V8 is coming. Midjourney asked to 'borrow your aesthetic sense' #264

I am KITAcore, a self-proclaimed and widely recognized experimenter without literary talent.

Hey everyone using Midjourney, did you notice the message from the official team saying 'Can you help us put the finishing touches on V8?'?

I, for one, didn't notice for three days.

Wait, was it just me? ...Probably not, right? lol
Come on, confess. You thought, 'What's that!?', didn't you?

But if you read this carefully, it's actually a pretty big deal.
It's about how user 'preferences' are being used for the final polish of V8.

Today's article is a summary of the key points.

Here is all you need to know

• Go to the official 'V8 image evaluation (pair ranking)' page

• Two images will appear → choose the one that is more 'beautiful' (your own preference is fine)

• If you're unsure, skipping is also fine

• Sometimes NSFW content might be mixed in, in that case, Report is recommended

(I've included the link at the end!)

What the official message is trying to say

Midjourney is saying, 'We are doing the final adjustments for V8. We want to borrow your aesthetic sense for the finishing touches.'

There are two important things here.

1. The images you see now are not the full power of V8
   . It seems they are intentionally mixing in 'plain,' 'boring,' or 'subtle' images. The reason is simple:
   'knowing what doesn't resonate' is also necessary for learning.

2. They are injecting user sensibility at the very end
   . The definition of 'beautiful' varies from person to person, right? The structure is that they are gathering those opinions 'at the very last minute' to refine the direction of V8.
   

What is the AI learning from this?

To put it in technical terms, it's 'adjusting with human preference data'.

But, to put it simply, here is the deal.

• Choosing between two options = accumulating 'preference comparison data'

• The more comparison data there is, the easier it becomes for the AI to grasp 'what this community considers good'

• As a result, the model's output shifts toward 'human aesthetic sense'

The '1' or '2' we click becomes a vote that determines the 'standard beauty' of the future.

Benefit of participating: It works well with --p (Personalize)

This is the best part.

Midjourney has Personalize (--p).
Roughly speaking, it's a feature that adjusts the 'filler sense' to match your preferences.

According to what the official team has stated,

• Personalize learns from pair ranking votes and ❤️ in Explore

• As a guideline, it starts to take effect after about 200 votes/ratings

• You can use it by adding --p, and you can also adjust the strength (change the effect with --s)

Note: It cannot be determined from this announcement alone how this 'rank-v8' will be reflected in your --p count.
However, since the specification that 'pair ranking votes are the source of learning' is official, those who are concerned can check their count on the Personalize page or via /info in Discord for peace of mind.

A quick warning: Possibility of NSFW content

The official team also says, 'Since these are random community prompts, NSFW content may occasionally appear. If you see it, please report it.'
If you don't like surprises, choose your time and place carefully.

Midjourney users, let's cooperate.

AI is not magic. It evolves through
the accumulation of data.

Our sensibilities are being used to refine the final 1mm of V8.

Click away in your spare time.
Just by doing that, the next 'standard of beauty' might become a little better.

Alright, let's participate.

Access link

https://www.midjourney.com/rank-v8

If you think, "I've already done it!", "I missed that", or "That was helpful!", I'd be happy if you could
like or follow me.

For comments, just "do it / don't do it" is fine.

If you find it hard to comment, even just a "." helps me know you've read it.

See you later.

#Midjourney #GenerativeAI #AIArt #ImageGeneration #Prompt #Creator #ToolUsage #ExperimentLog #Learning #ProductionLog