My memory is that the version the creator used for his son and the person who hired him to create the primer used a version that required a voice actor. The version that was used by the cast-off Chinese girls was an artificial voice.
Stephenson believed that the human voice would be something that a computer couldn't convincingly recreate, creating a market for actors and actresses to voice virtual parts. Experience right now seems to suggest he was right to a degree, but wrong in the ability of artificial voices to come close to the experience of a human voice.
I believe you're correct; it has been more than several years since I read that book, but now that you mention it, I thought at the time that it was an odd experiment to have only one control and one alternative. Having read more experimental and clinal psychology studies since then, sample sizes are almost always woefully minimal.
AI, like many technologies, will have great advantages and great disadvantages. A disadvantage from where I sit as a History professor is that in the past year, student posts and essays have gotten considerably better in their grammar. I’m certain they are using AI, but no matter what any of the “experts” at my institution say, it is difficult, if not impossible to prove. It’s almost refreshing to read something from a student with clunky language, but I’m guessing AI could do that too, if the student gave it the right prompt. I just cannot trust that what I’m reading is actual student work, which in small ways can limit the credibility of high-performing students. I honestly don’t know what to do other than grade what I read as if it is legitimate. A possible solution would be to track every keystroke and record video of their work for the class, but that would be expensive and time consuming to monitor.
AI can provide information, however, it still has to be valid in reality. If there were a way to pair or test the information that we learn from AI with some reference to the real world that we know is accurate (verified articles/books/videos/etc. or actual people with expertise), then that would mitigate some of the possible disadvantages. The more sources of information you can access, the better. The proliferation of AI might have the ironic effect of increasing the value we place on in-person experiences. Computers are great, but people are better.
I have come to believe that autodidacticism is likely the best method of learning. We have had general public education for some time in the US, with various education theories and methods in use. The end result has< I think, demonstrated that education does not equal learning; i.e., that learning is an internal process that may be informed by, but not necessarily due to the external education process. If AI (probably more like an expert system with deep learning algorithms) can help an individual with exploring the current knowledge bases on various subjects in a structured, systematic way, based on the individual's interests and abilities, then it would I think be positive. If an educational AI is a top-down, one-size-fits-all system, like current public education, then I expect it to be like most all government expert products: an expensive failure that does more harm than good.
Just one more thing parents need to be aware of. I’ve already seen how my kids have subtly countered the nonsense their kids (my grandkids) learn at school. Parental involvement will be the key to dealing with AI.
I'm not all that concerned about the more nefarious aspects of AI. I am concerned about the specific narrative you've described with kids because it's YouTube, TikTok and other things that are worse than a waste of time, for kids, on steroids.
You don't make it explicit, but you imply it: All education is indoctrination. Learning to read requires reading something; almost nothing is value free and almost none of that is interesting.
This is wonderfully naive: "truth-telling AI". It raises all sorts interesting questions around "what is truth?" Just as value-free text is rarely interesting, obviously "true" and obviously "false" is also rarely interesting - not to mention difficult to find. "A triangle has three sides" is true. What about "A triangle's internal angles add to 180 degrees"? Not so true, but "false" seems to overshoot a bit.
One assumes that if there are many AIs, they will be much like people: Some are trustworthy, if not always correct, and some are not. The trustworthy ones will also be much like trustworthy people: Difficult to find.
I really wish "Artificial Intelligence" were called "machine intelligence". It would be more accurate and it would allow for easier neologisms. "Artificial Stranger Danger" is not as catchy as "Mechanical Stranger Danger". We teach children about the danger of human strangers (which is fairly easy since it's almost built-in to humans). We will need to teach children about the danger of machine strangers - and how difficult it is to tell the two apart on the Internet.
At the other end of the age curve, in assisted living and nursing homes, this subtle guardrails thing is called “redirection”. If great-gran starts to get agitated about the political news, the staff brightly and changes the subject, cheerfully asking a question about their newgreat-grand-baby, or telling them about the upcoming sing-along activity, or whatever to head off any real meltdowns.
