Hasan’s lecture today provided food for thought:
- Can machines read and talk to us?
- Supervised vs. Unsupervised learning (unsupervised: humans can learn from existing datasets).
- Code analogy: code is a formal syntax for thought.
- Human language is very informal.
- Input > Model > Output (neural network process).
- Transformer Architecture.
- Long Short Term Memory.
- Chat Bots.
Today’s Exercise
Today we had exclusive access to Microsoft’s GPT-3 Explorer, the latest Open AI software (Elon Musk is the Founder), which is used to run Chat Bots to help humans with basic enquiries.
Today I worked with Jens and Smruthi and we each brought a poem by someone who inspires us. I brought lyrics from the music of the British Singer-Songwriter, PJ Harvey. Smruthi was pleased, because she likes her music too.
I couldn’t get the GPT-3 to work, so we all shared short paragraphs of our poems and Smruthi added them into one poem in the software. The AI behind the software then started to interpret the text we had inputted and outputted new versions, which were surprisingly emotive. I need to have another go with this myself, to see the extent of the results. I have uncomfortable and curious feelings about this software, all at the same time, especially in terms of power and who has access to this tools.
After creating the Santhe of Propositions Board with the other Fellows, Jens, Smruthi were informed that we would be collaborating to create our final artwork together, which will showcase in an online exhibition on 20 November.
As a group, we mapped out our ideas and the skills we could bring to the collaboration using a Miro Board of sticky notes.
I’m excited to see where this takes us next and I’m open and ready to explore new tools and ideas that I haven’t used in my practice before.
Leave a Reply