HumansAI.com NormanMacrae.net AIGames.solar EconomistDiary.com Abedmooc.com
What happened in 2012 blows my mind - 40 years after getting my MA in Stats at Cambridge maths lab DAMTP it turns out that everything we coded on computers was so inefficient to be practically worthless regarding sustainability of millennials. Awkwardly this also explains many conflicts top-down decision makers are drowning us all in. Turns out that computing efficiency begins with the biggest structure databank you dare invest computer time generating. For example if you tried to get a computer to recognise the 100 entities humans do most valuable work with , it might not be possible and it would certainly be very costly each new 100 entities you needed computers to work with.
what happened in 2012- after about a decade of work as a graduate, and 3 years of staging a global coopetition Dr FF Li demonstrated that computers can very economically train on recognising the 20000 entities we earhllings work with most provided we3 prepare a 20 million image databank - that's 1000 labeled (supervised learning) images per entity so as to make sure the computer doesnt think eg you are encountering a domestic cat when actually a Lion is about to devour you.
There are several parts to celebrating Dr Li's work as arguably the greatest maths breakthrough ever. Yes she needed other brilliant mathematicians to develop optimum algorithms at the heart of AI - shockingly the best algorithm had been known to Neumann-Einstein-Turing but almost discarded over half a century except by all but 3 Professors and alumni who are trilingual english, french and maths code.
Secondly it took hi-tech world to see Li's breakthrough before the work of the other greatest 21st c h=graduate Hassabis was valued. In his case computers now let us work with 250 million proteins - a biotech advance that might have taken a billion years of human work without Hassabis Deep Science AI
Although at www.aigames.solar we try to help celebrate anyone helping millennials be first sd generation; it does help to understand which huge investors did what first
Melinda gates and Nvidia ceo were first to sponsor dr Li to produce 10 by 8 hour courses showing that kids of any age could join in - well as soon as a kid learns to draw or dance- or as soon as a pre-adolescent girl decides women can be as great at engineering or maths as men. Oddly most us education systems havent started making use of this gift yet but giving fei-fei li her own ngo brand proved jolly useful. https://www.c-span.org/video/?447599-1/artificial-intelligence
It was her open ngo that first politely told congress in 2018 that ai alogorithms advnacing hunanity around the world couldnt be stopped- or at least all the biggest digital companies on the west coast couldnt be stopped from a race to chat. Now what these chats will do for us -in my controversial opinion may net be wh=oly understood to 2035. Many disagree but i see llms as h=gatewats to web3, metaverse virula reality constructs being only one of many interlinking ai games
Of all the games intellugence links in, my facsination for 50 years now is how it transforms education . For those wanting to help with that I recommend unserstanding 2012's breakthroughs and suporters before choosong who is the greatest caht guru of us all. After all the ai netwirks of ki and hassabis can see what the computer was pretrained on and collaboratively build ever more good wirk with the 2000 entities humans work with and the 250 million nature works with. In comp0arison quite what diferent chats are programed on seems to be top secret.Now i must admit a boas here- my family has been one of teh biographers of teh origian net neumann einstein turing since 1951. In there last 7 years they made it very clear that they hoped the first top secret apps of coding ie nuclear bomb rane would be teh last. For the simple reason that top secrets conflict with being transoarent oublic servants let alone trustworthy media operators. - see world class brand literatire which inpioneered from 1988 with The Economist if you want to see dig=ferences between ever more scary fake media and transprently good media.
Back to Economitlearning.com we are working on caraloguing nearly 365 good ai people or event every year from start of 2022; this needs one heck of a lot of cooperation - and has been one of the outpouts of year 1 of piloting AIgames. Some samples of good ai celebration every new dwan are at Economistlearning.com
By 2014 team google - probabkly Dean Pucjhia ans well as Page Brin had deceided Hassabis was ootentially the most valable trainer of natyral sxeince algorithms