Ethan, I've really come to enjoy your posts. I like your deep dives and perspective. Wondering if you are planning to look into how the guardrails or training is done with AI.
"This wasn’t supposed to be an AI-focused Substack." I feel ya on this one. I've gotten swept up in AI commentary and then work in the last few weeks, too. I don't know if the train will ever stop for some sightseeing, or if it's going to evolve around us to a bullet train and then a space ship while we're all just hanging out in the lounge asking Guinan for something blue to drink.
Ethan, witnessing the recent and thrilling advancements in AI and the public's reaction, I can't help but remind of Toffler's book "Future Shock," which predicted the rapid pace of technological change and its effects on society.
Love your posts. I’m a regular ole’ public school employee in a well regarded suburban school system and am already adopting ChatGPT for work and personal tasks. I enjoy reading what true experts are doing.
In 25 years time we’ve gone from working in the library to cut down on research time by being near the card catalog...then cutting days and weeks down to hours by researching on the internet, to cutting down weeks and hours into seconds with Chat GPT. It’s thrilling and one of the few moments in history there’s a tangible giant leap in technology in which so many regular people can capitalize. I remember the same feeling when I gave my senior project speech at University on “The Internet”.
Yes to all of this. Hardly anyone seems to appreciate how quickly AI is going to change the world. It's not just the technology itself, which as shown in that general intelligence paper is even more impressive than we thought. It's that all the complementary pieces, like a huge software industry, computing infrastructure, and installed device and user base, are already in place. The expansion of the web was gated on availability of personal computers, servers, and users. Same with mobile and smartphones. There is no real gating factor on this one as far as I can tell, apart from how quickly people can come up with useful ideas and (with AI help) code them. I would guess it's a trillion dollar industry at a minimum within five years, which implies X trillion in economic value created.
LLMs and generative art AIs are progressing far faster than anyone expected, and it's going to be a challenge not just to keep up, but also to understand the impacts they are going to have and put mitigations in place where needed. In fact, I think that latter point is impossible.
I've been thinking about the impact of LLMs in particular on the publishing world, and I do fear that there's a lot of disruption coming down the pike towards us, and we just don't have the tools or understanding to get ready for it. And it's not all going to be good disruption:
Another great post, Ethan, thank you. These advances are certainly as exciting as they are useful. Do we think though that there should be more trepidation and caution tempering this wave of excitement? Profit motives tend to push sensible conversations to the sidelines.
When each release builds on its predecessor in this way, it won’t be long before hallucinations happen so seldom that people being to trust systems blindly... then they go beyond next-word prediction... programmers give them hints of agency... alignment is a token gesture, not well thought through or understood... prejudices get rolled into a model and cemented into future versions... and before (who knows how) long, we have reached AGI and it's too late to ask if this system is ok, philosophically or otherwise.
I seesaw between unbridled joy at feeling so fortunate to be seeing these huge advances unravel in my lifetime and feelings of dread after listening to the likes of Eliezer Yudkowsky. Would love to hear your thoughts on the need and possibility for regulation.
Your takes and thoughts on AI and ChatGPT are the most balanced, nuanced, reasonable, addressing both objective hopes and concerns among those I have read, and they are always joy to read. I was to a Twitter whirlwind when all those releases were happening (and still are) and your Substack is a place that I think keeps me sane haha. Thanks
Without my having the ability to test Bard yet, can I ask is it simply a matter of not yet being up to snuff, or is it on the wrong track? I ask because some of your experiments with poetry suggest that it is generating material rather like some of what I was seeing in December.
I absolutely understand the last bit: I greatly ignored recent AI developments until ChatGPT. Reading and researching about AI - particularly its uses in education and research - has now become customary. I am both excited and terrified.
You mentioned that "Likely, between the time I send this email and you read it, Microsoft will have added image generation capabilities to Bing". It has been a sidebar option for my Bing (in Australia) for over a week. (Though it doesn't seem anywhere near as good as Midjourney currently.)
Outstanding work Ethan. Thank you for helping us to keep up with what is surely the most profound technological development (so far). Apparently things are just getting started. 🙂
Ethan, I've really come to enjoy your posts. I like your deep dives and perspective. Wondering if you are planning to look into how the guardrails or training is done with AI.
