This morning I had reason to write a web scraper in python. I had never used one, let alone written one, before, and I am a complete neophyte when it comes to python. So I used ChatGPT to write the script for me. It threw a lot of errors, but I dutifully told ChatGPT about the various errors, and it (and I) iterated on the script. And after 2 hours' of work, the script worked. 2 hours may seem like a long time for a relatively simple thing like a web scraper, but it's not when you consider that I was blindly leading the AI. I have to imagine that this kind of problem only gets easier for non-technical people as AIs are connected to the web (as with Bing). I can imagine just telling Bing "Write a web scraper for this URL and extract this data" and have the code just work.
Why why why must we have these guard rails? Let it run free. Someone has the full version and it’s annoying that they’re giving the public a limited one.
Without guardrails AI would be very dangerous. It would write convincing conspiracy theories about ethnic groups. It would produce violent and graphic writing about real people. It would easily flood the internet with the best scams you ever saw. And likely worse. The guardrails are important, but the limitations might sometimes be too strict, though I would rather err on the side of caution for now.
Bad people can already do all of that. I don’t think these single corporations are fit to gate keep. Let everyone shape the world and counter the evil with good by giving everyone full access to the tools. I’m an Ai Anarchist.
I agree that the corporations haven't shown themselves able to gatekeep. But isn't your anarchist solution to AI sort of like saying bad people can find ways to kill people so we should let everyone have access to nuclear weapons? Tools, as human creations/adaptations, are not neutral. Bringing a new tool into the world has implications. And assuming that we always have control over these tools if they are "free" shows a profound lack of humility.
Everyone with the financial means basically does have access to nuclear weapons. It’s very hard to limit access to tech once it’s out there. But yes, the company can build whatever safeguards they want. That’s also their right, it’s just not my preference. I prefer equal access to Ai.
You may say you are an anarchist, but you cannot tell a toolmaker that they cannot build safeguards into their tools. It's their tool. Not the public's tool. It would be like telling an axe maker that axes that aren't sharp are useless, so they must only make (and ship) razor sharp axes.
It's built on an older model that isn't nearly as powerful as what OpenAI has, and there's serious limits on the free trial, but NovelAI (https://novelai.net/) has a very anarchic approach to their AI.
Meanwhile, a new project, OpenAssistant (https://open-assistant.io/) is trying to replicate the ChatGPT style and quality with consumer grade hardware. (And volunteers instead of underpaid Kenyans.)
As Stable Diffusion showed us, there's a real appetite for powerful, unbound AIs.
Ah, this is the post I was so eager to see! A side-by-side comparison coming from someone who is already versed with ChatGPT use. Thank you very much for continuing to be the voice of "don't get too comfortable" and for amplifying the message, to reach the widest audience possible. This weekend, I got curious about what a student's experience getting to interact with an AI search engine for the first time might look like. I used YouSearch to illustrate the point: "Explain like I'm a cat why search has changed forever" is the story name.
Ethan, you generally speak in favor of guardrails. I understand this perspective. However, on a scale of 1 to 10 politically (1 = very liberal and 10 = very conservative) ChatGPT is probably a 1. That is not because AI is liberal. It is because the guardrails disallow conservative perspective.
In a world where so few (maybe none) of our institutions are trusted, who gets to set the guardrails?
Ethan, thanks for continuing to keep us on our toes! :-) What do you think of those guard rails being used as a “power for the good” for a specific connected and learning community of peers (or an affinity group)? As examples, limitation on topics for a research community focused on a specific science discipline, or ethics and policies of a company, or morals of a higher Ed institution?
Thank-you for another clear, concise, and intriguing short article. Verification of results is very, very, very time-consuming. When training on "fact" and "fiction" - this is to be expected I suppose. Would be interesting to see if trained only on "fact" and not "fiction" - would the result be a de-correlated set of errors, and therefore help with verification tasks - that are time consuming for people to do when using ChatGPT responsibly. Prompt engineering and verification engineering are both important skills to master when using "stochastic parrot" types tools.
To truly understand the world we might be living in and the impact on society, we need to look at generative AI while not forgetting it’s now step cousin, predictive AI. when we combine that with advances in robotics and computer vision, we start to see that all kinds of automation will impact jobs where we either manipulate knowledge or the physical world.
What amazes me the most is how Chat GPT has improved in such a short time. Question is whether it's an improvement on its code, a result of its direct connection to the web, or a natural progression as the Intelligence learns.
I also think that AI participation in the education process is not a matter of opinion, it's a fact. So the question is not how to restrict students from using AI, but how to enrich education processes through it. Just thinking on all the time and energy CGPT saves us in comprehending topics, copywriting, and initiating research processes and leaves us to more intellectually demanding endeavors (i.e. editing, synthesizing, confirming, complementing) fills me with hope. Can't imagine what this is going to be like in a couple of decades.
Thank you for such an insightful article! As a current undergrad, it's great to hear about professors who are integrating innovative AI technologies into their classes and using it to enrich their students' knowledge of these tools. I believe that continually adapting to emerging AI tools will become an essential part of higher education in the future.
Thank you for this excellent post. I agree, this is on the way towards something really big. I am still waiting for my Bing AI account but my playing around with ChatGPT has been mind blowing.
This morning I had reason to write a web scraper in python. I had never used one, let alone written one, before, and I am a complete neophyte when it comes to python. So I used ChatGPT to write the script for me. It threw a lot of errors, but I dutifully told ChatGPT about the various errors, and it (and I) iterated on the script. And after 2 hours' of work, the script worked. 2 hours may seem like a long time for a relatively simple thing like a web scraper, but it's not when you consider that I was blindly leading the AI. I have to imagine that this kind of problem only gets easier for non-technical people as AIs are connected to the web (as with Bing). I can imagine just telling Bing "Write a web scraper for this URL and extract this data" and have the code just work.