So with AI tutor/buddy/minder/monitor, first the little kids, and then everyone, will get a built in assisted living aid, working hard to prevent any undesirable emotional forays.
Seems like using AI buddies to nudge people in the”approved” direction would be a primary use of widespread AI bots. It’s going on now with banner ads to a lesser degree. Making it fun and ‘cute’ would confirm the expression that “the road to Hell is paved with good intentions” … and guided by attractive spokesmodels. CS Lewis and Screwtape also come to mind. I completely understand the Dad’s point and eagerness to have the ultimately knowledgeable and patient teacher for his son, but he risks not having the wonderful experience of teaching his son and watching the transformation that he himself assists into being - and there is no greater reward than that for a parent.
As if we don't already have a crumbling society in no small part a result of disintegrating family structures and the relinquishment of foundational instruction to the anti-human academy, we can now shove our kids in a closet with their AI buddy and complete the process of creating the Borg collective.
I’ve been reading a SciFi series called “Sentenced To War” and one of the premises is that the Space Marines have battle buddies implanted in them. Not only is the AI their buddy and connects them with knowledge and communications, but it is also capable of affecting their moods and emotions by manipulating their neuro chemicals. Imagine that as the next level @Glenn
A class on Comparative Intelligence would be interesting and, more importantly today, fun. Assign a small, focused topic for a student to research and report in two ways, AI and traditional study methods, then have the student compare and present the results. Extra credit given to the student who can poke holes in the AI result.
I linked this at Chicago Boyz, where I connected it with some thoughts on explicit/textual versus graphical/sensorial communication that were propounded by Neal Stephenson.
I use Grok frequently and find it very useful. But like any data source it needs to be filtered through your personal experiences and opinions. For example, here is Grok's answer to the question: "What is your opinion of Glenn Reynolds?"
The answer is too long for this post, but here's the summary at the end:
'Reynolds, therefore, is seen as a thought leader by some, providing critical analysis of current events and law from a conservative/libertarian perspective, while others might view his commentary as partisan or occasionally provocative. His influence in shaping online political discourse, especially through blogging, cannot be understated. However, like many public figures with strong opinions, his work invites both admiration and critique.'
AI is in its infancy and will provide a range of responses based on the resources it accesses and the prejudices of the individuals creating the program. Grok uses the information on “X” for the latest information on evolving issues. If you ask it for the latest information and restrict it’s access to “X” it will provide information that’s not necessarily current.
I think of this as AI version 1. As it evolves it will be possible to create AI "buddies" that reflect the attitudes, belief systems and mental orientation of the people using them. In other words, AI version 2 will reflect the people using it rather than the people creating it.
Which may not be all good. Most political communication channels have become what someone called 'comfort food', helping readers/viewers feel like they are Good People who have other Good People agreeing with them. AI buddies could intensify this to an even higher level.
Your prediction is scary because it is so plausible. We are truly living in a "Brave New World." Hard to predict exactly how everything will unfold, but I agree completely that alternative, truth-telling AI must be developed and used before ChatGPT and its successors ensnare an entire generation. I am so glad I grew up in the 80s and can remember a joyful time before smartphones and the Internet. Technology has been a blessing -- AND a curse. No easy answers or solutions.
I think I need to read *The Young Lady's Illustrated Primer* one more time, to see how its aged...
I thought the same thing, except the AI tutor in The Diamon Age had a real person operating behind the screen....
My memory is that the version the creator used for his son and the person who hired him to create the primer used a version that required a voice actor. The version that was used by the cast-off Chinese girls was an artificial voice.
Stephenson believed that the human voice would be something that a computer couldn't convincingly recreate, creating a market for actors and actresses to voice virtual parts. Experience right now seems to suggest he was right to a degree, but wrong in the ability of artificial voices to come close to the experience of a human voice.
