THE FASHION TECH BRIEFING
Without getting too meta… AI is Human after all… Isn't it?
Newsletter #28 | Read time • 3 mins
Share this Article on:
I was thinking about my day to day. I’m creating, writing (more and more!), questioning continually as I work on Aistetic. Startup life. As part of this, I use AI tools to challenge myself and my thinking. It got me thinking about what the starting point for AI is and the tools that are so prevalent today: LLMs, stable diffusion models and what we do as a company. They are both artificial but also fundamentally human. Or are they?
Reflecting our Intelligence
Take a deep learning model that is analysing imagery. The algorithms that power these AI systems are designed by human programmers, trained on human-generated data, and aimed at solving human problems. These models are a reflection of our minds - our programming, our data, what we put in, what we don’t put in, the mistakes or flaws are human. In much the same way that I reach a conclusion based on the available data that I have to hand, a deep learning algorithm reaches a conclusion based on the data that it is shown. Any bias or prejudices in that training data, will show up in the end conclusions and results. The models are reflecting our intelligence, our creativity…or in the worst cases they reflect the narrowness of our thinking, highlighting the limitations.
How a child learns
The actual process of learning is not dissimilar from how a child learns: it’s trial and error. Performance and learning is gradual, over a period of time. It’s repetition, with positive reinforcement facilitating progress and knowledge acquisition. This mirrors the way a child learns - through trying, experiencing and then acquiring that knowledge. Neural networks are designed to work in this way. Our systems and models at Aistetic are no different. We take large volumes of data both imagery and specific inputs (text) and we hone our models, to try, to learn and to acquire knowledge.
Creatively Fallible
What also makes AI interesting is its creativity and fallibility. When it gives you something that is not predictable, it is surprising. It's not always controllable…Distinctly human you can argue? I asked Firefly to create an image for the title of this week’s post and it gave me what you see above. Not what I was expecting. I did not ask or envision this image - I was expecting an AI reference, something obvious or predictable.
Personal Ethics
The ethical issues that AI presents are those that face us everyday. The trolley problem which explores the ethical dilemma of whether to sacrifice one person to save a larger number when faced with a life threatening situation is something that driverless cars now face. Should the AI learn its personalised ethics from the driver of the car and act like that human? Or should the ethics follow from societal norms? A human ethical issue. Interestingly, the German government looked at this in 2016 and found that personalised ethical standardised rules are not allowed - systems should follow society's norms.
The Ultimate Co-Pilot Ready to be Directed
Finally, the strongest case is how we work with it. Its potential is to be the copilot, the symbiotic partner to our human intelligence. AI might help transform the world. But we will be the ones leading it as it does so. As AI researcher Kai-Fu Lee suggests, "AI is going to change the world, but human beings are going to be the ones directing that change."
AI is born from our human ingenuity. It's shaped by our knowledge and our experiences. And it will continue to be shaped in this way. It mirrors our minds, complete with its strengths and limitations.
As a result, AI will hopefully help us better understand ourselves in the process.
And that can only be a good thing.
--
PS. When you are ready to dive into some AI-powered fashion tech, please do check out our AI Sizing, AI Listing and reach out for a chat.
Share this Article on: