GeekTechStuff

This week has seen several YouTube videos on Artificial Intelligence (AI) that caught my attention and, more importantly, managed to keep it. So I thought I would try something a little different and do a blog post sharing the videos and my key take away from each one.

This is not the AI we were promised – Professor Michael Wooldridge / The Royal Society

Professor Wooldridge discusses how Large Language Model (LLM) based AI works. This includes the positives, the negatives and how illogical LLM based AI can be. The jump from “state of the art” in 2016, where AI misidentified Van Gough’s Starry Night, to “state of the art” in 2020 where AI is answering 5/8 common sense based questions correctly, to “state of the art” 2024 where it is answering high end Maths and writing out a page of why the answer is correct, shows how far AI has come in recent years.

My key take away: LLM AI is not implementing an algorithm to solve a problem, it is pattern matching. LLMs are incapable of distinguishing facts.

AI Isn’t as Powerful as We Think – Professor Hannah Fry / New Scientist

Professor Fry has a new three part documentary around AI coming to the BBC at the end of February. This (roughly) 25 minute discussion with New Scientist discusses how some people have given up their job, ended human relationships or lost money due to AI. Its not all doom and gloom though, the positives of AI helping with mathematical breakthroughs. I liked the analogy of humans creating tools to complete tasks that once seemed improbable, and that does not make the tools godlike.

My key take away: We need more public discussion around AI and more conversation about the designs of systems (such as chat bots) that utilise AI.

Bad Bot Problem – Lewis Stuart / Computerphile

Following a report that 37% of all internet traffic being bot driven, specifically bad bots, Lewis delves into what bots are and how a bot net with sock puppet accounts can generate fake social media posts whilst driving interactions. AI is used to create more realistic social media posts, comments and responses. As well as AI this video touches on the Dead Internet Theory (https://en.wikipedia.org/wiki/Dead_Internet_theory) which has grown in interest in recent years.

My key take away: Using “War of the Worlds” on social media via bots could actually gain traction and see people believe it due to fake posts / fake interactions, just as those 1930’s believed Orson Welles radio broadcast.