I believe you're correct; it has been more than several years since I read that book, but now that you mention it, I thought at the time that it was an odd experiment to have only one control and one alternative. Having read more experimental and clinal psychology studies since then, sample sizes are almost always woefully minimal.
AI, like many technologies, will have great advantages and great disadvantages. A disadvantage from where I sit as a History professor is that in the past year, student posts and essays have gotten considerably better in their grammar. I’m certain they are using AI, but no matter what any of the “experts” at my institution say, it is difficult, if not impossible to prove. It’s almost refreshing to read something from a student with clunky language, but I’m guessing AI could do that too, if the student gave it the right prompt. I just cannot trust that what I’m reading is actual student work, which in small ways can limit the credibility of high-performing students. I honestly don’t know what to do other than grade what I read as if it is legitimate. A possible solution would be to track every keystroke and record video of their work for the class, but that would be expensive and time consuming to monitor.
AI can provide information, however, it still has to be valid in reality. If there were a way to pair or test the information that we learn from AI with some reference to the real world that we know is accurate (verified articles/books/videos/etc. or actual people with expertise), then that would mitigate some of the possible disadvantages. The more sources of information you can access, the better. The proliferation of AI might have the ironic effect of increasing the value we place on in-person experiences. Computers are great, but people are better.
I have come to believe that autodidacticism is likely the best method of learning. We have had general public education for some time in the US, with various education theories and methods in use. The end result has< I think, demonstrated that education does not equal learning; i.e., that learning is an internal process that may be informed by, but not necessarily due to the external education process. If AI (probably more like an expert system with deep learning algorithms) can help an individual with exploring the current knowledge bases on various subjects in a structured, systematic way, based on the individual's interests and abilities, then it would I think be positive. If an educational AI is a top-down, one-size-fits-all system, like current public education, then I expect it to be like most all government expert products: an expensive failure that does more harm than good.
I for one welcome our new AI overlords. Till now I’ve been in charge of my life and it’s been a heavy lift, so sure.. aliens, AI, whatever.
Just one more thing parents need to be aware of. I’ve already seen how my kids have subtly countered the nonsense their kids (my grandkids) learn at school. Parental involvement will be the key to dealing with AI.
I'm not all that concerned about the more nefarious aspects of AI. I am concerned about the specific narrative you've described with kids because it's YouTube, TikTok and other things that are worse than a waste of time, for kids, on steroids.
You don't make it explicit, but you imply it: All education is indoctrination. Learning to read requires reading something; almost nothing is value free and almost none of that is interesting.
This is wonderfully naive: "truth-telling AI". It raises all sorts interesting questions around "what is truth?" Just as value-free text is rarely interesting, obviously "true" and obviously "false" is also rarely interesting - not to mention difficult to find. "A triangle has three sides" is true. What about "A triangle's internal angles add to 180 degrees"? Not so true, but "false" seems to overshoot a bit.
One assumes that if there are many AIs, they will be much like people: Some are trustworthy, if not always correct, and some are not. The trustworthy ones will also be much like trustworthy people: Difficult to find.
I really wish "Artificial Intelligence" were called "machine intelligence". It would be more accurate and it would allow for easier neologisms. "Artificial Stranger Danger" is not as catchy as "Mechanical Stranger Danger". We teach children about the danger of human strangers (which is fairly easy since it's almost built-in to humans). We will need to teach children about the danger of machine strangers - and how difficult it is to tell the two apart on the Internet.
At the other end of the age curve, in assisted living and nursing homes, this subtle guardrails thing is called “redirection”. If great-gran starts to get agitated about the political news, the staff brightly and changes the subject, cheerfully asking a question about their newgreat-grand-baby, or telling them about the upcoming sing-along activity, or whatever to head off any real meltdowns.
