Incredible. I knew about modems back then, but we didn't have one... or at least, I don't remember us having one. It was hard to see what was about to happen, but we were definitely on the cusp.
Ugh. I really don’t like the phone app for Substack. I just wrote a very well thought out and BRILLIANT comment only for it to be erased because I dared to leave the app for a moment to check something on the Googles. And I’m too lazy to type it out again. Basically, my concern is going too fast… bah blah blah… speed kills. And my hope is that is remembering to teach the machines to be “inspired” by the human perspective and not theirs when they get too fast for us.
That happens to me all the time!! A lot of times I will wait until I get to my laptop if I know it’s going to be a long comment, or I will periodically copy and paste what I’m writing into a google doc on my phone lol
Yes Rudy, the app stinks. I'm using the Android version and it has all kinds of weird quirks.
A minor one, but although I run my phone on dark mode, there are certain apps and websites I prefer lighter. Generally that's not an issue but there's no simple light dark toggle. 🤦🏻
Wow, there's a lot of information in your article, Andrew. I skimmed it quickly because the topic engages me, but I'll read it again at my leisure.
Ray Kurzweil is a transhumanist. If he could, he'd probably take his own head apart and put it on a machine. Transhumanists overestimate technology because they have a lot of faith in it. Hm, well, they also need to get funding, ya know. But to reach the singularity, as Kurzweil means it, technology would have to develop self-consciousness. Not yet possible and probably never will be. This brings us to the concept of consciousness, and pops up huge scenarios.
I'm not so quick to dismiss the idea of emergent consciousness, but that's only because we really have no good working definition of what consciousness even is, and we know next to nothing about our own consciousness. That being said, I'm plenty skeptical of any wild claims that aren't backed by ample evidence, and a lot of this is still speculation.
You nailed it. Unfortunately, it is very difficult to find evidence to establish the nature of consciousness. You can look at it by field: psychology, philosophy, faith. We can also separate it from the realm of spiritual or human self-awareness. But I can't think of it as software.
Similarly, I can't think of it as "not software" or exclusive to the biological domain. Unfortunately, optimistic people who see things through rose-tinted lenses are easily fooled, and there's gonna be even more scamming than we've already seen to that end. Ugh!
Ah yes, I recall when Internet was dial-up and expensive, and accompanied by screeching modem noises when trying to download web pages. Good times.
As for the exponential pace of progress, I recall already sharing Tim Urban's analogy with bringing a person from the past into an arbitrary point in the future where so much progress has happened that this hypothetical person's brain simply cannot cope with it all and (s)he dies. And that gap has been shrinking. Right now, my mom is already hopelessly behind on this whole AI thing and it may as well be sci-fi mumno jumbo to her. It's entirely possible that within our lifetime this gap shrinks so much that even seemingly tech-savvy early adopters like myself will feel out of touch.
Fast forwarding to the end of the thesis, the idea is that only augmented folks (EG, some kind of direct brain-internet interface) will be able to keep up... and I have to say, I am completely convinced that this will happen. I'm not sure if it's like 30 years from now, or 10, or 5, or 50, or 100, but it's clearly in sight.
People keep saying the AI infection will slow, but what harm will be done in the mean time? And don't get me started on that Big Bang Theory nonsense. Humans limit their minds to beginnings and ends. They can't seem to accept that the universe just is and always was and always will be, and that it has no limits. They keep touting what they have been able to perceive through various devices and methodologies is the limits of that universe, but it is hubris to conclude nothing lies beyond. I will continue to hold onto my humanity and hope this AI madness stops in the remaining years of my lifetime.
My streaming shows were OnAir (indie music, still have a set of CDs) and Nelson’s World on MSN when it had that OnStage format in 1996-1999. Then it became the portal. Wiki had some information on this. They wayback machine certainly does too if you ever want to go spelunking into Microsoft’s early streaming efforts.
I had to fly with a real audio device that looked like a bomb, in carry on. I could because it was pre 9/11. I was encouraged to do this for Pop Mart tour but had just moved my grandmother in with me after my aunt abandoned her in Alaska. Otherwise I was freshly divorced and would have had a blast.
