the mass of bosonic dark matter particles based on their ability to be localized within a galaxy, that must also be taken into account. Despite these limitations, it is important to continue exploring and evaluating the potential of various particles as dark matter candidates in order to better understand this mysterious and essential component of the universe. It gives that there is a 1.58% only red. seems like there is a lot of need to GPT3 detector.
nilram 4y ago
Creating some random sentences, this came up as 98% chance as fake: My name is Aram. I have a large piece of fish in my backpack. I want to go to the mountains. I have no goals in life. This is all I can write. When I changed "I want..." to "I quite want...", the slider moved from 98% to 21% chance fake.

KenRuf 4y ago

There are a few potential clues that might indicate that a piece of text was generated by ChatGPT or a similar language model: Lack of context or coherence: Language models like ChatGPT are designed to generate text based on a given prompt or input, but they do not have a deep understanding of the context in which the text is being used. As a result, the text they generate may lack coherence or may not make sense in the context in which it is being used. Repetitive or unnatural language: Language models can sometimes produce repetitive or unnatural language patterns, particularly if they are not provided with sufficient context or training data. Inconsistencies or errors: Language models may produce errors or inconsistencies in the text they generate, especially if they are not trained on a diverse or comprehensive dataset. It is important to note that these clues are not definitive indicators that a piece of text was generated by ChatGPT or any other specific language model. There are many other factors that could contribute to these characteristics in text, and it is not always possible to determine with certainty whether a piece of text was generated by a specific model. FYI this is written by ChatGPT so you get an idea of its writing style. I asked it "what are some clues that might indicate something is written by chatgpt".

yccs27 4y ago
Thank you for the last sentence information. I definitely got that eerie "AI text" vibe, but as ChatGPT puts it, "it is not always possible to determine with certainty whether a piece of text was generated by [ChatGPT]".
dandelany 4y ago
A few more clues which are demonstrated, but not explained, by ^this response: * Begins with a slight rewording of your question, rephrased as the introduction to the answer * Ends with at least a sentence or two of CYA-type caveats, like "ultimately it will depend on", "there are many factors", etc. * Writes in full sentences even when not really necessary - eg. this bullet point would be phrased as "Text generated by a language model would be more likely than average to be written in full sentences..." * Avoids pronouns when any ambiguity is present, even if it would be obvious to a human. eg. If the first sentence is: "Robert and Joe were playing football when Joe injured his hand", a human might continue: "He tried to keep playing but it hurt too badly." ChatGPT would be more like: "Joe tried to keep playing but his hand hurt too badly". (this was NOT written by ChatGPT :)
- muttled 4y ago
  The regularity that it builds answers in this form makes me think they must have examples of it embedded in the prompt

ffssffss 4y ago

GPT-2 output detector flags it bright red (which is interesting since this was presumably done with GPT-3): https://huggingface.co/openai-detector/ https://huggingface.co/openai-detector/ I have a feeling we're going to have to stick this kind of thing in every piece of input processing middleware over the next couple years.

elashri 4y ago
This is an interested tool, thanks for sharing. It was accurate, I tried various ChatGPT generated responses. it gets very good results with all of them. Except when I went in depth about scientific topics. I wrote before this paragarah > "Interestingly, despite neutrinos being otherwise ideal as a dark matter candidate, this limit told us they could not be dark matter. We know that neutrinos are much lighter than 1-3 keV. So you literally cannot cram enough neutrinos into a galaxy for them to be the dark matter. Amusingly, this Pauli Exclusion Limit for fermions is also what supports a white dwarf or neutron star from gravitational collapse. If the star got smaller, there wouldn’t be enough room for the electrons (dwarfs) or neutrons (neutron stars) without doubling up on quantum states. If dark matter is a boson, you can’t use the Tremaine-Gunn bound, since bosons are okay with all sitting in the same quantum state. So in principle, you can fit an infinite number inside a galaxy. You can use another property of quantum fields to constrain bosonic dark matter, though. If the particle is too light, the quantum wavefunction is so dispersed it cannot be localized inside the galaxy. This puts a lower bound of ~10^-22 eV on bosonic dark matter. Rather less than the bound for fermionic dark matter. But at least this is even a “thing.” Of course, there are other reasons to eliminate neutrinos as a dark matter candidate, but yet we know more about what couldn’t be a dark matter, not what is the dark matter" I asked ChatGPT to `rewrite this in a more professional way` and got the following: > Neutrinos have been considered as a potential candidate for dark matter due to their neutral charge and low mass. However, the Tremaine-Gunn bound, which sets a limit on the maximum mass of fermionic dark matter particles based on their contribution to the gravitational potential of a galaxy, has shown that neutrinos cannot be the dominant form of dark matter. This is because their mass is significantly lower than the 1-3 keV range set by the bound. Additionally, the Pauli Exclusion Principle, which states that no two fermions can occupy the same quantum state simultaneously, also prevents neutrinos from being concentrated in sufficient quantities to serve as dark matter. On the other hand, bosonic dark matter particles, which can occupy the same quantum state, may be able to satisfy the constraints set by the Tremaine-Gunn bound. However, there are other considerations, such as the lower bound of approximately 10^-22 eV on the mass of bosonic dark matter particles based on their ability to be localized within a galaxy, that must also be taken into account. Despite these limitations, it is important to continue exploring and evaluating the potential of various particles as dark matter candidates in order to better understand this mysterious and essential component of the universe. It gives that there is a 1.58% only red. seems like there is a lot of need to GPT3 detector.

nilram 4y ago

Creating some random sentences, this came up as 98% chance as fake: My name is Aram. I have a large piece of fish in my backpack. I want to go to the mountains. I have no goals in life. This is all I can write. When I changed "I want..." to "I quite want...", the slider moved from 98% to 21% chance fake.