SYSTEM NOTICE

Auto translation by AI. Be sure, accuracy, nuances and authorial intent may not be fully reflected.
見出し画像

[Manufacturing] x [A.I.]. Images in your head made easy. We live in an amazing era.

I work in the jewelry industry,
and I also handle jewelry product development.

Regarding that product development,

for the past few years, I have been following a process of product development leading to crowdfunding, repeating a few successes and many failures every year.

And the flow of product development looks like this:
➡ In the case of jewelry

1) An idea comes to mind

2) Solidify the image somewhat

3) Request a data craftsman to create a prototype

4) Sample completion

5) Crowdfunding preparation

6) Crowdfunding project start

And
that is the flow,

and last year I did over 70 such projects,
and I don't think there is any other jewelry company doing more than one a week for crowdfunding,

so even so, it is incredibly fast,

but still, ideas are overflowing.

Of course, not all ideas are commercialized, and there are many strange ones, but

anyway,
I am suffering from
the "disease of wanting to release ideas to the world."

◆ Before meeting A.I.

About four years ago, I tried to look for someone to support this "disease of wanting to release ideas to the world"

on the business matching site Lancers,
and I posted a job opening for "a job to turn the things in my head into documents for one hour every day,"

and because it was during the COVID-19 pandemic,

I received messages almost every day from people who saw the job posting, and I actually worked with about 10 people, but turning what is in my head into documents requires organizing and recording the images and mechanisms that come to mind at that moment,

so is it better to have someone good at shorthand?
Is it better to have someone who can come up with ideas together?
Is it better to have someone who can write wonderful sentences?

I often find that the content ends up being tailored to the other person, and

it's quite difficult to quickly and easily put the images in my head into a document.
It's hard to quickly and easily capture the images in my head into a document.

In the end,

it just became a mess,

and although I tried for a while,
I eventually gave up.

-> By the way, this job posting is still up on Lancers, and I get a few applicants a year, so I sometimes ask them to do similar work that isn't exactly this job.

◆Meeting A.I.

I have written tons of such ideas from my head as bullet points in an idea list on an Excel sheet, but

when it comes to actually saying, 'Let's do this,'
it's hard to get motivated just by bullet points.

While I was doing that,
the era of 'that thing' arrived in 2023.

Yes, it's 'A.I.'

Specifically, it's ChatGPT,

and the first time I learned about it was
maybe on YouTube,

where I saw the topic that about 30% of American college students
are using ChatGPT for their homework,

so I immediately
tried using it.

Oh,
it's true.

Just giving it a prompt isn't much different from a search, but
it converts what you searched for into the text you are looking for.

And since its Japanese is perfect,
while I can't say it's 100%, if you make a few corrections,
it creates text that you can use.

However, it doesn't produce human-like text,
so that seems to be difficult,

but for things like researching,
creating explanatory text you need,
having it decide on a title,
or having it come up with new ideas,

it is extremely useful for those kinds of uses.
Previously, I used to hire people on Lancers or Coconala to research and write text about the background of products for product development.

When you outsource, as is often the case elsewhere, there are hits and misses, and if you don't convey the content you want written in detail, the results are often lackluster.

Such
frustration and cost-cutting
were resolved all at once and with ease

by CHAT.

◆ Images from imagination

And there is one more thing:
a program that can create "that" image you have in mind has started.

That is the image creation program called "DALL-E" which has been newly added to ChatGPT.

Image creation
has existed in other A.I.s as well,

and I have actually used Canva's image creation a few times, but
it had too much personality, and honestly, it wasn't something I could use.

Meanwhile,
"DALL-E" easily produces tasteful images for you.

By the way, ChatGPT-4 can also produce images, but
it only does one at a time,

whereas DALL-E can actually produce two at once.

With simple words, you get tasteful images.

This is shocking.

As a test,

I will have it create an image that says "An A.I. that can produce two images is very amazing."

Both images were a bit stiff, so when I asked for something more relatable,

this is what was created.

The image concept is a bit lackluster, but since the art is of professional quality, it's easy to use.

And when I asked it to create a character for this theme,

this is what was made,
and when I asked to make it cute,

This is how it turned out.
And
when I asked it to make that into "pure gold," which is one of our businesses,
it did this.

Something like this.
Furthermore,
when I asked it to make it into a card game I've been developing recently,

it turned out like this.

As you may have already noticed,
this A.I.

has an incredibly high capacity for "cuteness."

It ends up with a style that is very much to the Japanese taste.

Of course, there are ones that aren't cute,
and sometimes images come out that are completely different from what I intended.

And for some reason,

this A.I.
is incredibly stubborn,

so when I try to modify the initial image,
it ends up looking completely different.

For example,
if I get a good image,
and I ask it to use that as a reference, for instance, to have a dog playing baseball,
even if I say to keep the dog the same but change it to soccer, it will often give me a different dog instead of the same one.

Yes, it won't refine an image once it has been generated.

I assume this will be updated
at some point.

Also, there are limits if you use it too many times.

So, it ends up being a pace of about 30 images a day.

Even so,
the performance of this A.I.

is sufficient
for expressing the images in my head.

There are even some that I am considering using as-is
to release as a card game (currently in product development).

All of these images were created by A.I.
Some of them are impossible, but since it's a collaboration with A.I., I've decided to use them as they are.

Besides that,
there are times when I have the A.I. write parts of the text as well.

It really feels like having a wonderful partner.

Ah

Partner = Aibou
A.I.

I intend to continue the product development of jewelry and card games that I am working on in collaboration with such an A.I.
until my head is no longer overflowing with ideas.

Finally, as with the top image, I will post a few pictures of the cutest pure gold items created by A.I.

I asked for a kappa, but I wonder if it's good.
The texture is just like the real thing.
It looks like it would sell if I made it just like this.