I set up a user controlled web camera at local ISPs for interviews. I was paid overtime to sleep on planes. Bands I’d heard of made me tea in their homes and let me hold their babies.
In office I did point and click web testing before test tools existed. 💕
I wish I’d started source control before the 4 days it had my stuff totally broken (from working) and kept cycling through the same rejected approaches. I did it for my sanity and to quantify later :).
Good read, even though I do tend to disagree. I’ve written quite a bit about the nature of current machine intelligence. I think you’re right to zoom out, though. In the narrow view, most “exponential” growth ends up looking like a logistic curve after a while.
However, when you zoom out to the greater universe, life has a big pair of pants to grow into.
I don’t really think that neural networks are “intelligent”, but they definitely can expand on what we’ve got to work with.
True, it's much more about minimizing tedious tasks right now. Think it will improve enough to be useful to you? I like to keep in mind that these tools are as bad as they're ever going to be, but I don't have a great sense of how wide that gap might be for a very good coder.
If a language model has a 90-95% accuracy on tasks (which they don't yet), then I still have to check every single output because 1/20 times the output will be wrong.
This throws a big wrench in anything being autonomous, because we're seeing linear gains in performance with exponential increase in resource requirements, in transformers at least.
The copilot stuff is pretty useful as a reference. For developers, I think it's mostly used as a replacement for tools like Stackoverflow or other online bug-squashing workflows and tutorial content. This is because it's trained on these things already, and pretty good generally at regurgitating relevant content. However, it's also only so accurate, and oftentimes developers find themselves having to correct obscure bugs, because the model can produce things which look right, but actually aren't.
It's gotta be a step up for the times when you'd need to find code somewhere to copy and paste, right? I imagine it's a time saver for the most tedious sorts of tasks that don't require as much intense focus, but that's really just a guess.
Precisely, great guess lol! It's good at simple, repetitive things that don't take a lot of thinking, and mostly repeat the same logic while changing some parts. It is also easier to get answers to questions like “How do I do this x with y technology?”, or “Why is this bug arising with y technology”, etc.
Would you say that this ability (to reduce repetitive/mindless stuff) reduces your total time by a bit? If so, could you quantify how much time they save?
I will soon write about this. I am suffering through making ai help me write an app I could have finished by now and (now) keeping source control to quantify how bad it is. Very very bad, lol. It’s not saving anyone much time 🤣
One cool direction I see coming is a new reliance on nuclear energy, and hopefully we see innovations in that industry as well. This will do wonders for pivoting the US to a cleaner energy grid.
Thanks so much, Bill. I really figure whatever interests me will probably interest you, too. I think I'm right about that enough of the time to get away with this!
The article is interesting, but you missed the human factor. It may be that laws of nature permit continual improvement, but are humans really interested in continually pushing the boundaries of innovation?
Are the incentives right? We primarily pursued technological innovation to make our lives more comfortable. Are we comfortable enough now?
I continue Moores Law in both of my Sci-Fi books. When you can start breaking into quantum and organic another whole new world opens up.
In 20 years, the line will look vertical and the crazy progress we're making today will look horizontal. That's pretty wild.
I started out in 1985 online at 300bps. Not kbps mind you 🤯🤣
Incredible. I knew about modems back then, but we didn't have one... or at least, I don't remember us having one. It was hard to see what was about to happen, but we were definitely on the cusp.
Had to connect with my bare hands… throwing a switch at just the right time… in the snow… 300 bps both ways…
Kids these days have no idea…
That's nothing! Back in my day, we had to carry the individual electrons, one at a time, across the ocean if we wanted to send a text.
1982 here. My flip phone had that old dial up sound you heard when connecting as a ringtone…I thought it was funny.
Ugh. I really don’t like the phone app for Substack. I just wrote a very well thought out and BRILLIANT comment only for it to be erased because I dared to leave the app for a moment to check something on the Googles. And I’m too lazy to type it out again. Basically, my concern is going too fast… bah blah blah… speed kills. And my hope is that is remembering to teach the machines to be “inspired” by the human perspective and not theirs when they get too fast for us.