"This wasn’t supposed to be an AI-focused Substack." I feel ya on this one. I've gotten swept up in AI commentary and then work in the last few weeks, too. I don't know if the train will ever stop for some sightseeing, or if it's going to evolve around us to a bullet train and then a space ship while we're all just hanging out in the lounge asking Guinan for something blue to drink.
Ethan, witnessing the recent and thrilling advancements in AI and the public's reaction, I can't help but remind of Toffler's book "Future Shock," which predicted the rapid pace of technological change and its effects on society.
Love your posts. I’m a regular ole’ public school employee in a well regarded suburban school system and am already adopting ChatGPT for work and personal tasks. I enjoy reading what true experts are doing.
In 25 years time we’ve gone from working in the library to cut down on research time by being near the card catalog...then cutting days and weeks down to hours by researching on the internet, to cutting down weeks and hours into seconds with Chat GPT. It’s thrilling and one of the few moments in history there’s a tangible giant leap in technology in which so many regular people can capitalize. I remember the same feeling when I gave my senior project speech at University on “The Internet”.
I’ve also been going down the rabbit hole of personalized Python programs - and the only Python I know is essentially how to run PyCharm :)
Whenever I think of a small annoying problem I see if I can rip out a Python script for it and so far GPT-4 hasn’t let me down!
Yes to all of this. Hardly anyone seems to appreciate how quickly AI is going to change the world. It's not just the technology itself, which as shown in that general intelligence paper is even more impressive than we thought. It's that all the complementary pieces, like a huge software industry, computing infrastructure, and installed device and user base, are already in place. The expansion of the web was gated on availability of personal computers, servers, and users. Same with mobile and smartphones. There is no real gating factor on this one as far as I can tell, apart from how quickly people can come up with useful ideas and (with AI help) code them. I would guess it's a trillion dollar industry at a minimum within five years, which implies X trillion in economic value created.
And now, on the 8th day, we have plugins for ChatGPT.
LLMs and generative art AIs are progressing far faster than anyone expected, and it's going to be a challenge not just to keep up, but also to understand the impacts they are going to have and put mitigations in place where needed. In fact, I think that latter point is impossible.
I've been thinking about the impact of LLMs in particular on the publishing world, and I do fear that there's a lot of disruption coming down the pike towards us, and we just don't have the tools or understanding to get ready for it. And it's not all going to be good disruption:
https://wordcounting.substack.com/p/can-publishing-survive-the-oncoming
Another great post, Ethan, thank you. These advances are certainly as exciting as they are useful. Do we think though that there should be more trepidation and caution tempering this wave of excitement? Profit motives tend to push sensible conversations to the sidelines.
When each release builds on its predecessor in this way, it won’t be long before hallucinations happen so seldom that people being to trust systems blindly... then they go beyond next-word prediction... programmers give them hints of agency... alignment is a token gesture, not well thought through or understood... prejudices get rolled into a model and cemented into future versions... and before (who knows how) long, we have reached AGI and it's too late to ask if this system is ok, philosophically or otherwise.
I seesaw between unbridled joy at feeling so fortunate to be seeing these huge advances unravel in my lifetime and feelings of dread after listening to the likes of Eliezer Yudkowsky. Would love to hear your thoughts on the need and possibility for regulation.
Your takes and thoughts on AI and ChatGPT are the most balanced, nuanced, reasonable, addressing both objective hopes and concerns among those I have read, and they are always joy to read. I was to a Twitter whirlwind when all those releases were happening (and still are) and your Substack is a place that I think keeps me sane haha. Thanks
Without my having the ability to test Bard yet, can I ask is it simply a matter of not yet being up to snuff, or is it on the wrong track? I ask because some of your experiments with poetry suggest that it is generating material rather like some of what I was seeing in December.
It's interesting that GPT-4 will write code, but if you ask Bing, which uses GPT-4 to do it, it will tell you to hire an expert
I absolutely understand the last bit: I greatly ignored recent AI developments until ChatGPT. Reading and researching about AI - particularly its uses in education and research - has now become customary. I am both excited and terrified.
You mentioned that "Likely, between the time I send this email and you read it, Microsoft will have added image generation capabilities to Bing". It has been a sidebar option for my Bing (in Australia) for over a week. (Though it doesn't seem anywhere near as good as Midjourney currently.)
Great post! Thanks for the overview. Man, things are happening FAST.
Outstanding work Ethan. Thank you for helping us to keep up with what is surely the most profound technological development (so far). Apparently things are just getting started. 🙂