Stories
Slash Boxes
Comments

SoylentNews is people

posted by hubie on Monday November 20 2023, @03:53AM   Printer-friendly
from the dystopia-is-now! dept.

https://arstechnica.com/information-technology/2023/11/from-toy-to-tool-dall-e-3-is-a-wake-up-call-for-visual-artists-and-the-rest-of-us/

In October, OpenAI launched its newest AI image generator—DALL-E 3—into wide release for ChatGPT subscribers. DALL-E can pull off media generation tasks that would have seemed absurd just two years ago—and although it can inspire delight with its unexpectedly detailed creations, it also brings trepidation for some. Science fiction forecast tech like this long ago, but seeing machines upend the creative order feels different when it's actually happening before our eyes.

"It's impossible to dismiss the power of AI when it comes to image generation," says Aurich Lawson, Ars Technica's creative director. "With the rapid increase in visual acuity and ability to get a usable result, there's no question it's beyond being a gimmick or toy and is a legit tool."

[...] ChatGPT and DALL-E 3 currently work hand-in-hand, making AI art generation into an interactive and conversational experience. You tell ChatGPT (through the GPT-4 large language model) what you'd like it to generate, and it writes ideal prompts for you and submits them to the DALL-E backend. DALL-E returns the images (usually two at a time), and you see them appear through the ChatGPT interface, whether through the web or via the ChatGPT app.

[...] However, those scraped captions—written by humans—aren't always detailed or accurate, which leads to some faulty associations that reduce an AI model's ability to follow a written prompt.

To get around that problem, OpenAI decided to use AI to improve itself. As detailed in the DALL-E 3 research paper, the team at OpenAI trained this new model to surpass its predecessor by using synthetic (AI-written) image captions generated by GPT-4V, the visual version of GPT-4. With GPT-4V writing the captions, the team generated far more accurate and detailed descriptions for the DALL-E model to learn from during the training process.


Original Submission

This discussion was created by hubie (1068) for logged-in users only, but now has been archived. No new comments can be posted.
Display Options Threshold/Breakthrough Mark All as Read Mark All as Unread
The Fine Print: The following comments are owned by whoever posted them. We are not responsible for them in any way.
(1)
  • (Score: 0, Disagree) by Anonymous Coward on Monday November 20 2023, @04:46AM (8 children)

    by Anonymous Coward on Monday November 20 2023, @04:46AM (#1333576)

    The death of what meaning remains.
    First, they automated the menial, repetitive tasks.
    Now, they automate the mind.

    There will be nothing left for humanity.
    Our only choice will be to evolve.
    To become something other than human, but human at heart all the same.

    • (Score: 4, Insightful) by maxwell demon on Monday November 20 2023, @07:08AM (2 children)

      by maxwell demon (1608) on Monday November 20 2023, @07:08AM (#1333578) Journal

      Machines still only automate those tasks where the end product is removed fron the humans who produced it. There's still no stage play done by machines, and when that eventually arrives, it will be watched for what it is, not as replacement of human acting.

      Also note that it is still the humans who give the prompt, and decide whether or not the result is good or needs refinement.

      --
      The Tao of math: The numbers you can count are not the real numbers.
      • (Score: 3, Insightful) by Unixnut on Monday November 20 2023, @09:19AM (1 child)

        by Unixnut (5779) on Monday November 20 2023, @09:19AM (#1333589)

        > Also note that it is still the humans who give the prompt, and decide whether or not the result is good or needs refinement.

        And the fact the entire system relies on human input to initially train. At best it can regurgitate something based on the model it has developed from studying human input.

        As such I still don't consider what these tools do as Intelligence in the same vein as humans. It is machine learning, and the machine can only learn as well as what the humans train it to.
        The best you can say is that humans impart a bit of their intelligence into a machine via the ML process, which has its current uses and could well become more useful in future, but at its core it is just mimicking a narrow set of human capability.

        • (Score: 2, Interesting) by shrewdsheep on Monday November 20 2023, @11:04AM

          by shrewdsheep (5215) on Monday November 20 2023, @11:04AM (#1333594)

          But how to define intelligence, creativity, insight? Definitions can be made but they will always be arbitrary. The fact, that we seem to understand each and every step at the lower level, as we as humans created them, is not an argument. I believe that we indeed are entering a transition period towards robots becoming dominant, and probably for the good rather than bad.

          20-60 years ago the whole process has been thought through already. You can read Asimov, Hofstadter, Dennet and others. The discussion around the chinese room (https://en.wikipedia.org/wiki/Chinese_room) might be enlightening.

    • (Score: 2, Insightful) by pTamok on Monday November 20 2023, @07:43AM (3 children)

      by pTamok (3042) on Monday November 20 2023, @07:43AM (#1333581)

              First they came for the manual labourers, and I did not speak out—
                        Because I was not a manual labourer.