That happens to me all the time!! A lot of times I will wait until I get to my laptop if I know it’s going to be a long comment, or I will periodically copy and paste what I’m writing into a google doc on my phone lol
I almost never respond in any way other than from my laptop, and i regret it almost every time I break that protocol.
Yes Rudy, the app stinks. I'm using the Android version and it has all kinds of weird quirks.
A minor one, but although I run my phone on dark mode, there are certain apps and websites I prefer lighter. Generally that's not an issue but there's no simple light dark toggle. 🤦🏻
I only ever use Substack from my laptop. It's just too dumb and confusing otherwise.
Wow, there's a lot of information in your article, Andrew. I skimmed it quickly because the topic engages me, but I'll read it again at my leisure.
Ray Kurzweil is a transhumanist. If he could, he'd probably take his own head apart and put it on a machine. Transhumanists overestimate technology because they have a lot of faith in it. Hm, well, they also need to get funding, ya know. But to reach the singularity, as Kurzweil means it, technology would have to develop self-consciousness. Not yet possible and probably never will be. This brings us to the concept of consciousness, and pops up huge scenarios.
I'm not so quick to dismiss the idea of emergent consciousness, but that's only because we really have no good working definition of what consciousness even is, and we know next to nothing about our own consciousness. That being said, I'm plenty skeptical of any wild claims that aren't backed by ample evidence, and a lot of this is still speculation.
You nailed it. Unfortunately, it is very difficult to find evidence to establish the nature of consciousness. You can look at it by field: psychology, philosophy, faith. We can also separate it from the realm of spiritual or human self-awareness. But I can't think of it as software.
Similarly, I can't think of it as "not software" or exclusive to the biological domain. Unfortunately, optimistic people who see things through rose-tinted lenses are easily fooled, and there's gonna be even more scamming than we've already seen to that end. Ugh!
Oh, yes. Definitely.
Ah yes, I recall when Internet was dial-up and expensive, and accompanied by screeching modem noises when trying to download web pages. Good times.
As for the exponential pace of progress, I recall already sharing Tim Urban's analogy with bringing a person from the past into an arbitrary point in the future where so much progress has happened that this hypothetical person's brain simply cannot cope with it all and (s)he dies. And that gap has been shrinking. Right now, my mom is already hopelessly behind on this whole AI thing and it may as well be sci-fi mumno jumbo to her. It's entirely possible that within our lifetime this gap shrinks so much that even seemingly tech-savvy early adopters like myself will feel out of touch.
Fast forwarding to the end of the thesis, the idea is that only augmented folks (EG, some kind of direct brain-internet interface) will be able to keep up... and I have to say, I am completely convinced that this will happen. I'm not sure if it's like 30 years from now, or 10, or 5, or 50, or 100, but it's clearly in sight.
I've played the Bioshock series, I know how this goes. If I wanna keep up, I gotta start splicing. This can only be a good thing, right?
What could go wrong?
Nothing at all, I am told by reliable sources. (My ADAM-addled brain.)
People keep saying the AI infection will slow, but what harm will be done in the mean time? And don't get me started on that Big Bang Theory nonsense. Humans limit their minds to beginnings and ends. They can't seem to accept that the universe just is and always was and always will be, and that it has no limits. They keep touting what they have been able to perceive through various devices and methodologies is the limits of that universe, but it is hubris to conclude nothing lies beyond. I will continue to hold onto my humanity and hope this AI madness stops in the remaining years of my lifetime.
I remember MSN was streaming media in 1996 when most had 9600 and 14400. I was a tester and tech support for OnAir then. Remember the U2 Pop Tour?
I do not, but I am familiar with some very early video streaming efforts. What was your work like?
My streaming shows were OnAir (indie music, still have a set of CDs) and Nelson’s World on MSN when it had that OnStage format in 1996-1999. Then it became the portal. Wiki had some information on this. They wayback machine certainly does too if you ever want to go spelunking into Microsoft’s early streaming efforts.
