Monday 22 May 2023

AI & US

"God-like AI" by Midjourney

AI is beginning to scare the bejeezus out of US. Elon Musk and other “luminaries’ urged companies to slow down “God-like” AI. Approx. 300m, mostly white-collar jobs will be taken over by AI. Teachers, telemarketers, and traders are at the top of the hit list. ChatbotGPT4, the must-have AI app speaks 5,000 languages (LLMs), but hallucinates and spouts fake info.

"DABUS" by Midjourney

A St. Lois Mo inventor, Stephen Thaler named his creation DABUS (Device for the Autonomous Botsstrappping of Unified Sentience). He thinks it thinks and feels. Thaler loves DABUS. When he shuts up shop and goes home he worries it’s lonesome. Is he nuts?

Yes, that's me on actual ink and paper... -Eki

Eki tracked AI’s speed-of-light catapult into the mainstream for Aamulehti. Journalist, Markus Määttänen referred to Kierkegaard to give a 19th C perspective on the 4th Industrial Revolution. In the Economist, historian/philosopher Yuval Noah Harari writes that AI has hacked the operating system of human civilization: “Unless we regulate AI it will destroy democracy”.  Geoffrey Hinton, AI's godfather who worked with the technology for 50 years tells us “There’s danger ahead”. Hold on to your seats folks, it's gonna be a wild ride.

* The Aamulehti AI article (translated)

Note: six AI illustrations for Gen Z LUDDITES  took approx. five minutes. Eki said pre-AI, if he could draw, they would have taken about two weeks.

Sources: Aanulehti, Economist, Financial Times, Washington Post, New York Times

Next week: FINLAND: the next RIVIERA?


Note: I guess the main fear is the fear of the unknown. We've never seen something like this as a species before (unless the Neanderthals were smarter than us). But here we are, and the unknown is hitting us with the speed of, well, something really fast.

When it comes to sentience, I still maintain the same stance as I did in my 15 minutes of Aamulehti fame: it's a matter of definition. Is AI sentient in the exact same way as we humans are? Of course not. Is it able to monitor its environment, make observations, use logic, and act accordingly... sure it is. Or, something to that order. 

There are AI-enabled robots that can pretty much do the whole enchilada, but things like ChatGPT are more limited for now. This said ChatGPT has gained the new ability to surf the internet, and there are plugins for other stuff too: for example, ChatGPT itself is unreliable at math, but it can now tap into Wolfram Alpha math engine, and suddenly it outperforms more or less every human at math. 

It's also getting more and more reliable. This can actually be a bad thing, because constantly hallucinating AI keeps the users on their toes, reminding us that the information may not be accurate. But what about the near future, when AI indeed is completely correct something like 99.9% of the time? That 0.01% error rate doesn't sound like much, but it can really cause big problems if everything the AI outputs is taken at face value when we're used to it "always" providing the perfect answer.

CU
--
Eki


No comments:

Post a Comment