33 ms·

These AI Models will make an impact in many domains and affect the distribution of job types available in art, legal, software, and other professions. To prepa

These AI Models will make an impact in many domains and affect the distribution of job types available in art, legal, software, and other professions. To prepare for the upcoming major changes, we need to urgently make creativity and innovation much more prominent in curricula across the board. Kids and adults will need to learn to adapt so they can thrive with such AI tools in the market. Skills such as creativity, collaboration, communication, innovation using new technologies will be even more crucial to a large number of people in the coming years. ============== I found this article to be a good discussion of many points involved. The author demonstrates pretty good understanding of the powers and limitations of these narrow AI models, although I'd say DALL-E-like models do have some level of understanding of how people use objects, based on both their textual and visual training data. https://betterprogramming.pub/dall-e-2-will-disrupt-art-deeper-than-photography-did-b900b52c1ea2 https://betterprogramming.pub/dall-e-2-will-disrupt-art-deep... "...What these systems have is the meta-ability of learning to paint — anything in any style, already invented or to be invented. This is a crucial difference and it’s what distinguishes DALL·E 2 from any previous disruptive single-purpose technology. ...Artists can escape a camera because it’s static in its abilities. It can take pics and nothing else. ... AI wouldn’t replace just one style, it would replace all styles. .... The most important one is that it doesn’t understand that the objects it paints are a reference to physical objects in the world. If you ask DALL·E 2 to draw a chair it can do it in all colors, forms, and styles. And still, it doesn’t know we use chairs for sitting down. Art is only meaningful in its relationship with either the external physical world or our internal subjective world — and DALL·E 2 doesn’t have access to the former and lacks the latter. ... This means it can’t understand the underlying realities in which its paintings are based and generalize those into novel situations, styles, or ideas it has never seen before. This limitation is key because we humans can do it very easily. ..."

rblatz 4y ago
I’ve seen som hilarious DALL-E images like a razer gaming coffin. That are exactly what you would expect and also not I would say fairly novel images.
Lwepz 4y ago
>it can’t understand the underlying realities in which its paintings are based and generalize those into novel situations, styles, or ideas it has never seen before As long as we keep putting a greater emphasis on the final output than we do on the process that led to the output, this "limitation" won't be a factor.
acoard 4y ago
> And still, it doesn’t know we use chairs for sitting down. I imagine if I gave it the prompt "chair being used by woman" then a majority of the pics would have someone sitting in it. I realize your point is more of a "Chinese Room" philosophical objection. But it's important to note the level of sophistication these models have. I'll also add that Nvidia is currently doing research on 3D model based generation, so that the model is aware of the 'skeleton' if the subject before illustrating it.

rblatz 4y ago

I’ve seen som hilarious DALL-E images like a razer gaming coffin. That are exactly what you would expect and also not I would say fairly novel images.

Lwepz 4y ago

>it can’t understand the underlying realities in which its paintings are based and generalize those into novel situations, styles, or ideas it has never seen before As long as we keep putting a greater emphasis on the final output than we do on the process that led to the output, this "limitation" won't be a factor.

acoard 4y ago

> And still, it doesn’t know we use chairs for sitting down. I imagine if I gave it the prompt "chair being used by woman" then a majority of the pics would have someone sitting in it. I realize your point is more of a "Chinese Room" philosophical objection. But it's important to note the level of sophistication these models have. I'll also add that Nvidia is currently doing research on 3D model based generation, so that the model is aware of the 'skeleton' if the subject before illustrating it.