So with AI tutor/buddy/minder/monitor, first the little kids, and then everyone, will get a built in assisted living aid, working hard to prevent any undesirable emotional forays.
Hmmm.
Seems like using AI buddies to nudge people in the”approved” direction would be a primary use of widespread AI bots. It’s going on now with banner ads to a lesser degree. Making it fun and ‘cute’ would confirm the expression that “the road to Hell is paved with good intentions” … and guided by attractive spokesmodels. CS Lewis and Screwtape also come to mind. I completely understand the Dad’s point and eagerness to have the ultimately knowledgeable and patient teacher for his son, but he risks not having the wonderful experience of teaching his son and watching the transformation that he himself assists into being - and there is no greater reward than that for a parent.
As if we don't already have a crumbling society in no small part a result of disintegrating family structures and the relinquishment of foundational instruction to the anti-human academy, we can now shove our kids in a closet with their AI buddy and complete the process of creating the Borg collective.
I’ve been reading a SciFi series called “Sentenced To War” and one of the premises is that the Space Marines have battle buddies implanted in them. Not only is the AI their buddy and connects them with knowledge and communications, but it is also capable of affecting their moods and emotions by manipulating their neuro chemicals. Imagine that as the next level @Glenn
A few notes on things to come:
* We need decentralized AI and it is happening. I wrote a post on X last year on the hacker revolution in AI.
https://x.com/jsimonandersson/status/1723708130146549899
* We can make our own conversational knowledge bases to share the knowledge that Google or the educational system will not promote. A few thoughts here: https://x.com/jsimonandersson/status/1745039012627104235
* The knowledge bases should be linked, so you can talk to one and get the knowledge of a larger network. A first attempt at building that technology is here: https://x.com/jsimonandersson/status/1736347442122047580
I'm currently working on conversational AI that makes genomics accessible and lets you chat with your own DNA.
I think we are at the beginning of a golden age for learners of all ages.
A class on Comparative Intelligence would be interesting and, more importantly today, fun. Assign a small, focused topic for a student to research and report in two ways, AI and traditional study methods, then have the student compare and present the results. Extra credit given to the student who can poke holes in the AI result.
I linked this at Chicago Boyz, where I connected it with some thoughts on explicit/textual versus graphical/sensorial communication that were propounded by Neal Stephenson.
https://chicagoboyz.net/archives/72470.html
I use Grok frequently and find it very useful. But like any data source it needs to be filtered through your personal experiences and opinions. For example, here is Grok's answer to the question: "What is your opinion of Glenn Reynolds?"
The answer is too long for this post, but here's the summary at the end:
'Reynolds, therefore, is seen as a thought leader by some, providing critical analysis of current events and law from a conservative/libertarian perspective, while others might view his commentary as partisan or occasionally provocative. His influence in shaping online political discourse, especially through blogging, cannot be understated. However, like many public figures with strong opinions, his work invites both admiration and critique.'
AI is in its infancy and will provide a range of responses based on the resources it accesses and the prejudices of the individuals creating the program. Grok uses the information on “X” for the latest information on evolving issues. If you ask it for the latest information and restrict it’s access to “X” it will provide information that’s not necessarily current.
I think of this as AI version 1. As it evolves it will be possible to create AI "buddies" that reflect the attitudes, belief systems and mental orientation of the people using them. In other words, AI version 2 will reflect the people using it rather than the people creating it.
Which may not be all good. Most political communication channels have become what someone called 'comfort food', helping readers/viewers feel like they are Good People who have other Good People agreeing with them. AI buddies could intensify this to an even higher level.
Your prediction is scary because it is so plausible. We are truly living in a "Brave New World." Hard to predict exactly how everything will unfold, but I agree completely that alternative, truth-telling AI must be developed and used before ChatGPT and its successors ensnare an entire generation. I am so glad I grew up in the 80s and can remember a joyful time before smartphones and the Internet. Technology has been a blessing -- AND a curse. No easy answers or solutions.