Dorchester — May 9th 2023

Euan’s presentation was a tour de force, the result of hours of reading and research consolidated into a very colourful and dynamic Keynote, intended to make us all start to think more broadly about AI.

He started by questioning what we mean by Intelligence, ‘clever’ is not necessarily ‘intelligent’. Nature has built in intelligence. Trees have communication networks passing information in the air as well as by fungi and roots.

ChatGPT may have brought AI to public awareness, but AI has been around for some time on our Apple devices, meaning ever larger Neural Engines on Apple Silicon. While the Large Language model of AI of which Chat GPT is an example is making the headlines, equally important is the Calculation model. Euan compared these two, with Wolfram Language, a calculable symbolic language which makes it possible to quickly express complex ideas in computational form.

Euan showed how Language model ChatGPT develops text, and provided examples of both Natural and man-made (AI) intelligence, including its use in sound, graphics and video creation. While Language model AI is close to passing the Turing test, he pointed out that the current Language model AI is very Western/English language biased and can create plausible false references (“Hallucinations”) which make it easy for malign actors to develop lies with consequence. Ethical and other assumptions need to be challenged.

Below are some links for further reading:-

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8217795/

https://futureoflife.org/open-letter/pause-giant-ai-experiments/

https://www.forbes.com/sites/lanceeliot/2023/01/05/people-using-generative-ai-chatgpt-are-instinctively-making-this-ai-rookie-mistake-a-vexing-recipe-for-ai-ethics-and-ai-law/

‘A New Approach to Computation Re-imagines Artificial Intelligence’ Quanta, Anil Ananthaswamy. April 13, 2023

‘Recalibrating Assumptions on AI’; Arthur Holland Michel — Chatham House Research Paper

Comments

Page 1