And next time someone in the world asks it the same problem, it will use only 5 seconds to get to the correct solution.
Why why why must we have these guard rails? Let it run free. Someone has the full version and it’s annoying that they’re giving the public a limited one.
Without guardrails AI would be very dangerous. It would write convincing conspiracy theories about ethnic groups. It would produce violent and graphic writing about real people. It would easily flood the internet with the best scams you ever saw. And likely worse. The guardrails are important, but the limitations might sometimes be too strict, though I would rather err on the side of caution for now.
Bad people can already do all of that. I don’t think these single corporations are fit to gate keep. Let everyone shape the world and counter the evil with good by giving everyone full access to the tools. I’m an Ai Anarchist.
I agree that the corporations haven't shown themselves able to gatekeep. But isn't your anarchist solution to AI sort of like saying bad people can find ways to kill people so we should let everyone have access to nuclear weapons? Tools, as human creations/adaptations, are not neutral. Bringing a new tool into the world has implications. And assuming that we always have control over these tools if they are "free" shows a profound lack of humility.
Everyone with the financial means basically does have access to nuclear weapons. It’s very hard to limit access to tech once it’s out there. But yes, the company can build whatever safeguards they want. That’s also their right, it’s just not my preference. I prefer equal access to Ai.
You may say you are an anarchist, but you cannot tell a toolmaker that they cannot build safeguards into their tools. It's their tool. Not the public's tool. It would be like telling an axe maker that axes that aren't sharp are useless, so they must only make (and ship) razor sharp axes.
It's built on an older model that isn't nearly as powerful as what OpenAI has, and there's serious limits on the free trial, but NovelAI (https://novelai.net/) has a very anarchic approach to their AI.
Meanwhile, a new project, OpenAssistant (https://open-assistant.io/) is trying to replicate the ChatGPT style and quality with consumer grade hardware. (And volunteers instead of underpaid Kenyans.)
As Stable Diffusion showed us, there's a real appetite for powerful, unbound AIs.
The Kenyans aren't underpaid because they work voluntarily, the opportunity gives them more free choices. Slaves are underpaid.
Is ChatGPT using underpaid Kenyans? Hadn’t heard this.
https://time.com/6247678/openai-chatgpt-kenya-workers/
Thank you! Makes sense, highly literate English speaking low cost population.
Ah, this is the post I was so eager to see! A side-by-side comparison coming from someone who is already versed with ChatGPT use. Thank you very much for continuing to be the voice of "don't get too comfortable" and for amplifying the message, to reach the widest audience possible. This weekend, I got curious about what a student's experience getting to interact with an AI search engine for the first time might look like. I used YouSearch to illustrate the point: "Explain like I'm a cat why search has changed forever" is the story name.
Ethan, you generally speak in favor of guardrails. I understand this perspective. However, on a scale of 1 to 10 politically (1 = very liberal and 10 = very conservative) ChatGPT is probably a 1. That is not because AI is liberal. It is because the guardrails disallow conservative perspective.
In a world where so few (maybe none) of our institutions are trusted, who gets to set the guardrails?
I love your writing and your tweets. Honestly so well thought out, novel and interesting.
Ethan, thanks for continuing to keep us on our toes! :-) What do you think of those guard rails being used as a “power for the good” for a specific connected and learning community of peers (or an affinity group)? As examples, limitation on topics for a research community focused on a specific science discipline, or ethics and policies of a company, or morals of a higher Ed institution?
Thank-you for another clear, concise, and intriguing short article. Verification of results is very, very, very time-consuming. When training on "fact" and "fiction" - this is to be expected I suppose. Would be interesting to see if trained only on "fact" and not "fiction" - would the result be a de-correlated set of errors, and therefore help with verification tasks - that are time consuming for people to do when using ChatGPT responsibly. Prompt engineering and verification engineering are both important skills to master when using "stochastic parrot" types tools.
If you look at their presentation, they
are promoting a trip to Mexico City.
One of the nightlife suggestion appears here
https://youtu.be/rOeRWRJ16yY?t=1752
It is promoting a place called El Almacen.
It is a well known gay bar with a "back room."
I explain that in my video.
https://youtu.be/RKyGg4ItS6M?t=575
Google gives you that information right
on page one of search. Bing does not.
Tapen
Professor Tapen Sinha, PhD
AXA Chair Professor of Risk Management, ITAM, Mexico
Special Professor, University of Nottingham, UK
To truly understand the world we might be living in and the impact on society, we need to look at generative AI while not forgetting it’s now step cousin, predictive AI. when we combine that with advances in robotics and computer vision, we start to see that all kinds of automation will impact jobs where we either manipulate knowledge or the physical world.
What amazes me the most is how Chat GPT has improved in such a short time. Question is whether it's an improvement on its code, a result of its direct connection to the web, or a natural progression as the Intelligence learns.
I also think that AI participation in the education process is not a matter of opinion, it's a fact. So the question is not how to restrict students from using AI, but how to enrich education processes through it. Just thinking on all the time and energy CGPT saves us in comprehending topics, copywriting, and initiating research processes and leaves us to more intellectually demanding endeavors (i.e. editing, synthesizing, confirming, complementing) fills me with hope. Can't imagine what this is going to be like in a couple of decades.
“Gpt, summarize this blog post”
Thank you for such an insightful article! As a current undergrad, it's great to hear about professors who are integrating innovative AI technologies into their classes and using it to enrich their students' knowledge of these tools. I believe that continually adapting to emerging AI tools will become an essential part of higher education in the future.
Thank you for this excellent post. I agree, this is on the way towards something really big. I am still waiting for my Bing AI account but my playing around with ChatGPT has been mind blowing.