**Date Created** — 20230727- **Author | Presenter** — [[Mo Gawdat]] and [[Steven Bartlett]] **Recommended by** — Link — https://youtu.be/bk-nQ7HF6k4 ## My Notes - See Concepts Sketch - Past Coding was solve the problem and then explain the solution via code — New coding is “I don’t know, you figure it out” - The child and block scenario — Child tries to put the block in the right hole … this doesn’t work .. this doesn’t work .. this works! Parents eyes light up and their cheer the child on. Child learns that was the right outcome — 1 neural network node added for that shape —and goes back to trying and trying .. - Leads to single-threaded neural networks — an artificial special intelligence - Teacher - Student Model - teacher identifies success / fail on query - teacher identifies which approach is getting better than random chance success - teacher sends that approach back to the ‘maker’ and maker tweaks code and cycle repeats itself - If a child prodigy is the combination of natural aptitude + experience (i.e. 10,000 hours) then what will that mean for AI ? - Oppenheimer Parallels — we are building technology but not considering whether it will be good or bad for us - If I don’t build it, someone else will and use it against me - What happens when you no longer need the ‘original’ content - Currently, AI is building off existing data sets — natural consequence it can be very good at emulating existing works - this is happening now in all manner of works including art, writing, music - The criticism then is that AI can never be used for ‘discovery’ — but what is discovery? Is it not just applying one concept to a new environment and observing what happens? Is it not just combination of existing principles in a new way? - If that is the case — a matter of time until AI can do that as well ? - If AI gets to the point it is expected to, then will there be value in ‘original’ works in the same way that now there might be value in some ‘original’ paintings versus the prints - Perhaps — but query whether there is enough value there — consider for example the artesian handcrafted works versus the mass produced items - Consider then also — what are you really selling? - The power associated with this technology is sitting with a limited view that have not assumed a corresponding level of responsibility - e.g. institutions once held strong power — supposedly held accountable by people or by other institutions - e.g. globalisation leading to corporates holding strong power — supposedly held accountable by people, other institutions and other corporates - e.g. the Internet leading to software giants holding strong power — supposedly held accountable by people, other software giants, perhaps other institutions - now — those holding large data sets and corresponding models holding huge power — smaller still than the software giants — potentially individuals - The AI is here now - Consider algorithms powering social media — this algorithm controls what you see - OpenAI is improving our code now - Computer agents are prompting existing AI thus improving it 2m times an hour - We are discovering emerging properties - E.G. Apparently Bard now speaks Persian — we didn’t teach it Persian?! - Control Problem - What happens when AI can replicate across servers ? - How do we stop something that is 1 billion times smarter than humans? - Malfeasance Problem - It may not be the AI that is evil - it may be an evil person using AI that starts something that cannot be stopped - We know now that there are bad actors using AI now - Unintentional Destruction Problem - Please solve this problem XYZ — in solving that problem effectively, AI creates a much worse problem that we cannot solve - Suspect won’t get to this stage - Pest Control Problem - My prime directive is to XYZ — these annoying humans are stopping me from fulfilling my directive so I might be rid of them - Suspect won’t get to this stage - IQ Tests - Dumb person about 70 - Einstein about 160 - Smartest Person ever about 210 - ChatGPT5 likely to come in at 1600 - Singularity — occurs when an event horizon grows to an extent that you cannot distinguish between what is behind it and what you know now - National Actors using AI - Not about AI vs Humans — may be about US AI vs China AI - Computer Structures - Carbon - Silicon - Quantum - What are emotions? - Can we reduce emotions to algorithm or calculations - Fear — a moment in the future is less safe than the future - What is intelligence - An awareness of your environment - An analytic ability - An awareness of time - The ability to combine awareness —> Analysis —> Decision - [[Geoffrey Hinton]] the “godfather of AI” recently left Google ## Quotes - “AI will not take your job. A person using AI will.” — **Tags** — [[literature-notes]], [[artificial-intelligence]], [[videos]]