We love your newsletters and we have a good news for you!
We're also running a beehiiv newsletter based on AI with 19K+ readers, and we are inviting writers to share their best knowledge on any topic that you love on our next issue, completely dedicated for you!
If it is tech/AI, our audience will be more than happy to read your work on our newsletter.
We can explore cross-promotion opportunities, if you're interested?
Or if you have any other proposal, we can proceed with it too.
Hi! I'd like to try Origin, but I'm hesitant to incorporate software with such a broad (and vague) permission request into my browser. I am a researcher who works with confidential data. What –exactly– does: "It can: Read and change all your data on all websites" mean?
GPTZero won't last, and neither will pretty much any AI detector. I think you have less than a year to hype it up until people realize that AI detectors can never be a perfect and a working tool. Schools will have to be the ones to adapt to these changes and completely change the education system.
I applaud you for being able to create a tool so quickly that some teachers could believe in it, and falsely accuse students of using AI.
Yes, AI detectors will still be a thing, and they'll usually be fine at finding AI-created stuff on the internet, like articles and SEO. However, AI detectors should never be implemented within the education system due to their inherent issues they will always have. Hopefully, this will provide the necessary impetus for our outdated education system to undergo much-needed changes.
I look forward to seeing what you guys do outside of AI detection in education, as we all know that will never work out and detectors will just keep creating many more problems if you continue. Fact-checking and content verification is great though, I’d love to see you working on that.
Hi Crystal. I respectfully disagree. Here's a theory paper that says detection will always be possible with more data (https://arxiv.org/pdf/2304.04736.pdf). The entire field of detection is so new, and most others detector are either playing catch-up to us, or repurposing old research papers not meant for practical applications. Detection is a new field that will need funding beyond academia to actually succeed. Coincidentally, we just raised that funding. Give us the chance to build the technology for real-world use.
I actually have read the research paper, and it does make interesting points but still shows the many flaws in AI detection.
They're suggesting that it should almost always be doable to spot text that's been churned out by an AI, unless the patterns in human-written and AI-written texts are totally identical. This idea actually comes from basic principles in the field of information theory. It's tied to the notion that if AI-generated text is becoming more like human writing, we're going to need more examples to spot the difference.
However, there's a catch to this method. It depends on having a large enough pool of examples to spot the differences between human and AI-produced text. The issue is, as AI gets better at writing and mimics human style more closely, spotting the difference could get trickier if we only have a few samples to look at. It will never be possible to always be getting important samples that are ever-changing. We are also completely forgetting that a lot of people write similarly to AI, and so there will always be false positives. Thats why I don’t think AI detectors should be a tool to “save the education system” as majority of teachers want simple and easy-to-use tool that tells them if something was written by AI or not, this creates huge problems that are already being shown, as these AI detectors will never be perfect, and this applies not just to GPTZero. There are too many false positives and false negatives, for this to be a widely used tool in academia. The answer to this will have to be the already outdated education system, changing their ways for the better by how they approach assignments, and teach. There are still many more inherent limitations and challenges, and this is something that future research will need to explore further.
The education system, in general, is a complex and powerful system with many issues already. If you want to advertise GPTZero to them, you must heavily acknowledge that, like many AI detectors, GPTZero should not be entirely trusted. We have already witnessed numerous instances of students being falsely accused by their schools using GPTZero and other AI detectors such as Turnitin. Remember majority of teachers desire a simple tool, they don't have time to examine everything. Thats why I still think that we need less focus on tools like AI detectors, and more focus on changing the education system as a whole.
They also do not discuss the potential logistical and ethical challenges of collecting more text samples, which is a significant oversight given the privacy concerns and data governance issues associated with large scale data collection. Especially with what we saw at the senate today with Sam Altman, it's going to be a lot harder! lol
Hey Edward, hope you are doing great!
I am Soumya from ByteBrief (bytebrief.co)
We love your newsletters and we have a good news for you!
We're also running a beehiiv newsletter based on AI with 19K+ readers, and we are inviting writers to share their best knowledge on any topic that you love on our next issue, completely dedicated for you!
If it is tech/AI, our audience will be more than happy to read your work on our newsletter.
We can explore cross-promotion opportunities, if you're interested?
Or if you have any other proposal, we can proceed with it too.
Contact us: hello@bytebrief.co
Thanks
Hi! I'd like to try Origin, but I'm hesitant to incorporate software with such a broad (and vague) permission request into my browser. I am a researcher who works with confidential data. What –exactly– does: "It can: Read and change all your data on all websites" mean?
GPTZero won't last, and neither will pretty much any AI detector. I think you have less than a year to hype it up until people realize that AI detectors can never be a perfect and a working tool. Schools will have to be the ones to adapt to these changes and completely change the education system.
I applaud you for being able to create a tool so quickly that some teachers could believe in it, and falsely accuse students of using AI.
Yes, AI detectors will still be a thing, and they'll usually be fine at finding AI-created stuff on the internet, like articles and SEO. However, AI detectors should never be implemented within the education system due to their inherent issues they will always have. Hopefully, this will provide the necessary impetus for our outdated education system to undergo much-needed changes.
I look forward to seeing what you guys do outside of AI detection in education, as we all know that will never work out and detectors will just keep creating many more problems if you continue. Fact-checking and content verification is great though, I’d love to see you working on that.
Hi Crystal. I respectfully disagree. Here's a theory paper that says detection will always be possible with more data (https://arxiv.org/pdf/2304.04736.pdf). The entire field of detection is so new, and most others detector are either playing catch-up to us, or repurposing old research papers not meant for practical applications. Detection is a new field that will need funding beyond academia to actually succeed. Coincidentally, we just raised that funding. Give us the chance to build the technology for real-world use.
I actually have read the research paper, and it does make interesting points but still shows the many flaws in AI detection.
They're suggesting that it should almost always be doable to spot text that's been churned out by an AI, unless the patterns in human-written and AI-written texts are totally identical. This idea actually comes from basic principles in the field of information theory. It's tied to the notion that if AI-generated text is becoming more like human writing, we're going to need more examples to spot the difference.
However, there's a catch to this method. It depends on having a large enough pool of examples to spot the differences between human and AI-produced text. The issue is, as AI gets better at writing and mimics human style more closely, spotting the difference could get trickier if we only have a few samples to look at. It will never be possible to always be getting important samples that are ever-changing. We are also completely forgetting that a lot of people write similarly to AI, and so there will always be false positives. Thats why I don’t think AI detectors should be a tool to “save the education system” as majority of teachers want simple and easy-to-use tool that tells them if something was written by AI or not, this creates huge problems that are already being shown, as these AI detectors will never be perfect, and this applies not just to GPTZero. There are too many false positives and false negatives, for this to be a widely used tool in academia. The answer to this will have to be the already outdated education system, changing their ways for the better by how they approach assignments, and teach. There are still many more inherent limitations and challenges, and this is something that future research will need to explore further.
The education system, in general, is a complex and powerful system with many issues already. If you want to advertise GPTZero to them, you must heavily acknowledge that, like many AI detectors, GPTZero should not be entirely trusted. We have already witnessed numerous instances of students being falsely accused by their schools using GPTZero and other AI detectors such as Turnitin. Remember majority of teachers desire a simple tool, they don't have time to examine everything. Thats why I still think that we need less focus on tools like AI detectors, and more focus on changing the education system as a whole.
They also do not discuss the potential logistical and ethical challenges of collecting more text samples, which is a significant oversight given the privacy concerns and data governance issues associated with large scale data collection. Especially with what we saw at the senate today with Sam Altman, it's going to be a lot harder! lol