Extended Intelligence

Plotting Our Model

Throughout the week we met and discussed which possible forms of extended intelligence could be used. As we choice to work within our intervention group we wanted to focus it on either energy of food waste. For this class we choice to see how we could create and use an AI to take pictures of food and then create cards that give the user a step by step how to repurpose their waste. The first step we did once we formed our idea was to plot all the aspects of our model in the figure below. This was a bit of a challenge as we needed to consider all the elements we would ideally use to create the model and how they were connected. This led to some interesting conversations and also in the end forced us to make some choices for the simple code demo.

Example of food waste identifying dataset which ideally would be used to train a new model if building from scratch.

Researching Datasets & APIs

This part was quite interesting for me. It led me to realized how many existing neural networks and models already exist that you can use as a designer. But it also showed me the limitations of them. When I was coding the AI it really hit me that with the APIs that require a prompt the prompt must be relatively specific so that the output is what you desire. This led me to thinking of the benefits of creating your own model training it through datasets that already out there, but even then with food waste the existing datasets tended to either identify waste to generally or there were not enough to train a model. I feel that if in the future I were to continue with this project it would require me to train a new model using a created datasets. However the downside of this for this project is the time required and the outputted result. For that reason we choice to go with existing APIs.

Coding the Functions

Before even starting the coding process I decided to together with my group go through the various steps that needed to be completed to reach the desired output for the user. We did this in such a way that there were four main steps that the AI needed to do. These were as followed:

Step 1: Get input image from user and output food waste spotted.

Step 2: Take the various ingredients and search for most relevant repurpose method.

Step 3: Create a generated image based on the repurposing method.

Step 4: Bring the outputs of step 2 and step 3 together and show the user a printed card with image and reuse together.

Once these had been determined we searched for relevant APIs and tweaked them for them to work within our contexts. I then set out to write the code and bring the final AI to life.

Reflecting on Extended Intelligence

For me this was one of the most interesting weeks in term one. As I started to learn to code in the past summer it was so much fun and interesting how you could grow so much with the help of using APIs and finding interesting ways to create new AIs. It taught me a lot about the various problems that one faces when exploring new technologies. From the logistics of coding the system and training the model to seeing how you can avoid biases.

I must also say that as AI is rapidly growing and changing it is hard to know it all. I feel that as a designer I need to not only figure out if/how extended intelligence might play a role in my designs but also what the implications of this might be to others. We learned in the past course how to design within communities and that is something I strongly believe in, yet with Extended Intelligence you are training models off existing data which has its problems. You are not using the connection to the people nowadays but past data, and the world, especially in the past, has had various negative biases that are now embedded in our data. Take for example the AI exhibit we visited last week. Here we saw various examples of how biased data led to more biased results with different search prompts. This was not as big of a problem within the group project I did during this course but in the future if I want to use extended intelligence within communities I must take this into account and find ways to work against this.

With our project in mind I feel that I grew a lot with my knowledge in such a short time. I learned the basics of coding this past summer and since starting MDEF have challenged myself to grow and learn as much within this field as I can. Coding is something that I feel gives me freedom to express myself creatively but also create new systems that can simplify problems for others. I really like that during this course we were given the ability to create systems using APIs as it gave more possibilities then trying to train a model from scratch in such a short amount of time. In regards to what we designed I feel that we were trying to give people a new way of looking at their foodscraps and allowing them to try and repurpose it in a new way. With the simple code we made we used extended intelligence for many parts, however like discussed during the presentations if this were taken further I feel that some things could be done through regular code. This is something I feel might be important as while AI is super interesting, the amount of energy needed to train the models and run the system is great. Next to that the water used to cool the system as of today is quite high. This is something that I am considering when designing a system as you want to facilitate new experiences but you also must try and find a balance between ordinary coding versus AI to create the most fluid and sustainable system for the future.

All in all, this course gave me the tools needed to start exploring the creation of new intelligent systems while also opening my eyes to the ethical and sustainability problems that can arise by using AI. I feel that in the future I will need to learn how I can use these systems in a balanced way in my designs. In the way the world is growing there is no way around the fact, that AI will be a part of the designs of the future but hopefully I can do it in a way that fits in with my design principals.