              Then they came for the knowledge workers, and I did not speak out—
                        Because I was not a knowledge worker.

              Then they came for the owners of the means of production, and I did not speak out—
                        Because I was not an owner of the means of production.

              Then they came for me—and there was no one left to speak for me.

      • (Score: 2, Touché) by khallow on Monday November 20 2023, @06:42PM

        by khallow (3766) Subscriber Badge on Monday November 20 2023, @06:42PM (#1333643) Journal
        Nobody is coming for manual laborers, knowledge workers, owners of means of production, or you. So speak out against what? The future being more uncertain than you would like?
      • (Score: 2) by Reziac on Tuesday November 21 2023, @02:51AM

        by Reziac (2489) on Tuesday November 21 2023, @02:51AM (#1333681) Homepage

        +1 Ouch

        --
        And there is no Alkibiades to come back and save us from ourselves.
      • (Score: 2) by ledow on Wednesday November 22 2023, @08:26AM

        by ledow (5567) on Wednesday November 22 2023, @08:26AM (#1333836) Homepage

        It's 2023. "Manual" labourers are using JCBs, power tools, pneumatic drills, etc. and are subject to health and safety laws that basically require them not to do heavy lifting, stupid things or dangerous stuff with a lot of protection and assistance.

        I don't see anything bad in that.

    • (Score: 2) by ledow on Wednesday November 22 2023, @08:23AM

      by ledow (5567) on Wednesday November 22 2023, @08:23AM (#1333835) Homepage

      Or we could just pay people universal basic income (never had an unsuccessful trial), automate everything that anyone wants automated, and let humans exist however they choose to with no obligations.

  • (Score: 2) by r_a_trip on Monday November 20 2023, @07:55AM (4 children)

    by r_a_trip (5276) on Monday November 20 2023, @07:55AM (#1333583)

    Yeah, I can't really feel any sympathy for our poor artists. Where were they when the industrialization and mechanization made thousands of tasks inaccessible for humans. Safely watching from the sidelines, because machines couldn't do what they do. Oh yes, there was enough paint smeared on canvasses and other mediums used to comment, but it stayed at that, commenting. Well, now it's time for the low to midlevel artists to be automated away. Maybe the workers who previously did the mechanized tasks will comment.

    • (Score: 2) by HiThere on Monday November 20 2023, @01:56PM

      by HiThere (866) Subscriber Badge on Monday November 20 2023, @01:56PM (#1333602) Journal

      I think you have seen a severely curated selection of artistic images. Not all artists focused on the "plight of the workers", but a considerable number did, on various sides.

      --
      Javascript is what you use to allow unknown third parties to run software you have no idea about on your computer.
    • (Score: 3, Touché) by Tork on Monday November 20 2023, @05:56PM

      by Tork (3914) Subscriber Badge on Monday November 20 2023, @05:56PM (#1333639)

      Yeah, I can't really feel any sympathy for our poor artists.

      You mean you don't want to and formed a strange rationalization to justify it. In practical terms you ain't helpin' yourself if you're being frivolously picky about who your allies are.

      --
      🏳️‍🌈 Proud Ally 🏳️‍🌈
    • (Score: 1) by khallow on Monday November 20 2023, @06:09PM

      by khallow (3766) Subscriber Badge on Monday November 20 2023, @06:09PM (#1333640) Journal

      Where were they when the industrialization and mechanization made thousands of tasks inaccessible for humans.

      Rather they made thousands of tasks accessible [soylentnews.org] to humans. In that link, I describe two problems that were made not merely accessible, but easily accessible.

      [Azuma Hazuki:] We really are limited by our biology, just nowhere near to the extent people who choose to wallow in the worst aspects of it insist.

      [...]

      We're reaching the limits of what we are capable of by nature.

      So which is it? Are we reaching the limits as the wallowers claim or not?

      My take is that we're already well beyond the limits of our biology. A few weeks back I was driving 80 MPH, merely to get from point A to B. Up to about 170 years when sufficiently fast mechanical travel was developed [wikipedia.org], going 80 MPH meant you most likely were seconds from death, because you were in some sort of accident or disaster.

      For another example, two or three decades back, I implemented one of the Mersenne prime finders on computer. Over a single night, I generated all of the Mersenne primes [wikipedia.org] (they are all of the form two to a power of a prime minus one, simple examples are 3, 7, and 31) that anyone had computed by hand plus a little more. In other words, many year of hard work done overnight as a lark.

      No task can be made inaccessible by a better way to do the task. It just doesn't make sense to do it the old way when you can do much better.

    • (Score: 1, Touché) by Anonymous Coward on Monday November 20 2023, @06:20PM

      by Anonymous Coward on Monday November 20 2023, @06:20PM (#1333641)
      printing press, camera, mechanical pencils, erasers, a mountain of drawing tools, computers, printers, copiers, photoshop, kai's power goo....
(1)