5 ms·

"We tend to overestimate the effect of a technology in the short run and underestimate the effect in the long run." - Roy Amara. This is the early overestimati

"We tend to overestimate the effect of a technology in the short run and underestimate the effect in the long run." - Roy Amara. This is the early overestimation.

thinkingkong 4y ago
On the contrary. The overestimation was probably in the previous AI era which couldnt live up to any of the hype.
- QuantumGood 4y ago
  I think "a technology" can be extended to "each wave of technological development". "A technology" does not begin at some point and extend into the future infinitely in the form it was originally defined. Each new wave runs into the overestimation/underestimation problem.
- aiappreciator 4y ago
  This, image AI has been a decade in the making already. Previously it was GANs for facial recognition, for edge detection, ultimately for self-driving. The impact of that on society has been insignificant, so a big overestimation. Then came transformers, then came combining transformers with diffusions, an innocent little combination, that now becomes a tsunami in image generation. Artists are launching mass protests online in extreme panic and anguish, tsunami-tide of consumer applications are coming each month, many of them extremely profitable and don't even need VC money. This is the long-term impact phase already.
  - woah 4y ago
    > many of them extremely profitable and don't even need VC money Really? Do you have some examples?
  - canadianfella 4y ago
Dove 4y ago
While I don't agree with the doomsaying of the article, I do think the current round of thinking falls into the "late underestimation" category, in the grand scheme of things. Science fiction from the 1960's through about the 1990's was of the opinion that AI would produce intelligence indistinguishable from humans, and would do so extremely quickly (we should have had it by now), and that the most pressing questions would be things like whether those intelligences deserved voting rights and whether they made humanity obsolete. That was a wild overestimation of what the technology was capable of. But the widespread application of AI, either in search results, social media moderation, propaganda and disinformation, bots, fake reviews, or just straight up spam, forms an absolute assault on the nature of trust and information in society, overturning heuristics that have been in effect for all of history (that people are usually trustworthy, that information is usually representative, that lying is hard). While I don't think the threat is apocalyptic, I do think no one even remotely saw this coming.
- PinkMilkshake 4y ago
  > I do think no one even remotely saw this coming. This is just outside of your time range but Metal Gear Solid 2: Sons of Liberty came out in 2001, and predicted much of this. The main example of this pertains to the plot and themes of Metal Gear Solid 2: Sons of Liberty, released on November 13, 2001, which delved into ideas and concepts that would become culturally significant in the 2010s. Among these themes were post-truth politics, alternative facts, echo-chambers, fake news, AI-curated news feeds, information overload in the Information Age, and political correctness.[140] While the game received universal acclaim upon release for its gameplay and attention to detail, the plot became a divisive topic among critics, with some calling it "absurd" and "stupid". - https://en.wikipedia.org/wiki/Hideo_Kojima#Legacy https://en.wikipedia.org/wiki/Hideo_Kojima#Legacy

"We tend to overestimate the effect of a technology in the short run and underestimate the effect in the long run." - Roy Amara. This is the early overestimation.

thinkingkong 4y ago

On the contrary. The overestimation was probably in the previous AI era which couldnt live up to any of the hype.

QuantumGood 4y ago
I think "a technology" can be extended to "each wave of technological development". "A technology" does not begin at some point and extend into the future infinitely in the form it was originally defined. Each new wave runs into the overestimation/underestimation problem.
aiappreciator 4y ago
This, image AI has been a decade in the making already. Previously it was GANs for facial recognition, for edge detection, ultimately for self-driving. The impact of that on society has been insignificant, so a big overestimation. Then came transformers, then came combining transformers with diffusions, an innocent little combination, that now becomes a tsunami in image generation. Artists are launching mass protests online in extreme panic and anguish, tsunami-tide of consumer applications are coming each month, many of them extremely profitable and don't even need VC money. This is the long-term impact phase already.
- woah 4y ago
  > many of them extremely profitable and don't even need VC money Really? Do you have some examples?
- canadianfella 4y ago

Dove 4y ago

While I don't agree with the doomsaying of the article, I do think the current round of thinking falls into the "late underestimation" category, in the grand scheme of things. Science fiction from the 1960's through about the 1990's was of the opinion that AI would produce intelligence indistinguishable from humans, and would do so extremely quickly (we should have had it by now), and that the most pressing questions would be things like whether those intelligences deserved voting rights and whether they made humanity obsolete. That was a wild overestimation of what the technology was capable of. But the widespread application of AI, either in search results, social media moderation, propaganda and disinformation, bots, fake reviews, or just straight up spam, forms an absolute assault on the nature of trust and information in society, overturning heuristics that have been in effect for all of history (that people are usually trustworthy, that information is usually representative, that lying is hard). While I don't think the threat is apocalyptic, I do think no one even remotely saw this coming.

PinkMilkshake 4y ago
> I do think no one even remotely saw this coming. This is just outside of your time range but Metal Gear Solid 2: Sons of Liberty came out in 2001, and predicted much of this. The main example of this pertains to the plot and themes of Metal Gear Solid 2: Sons of Liberty, released on November 13, 2001, which delved into ideas and concepts that would become culturally significant in the 2010s. Among these themes were post-truth politics, alternative facts, echo-chambers, fake news, AI-curated news feeds, information overload in the Information Age, and political correctness.[140] While the game received universal acclaim upon release for its gameplay and attention to detail, the plot became a divisive topic among critics, with some calling it "absurd" and "stupid". - https://en.wikipedia.org/wiki/Hideo_Kojima#Legacy https://en.wikipedia.org/wiki/Hideo_Kojima#Legacy