Unveiling the wonders of AI...with C3PO
Breaking down the neural networks and construction of AI to C3PO gangsta rapping. In a ten-minute microtraining for women in digital #Droidlife
Today I spoke to the women in digital in Lancashire. #LWID
I was delivering training on AI.
In 10 mins.
I spoke about AI, and it’s wonders, but of course, as you’ve come to expect… I didn’t try to hide the truths.
AI is like marmite, people either love it, or hate it. BUT it’s here, and it’s coming in great big waves, so lets talk.
I want them to have value in AI and to understand, because these women are innovators, they will lead the next generation businesses, and to understand the AI fundamentals; why one AI is better than another, or why things look great on the outside, but no substance on the inside, it was important to take them to a workshop by MIT and Facebook, these brains talk about it much more, but I summarise it in my post on analytics, data and intuition.
Then, of course DigiEnable, our host and founder of Lancashire Women In Digital, Liz Hardwick has integrated AI solutions to help her clients, and customers to be more productive, as well as to grow, so I thought I’d introduce the LWID group to the marvels of image AI and the “Gold Gang”
Introduced to me by my friend Wayne Kennedy, a VFX expert, we met at ILM. Years ago.
C3PO on AI (video).
This is a fun 2-min video which is excellent of showing the potential and the limits of current AI apps.
Created by Daniel Eckler and his team of 7.
This is a fantastic example of the marvels of what AI can achieve.
These digitally created pictures eliminated teams and scores of people/talent like photographers, lighting professionals, a director, talent, videographers, audio engineers - you get the picture, ahem, so it eliminates cost, but also it eliminates jobs.
Sora hadn’t come out at this point, or it did during production. Sora is further fuelling the argument AI is taking creative skills, and jobs away from people.
However, some innovators would have us believe that this could all be achieved by a simple prompt.
It cannot.
Eckler and his highly skilled team, who’d worked on several other pieces of work, stitched this together using 28 different AI and tools.
Images
Using Figma and Midjourney over a thousand images were created for this video, refining prompts and adding to the prompts to ensuring consistency; they “killed most” and used the rest.
The rough cut was in AI, and it wasn’t perfect in the least. Eckler and his team needed to work on perfecting it to give it the ‘feeling’ they were looking for.
Importantly, they added effects to “fudge the AI weirdness”, using Eyecandy.
It is so important to understand this, because we are given the proposal that using AI will make it perfect. When AI still needs the human touch so that tweaks and strangeness are eliminated.
Lyrics
The lyrics are written by ChatGPT, using multiple prompts, getting 100s of sets of lyrics. Hundreds of lyrics were generated and then to a real human going through these to whittle them down, get them punchy and ‘right’.
Eckler explains in his thread on X how he auditioned rappers from Fiverr to get the “right” British accent - a person got the job there.
Yes, it is possible to create an AI voice after training it with thousands of voice clips, such as your podcast, or your voice notes, but that voice might still not be the one you need, to say, voice a C3PO.
Celeb faces
The generation of the images went through several iterations in Midjourney, and then adding the faces of Keanu Reeves, Donald Glover and even Yoda, by using - Faceswap another Ai app, using thumbnails. They did this to a multitude of images and then got rid of any that didn’t work.
Languages
Innovation of using AI to use language and synch the mouth allows the audience to listen in their own choice of language. This is game changing, taking away the need for subtitles. Different AI apps worked better for different effects. So, it needed human oversight.
High Res, and clean up
Much of the images are cleaned up after Ai has generated them, proving you can only get part of the way there with AI. The skill and talent of Ecklers team was vital to ensure it was a seamless production.
Opportunities
In conclusion, Ai is moving fast, yes.
AI is going to be the way for the next generation, and they will be familiar with it as you are with texting your friends.
Singapore is wise then to ensure its 40+ population get retrained offering them subsidies to work alongside the newly graduated, ready to stay on in the workforce applying their newly acquired skills.
I’d say AI will change things.
Dramatically.
But, it will also hold opportunities for those of us with existing skill base and talents.
AI might take your job, but your skills will enhance whatever AI can do.
Technology must meet strategy.
Use it as a tool, to make you even better at your job, building your business, always keeping in mind the way this particular AI is constructed, it’s input values and hidden layers of neural networking, to know if its AI of value, or only has embers of potential.
Trust yourself, and always add the human touch.
Daniel Ecklers twitter/x thread detailing the entire process and specifics of AI tools used in Gold Gang here.