I had to fly with a real audio device that looked like a bomb, in carry on. I could because it was pre 9/11. I was encouraged to do this for Pop Mart tour but had just moved my grandmother in with me after my aunt abandoned her in Alaska. Otherwise I was freshly divorced and would have had a blast.
I set up a user controlled web camera at local ISPs for interviews. I was paid overtime to sleep on planes. Bands I’d heard of made me tea in their homes and let me hold their babies.
In office I did point and click web testing before test tools existed. 💕
That might be fun to write about!
I wish I’d started source control before the 4 days it had my stuff totally broken (from working) and kept cycling through the same rejected approaches. I did it for my sanity and to quantify later :).
Good read, even though I do tend to disagree. I’ve written quite a bit about the nature of current machine intelligence. I think you’re right to zoom out, though. In the narrow view, most “exponential” growth ends up looking like a logistic curve after a while.
However, when you zoom out to the greater universe, life has a big pair of pants to grow into.
I don’t really think that neural networks are “intelligent”, but they definitely can expand on what we’ve got to work with.
Also some people write better code with AI but I have not had that experience 😂
True, it's much more about minimizing tedious tasks right now. Think it will improve enough to be useful to you? I like to keep in mind that these tools are as bad as they're ever going to be, but I don't have a great sense of how wide that gap might be for a very good coder.
If a language model has a 90-95% accuracy on tasks (which they don't yet), then I still have to check every single output because 1/20 times the output will be wrong.
This throws a big wrench in anything being autonomous, because we're seeing linear gains in performance with exponential increase in resource requirements, in transformers at least.
The copilot stuff is pretty useful as a reference. For developers, I think it's mostly used as a replacement for tools like Stackoverflow or other online bug-squashing workflows and tutorial content. This is because it's trained on these things already, and pretty good generally at regurgitating relevant content. However, it's also only so accurate, and oftentimes developers find themselves having to correct obscure bugs, because the model can produce things which look right, but actually aren't.
It's gotta be a step up for the times when you'd need to find code somewhere to copy and paste, right? I imagine it's a time saver for the most tedious sorts of tasks that don't require as much intense focus, but that's really just a guess.
Precisely, great guess lol! It's good at simple, repetitive things that don't take a lot of thinking, and mostly repeat the same logic while changing some parts. It is also easier to get answers to questions like “How do I do this x with y technology?”, or “Why is this bug arising with y technology”, etc.
Would you say that this ability (to reduce repetitive/mindless stuff) reduces your total time by a bit? If so, could you quantify how much time they save?
I will soon write about this. I am suffering through making ai help me write an app I could have finished by now and (now) keeping source control to quantify how bad it is. Very very bad, lol. It’s not saving anyone much time 🤣
Fascinating! I look forward to seeing this!
Same! Please feel free to tag me whenever the post goes live.
Actually, consequences exist for this new world, don't they? The energy to power the process scares me.
One cool direction I see coming is a new reliance on nuclear energy, and hopefully we see innovations in that industry as well. This will do wonders for pivoting the US to a cleaner energy grid.
I agree. One way could be through fusion, merging atoms together to create energy.
Great and thoughtful essay Andrew. Well done. It is up to each of us to be excited or frightened by the possibilities—because we cannot stop it.
Thanks, Dee! I agree completely.
Perhaps it calls for a measurement other than time…
I'm continually stunned by the wide variety of topics in your posts. Is there anything you don't write about? (Rhetorical question).
Thanks so much, Bill. I really figure whatever interests me will probably interest you, too. I think I'm right about that enough of the time to get away with this!
Similar thoughts... Exponential Everywhere. https://www.nextcolabscommunity.io/blog/exponential-ai
The article is interesting, but you missed the human factor. It may be that laws of nature permit continual improvement, but are humans really interested in continually pushing the boundaries of innovation?
Are the incentives right? We primarily pursued technological innovation to make our lives more comfortable. Are we comfortable enough now?
Innovation is not guaranteed.
https://open.substack.com/pub/lithiumhorizons/p/the-paradox-of-technological-advancement?r=68sw4&utm_medium=ios