AI has turned out very differently than most of the originators expected.
I studied AI in college in the 90s under the label “symbolic systems” and that was the predominant idea of what AI would be — capturing semantic content in some sort of regularized notation, such that thinking could be produced in some logical and parameterized way.
The resulting AIs were imagined to be coldly rational, but without any sense of emotional content. They might work as physical laborers but would struggle to handle language, especially any kind of subtext or ambiguity.
As a simple example, it would have been considered somewhat ridiculous to imagine that an AI might produce a realistic looking painting in a certain style, with the wrong number of fingers on a person’s hand.
This is one of the questions Alexander Krauss addresses in his new book The Engine of Scientific Discovery: How New Methods and Tools Spark Major Breakthroughs. Although the book is new, it’s already been acknowledged as an important contribution to the philosophy of science and the science of science. It’s very good and I highly recommend it. He address your particular question in the later part of the book.
The flight example is interesting because many people (eg. lord kelvin) were convinced that heavier than air flying machines were completely impossible despite having seen a bird.
Futurism is the attempt to make this kind of projection semi-scientific. Futurism extends from the field of economics, mainly. And, much like economics, it has widely failed to make any accurate predictions at all. As a social science grounded in statistical projection, it is extremely fuzzy.
Adherents treat Kurzweil like some kind of soothsayer. But his predictions are accurate like horoscopes are accurate, just with added statistics around them to give them some truthiness.
Futurism is basically Asimov’s “psychohistory” mixed with Heinlein’s “future history,” paired with speculative market practices. It’s glorified pseudoscience.the tech industry loves it because it’s economics, and their entire game is one of finance and stock valuations.
As to your specific examples, I’d say the important projections about AI are mainly Vernor Vinge’s. The earlier stuff about automata in the 1800s even before that imagined sentient machines. There’s even hints of the idea back in the late 1500s and early 1600s with the rise of the “mechanical” model from which we inherit modern physics.
QM has a major delay because Einstein considered it a blunder. So it took a while before it got much further than then initial equations that suggested the interpretation. It’s also of limited practical value until you have the technological development to do anything with or about it.
Consider something like the MRI/fMRI issue where basically the entire history of published science in the field had to be retracted because we discovered the firmware was bad and all its measures were wrong. Any projections we might have made would be invalidated. And so the reflexive and iterative development of science itself is part of the delay, and challenges our capacity of prediction.
I am familiar with the name but not with the work. Where should I begin? My understanding is that these technological-breakthrough forecasts tend to be good when the end state is already known, and tend to be bad when the path runs into a combinatorial explosion of required innovations. For example, quantum computers were only possible once it became cheap to perform certain quantum-state calculations/manipulations , and that cheapness only arrived much later, after a long chain of intermediate discoveries, which could not be forcasted well.
Still OP question in my opinion is so generic and mudded with random concepts that this is the only possible answer, waiting for possibly a more articulated question.
amumpsimus | a day ago
AI has turned out very differently than most of the originators expected.
I studied AI in college in the 90s under the label “symbolic systems” and that was the predominant idea of what AI would be — capturing semantic content in some sort of regularized notation, such that thinking could be produced in some logical and parameterized way.
The resulting AIs were imagined to be coldly rational, but without any sense of emotional content. They might work as physical laborers but would struggle to handle language, especially any kind of subtext or ambiguity.
As a simple example, it would have been considered somewhat ridiculous to imagine that an AI might produce a realistic looking painting in a certain style, with the wrong number of fingers on a person’s hand.
DataMonster01 | a day ago
This is one of the questions Alexander Krauss addresses in his new book The Engine of Scientific Discovery: How New Methods and Tools Spark Major Breakthroughs. Although the book is new, it’s already been acknowledged as an important contribution to the philosophy of science and the science of science. It’s very good and I highly recommend it. He address your particular question in the later part of the book.
phrotozoa | 10 hours ago
The flight example is interesting because many people (eg. lord kelvin) were convinced that heavier than air flying machines were completely impossible despite having seen a bird.
XanderOblivion | 2 hours ago
Futurism is the attempt to make this kind of projection semi-scientific. Futurism extends from the field of economics, mainly. And, much like economics, it has widely failed to make any accurate predictions at all. As a social science grounded in statistical projection, it is extremely fuzzy.
Adherents treat Kurzweil like some kind of soothsayer. But his predictions are accurate like horoscopes are accurate, just with added statistics around them to give them some truthiness.
Futurism is basically Asimov’s “psychohistory” mixed with Heinlein’s “future history,” paired with speculative market practices. It’s glorified pseudoscience.the tech industry loves it because it’s economics, and their entire game is one of finance and stock valuations.
As to your specific examples, I’d say the important projections about AI are mainly Vernor Vinge’s. The earlier stuff about automata in the 1800s even before that imagined sentient machines. There’s even hints of the idea back in the late 1500s and early 1600s with the rise of the “mechanical” model from which we inherit modern physics.
QM has a major delay because Einstein considered it a blunder. So it took a while before it got much further than then initial equations that suggested the interpretation. It’s also of limited practical value until you have the technological development to do anything with or about it.
Consider something like the MRI/fMRI issue where basically the entire history of published science in the field had to be retracted because we discovered the firmware was bad and all its measures were wrong. Any projections we might have made would be invalidated. And so the reflexive and iterative development of science itself is part of the delay, and challenges our capacity of prediction.
Cybtroll | a day ago
...didn't Thomas Kuhn identified and explained EXACTLY this?
Sharp-Excitement-664 | a day ago
yes, but whether or not one reads his explanation as sufficient is an entirely different thing.
[OP] Tiny-Switch5965 | a day ago
I am familiar with the name but not with the work. Where should I begin? My understanding is that these technological-breakthrough forecasts tend to be good when the end state is already known, and tend to be bad when the path runs into a combinatorial explosion of required innovations. For example, quantum computers were only possible once it became cheap to perform certain quantum-state calculations/manipulations , and that cheapness only arrived much later, after a long chain of intermediate discoveries, which could not be forcasted well.
Cybtroll | a day ago
Still OP question in my opinion is so generic and mudded with random concepts that this is the only possible answer, waiting for possibly a more articulated question.
Immediate_Chard_4026 | a day ago
Si. Se llama deterministica y probabilistica.
BagsYourMail | a day ago
Yes, funding