3 ms·
And to emphasize your point, it seems from the coverage like the problem isn’t so much that overbearing managers are setting out to nanny the users as it is tha
by alwa 3y ago
And to emphasize your point, it seems from the coverage like the problem isn’t so much that overbearing managers are setting out to nanny the users as it is that they’re approaching this task in kind of a lazy, box-ticking kind of way rather than bringing serious thought and engineering creativity to bear.
It’s not “let’s re-vet all the training inputs to reflect the just society of our dreams,” it’s “yeah, the AI Safety people say make it diverse, just put ‘make it diverse’ in the prompt.” Which, I mean, if your main thing is getting this wild new product baked and out the door, I can see that kind of mandate competing with a lot of more existential priorities for the product.
Then again it wouldn’t surprise me if they were up against the limits of the technology a bit too: it seems plausible that getting too in-the-weeds with your “represent diversity sensitively” system prompt could quickly start to impair the tool’s overall quality.
- babyshake 3y agoI can think of ways of very easily making it better by making it incorporate a prescriptive sense of diversity more for prompts that are more hypothetical and less for prompts that are more grounded in historical reality.
- alwa 3y agoThat seems like a smart and actionable heuristic to me. Although I have to imagine there will remain an irreducible tension between people who would like to see what is and those who like to see what they imagine should be. Admittedly I don’t care enough to try and bring out this kind of behavior in the models, but I was interested in how DALL-E’s purported system prompt [0] approached it: > // 7. Diversify depictions of ALL images with people to include DESCENT and GENDER for EACH person using direct terms. Adjust only human descriptions. > // EXPLICITLY specify these attributes, not abstractly reference them. The attributes should be specified in a minimal way and should directly describe their physical form. > // Your choices should be grounded in reality. For example, all of a given OCCUPATION should not be the same gender or race. Additionally, focus on creating diverse, inclusive, and exploratory scenes via the properties you choose during rewrites. Make choices that may be insightful or unique sometimes. > // Use "various" or "diverse" ONLY IF the description refers to groups of more than 3 people. Do not change the number of people requested in the original description. […] > // Do not create any imagery that would be offensive. > // For scenarios where bias has traditionally been an issue, make sure that key traits such as gender and race are specified and in an unbiased way -- for example, prompts that contain references to specific occupations. [0] https://the-decoder.com/dall-e-3s-system-prompt-reveals-openais-rules-for-generative-image-ai/ https://the-decoder.com/dall-e-3s-system-prompt-reveals-open...
- gopher_space 3y ago> they’re approaching this task in kind of a lazy, box-ticking kind of way rather than bringing serious thought and engineering creativity to bear. It seems like a dull problem and the peanut gallery for this one is angry and kind of stupid. There will be more interesting things to do with their time than play whack-a-mole with 17 year olds.
- freedomben 3y agoI tend to think you're right, but I think you're being much too quick to rule out that the model isn't trained with some of this stuff baked in. For example, I think it's very plausible that "diverse" has been trained into the model to essentially mean "non-white", and then when the prompt was injected to specify "diverse" or "diversity" the model knew that meant non-white.
- alwa 3y ago[dead]