Should We Slow Down AI Progress?

Humans are driven by profit and power - look at the lack of controls on GENE modification, not just of specific genes, but on large areas of genomes. What happened with Covid-19 [accident or not] is an EXPECTED outcome of careless or even reasonably careful genetic modification. So the question may very well be -- "Which of these two or more technologies are or will be the genie out of the bottle? -- remembering this question from the 1950's with knowledge of the atom and potential destruction of the earth - or -- the lack of rationale and reasonable use of the planets limited, highly interactive, and small planet with unknown environmental switches for radical change. Human beings generally do not protect the larger environment. One question we may need to answer is should the government of the earth be a dictatorship structure to limit the damage we are doing.

My (uninformed) opinion. I think about the weakness of the internet – when stuff is free, then their product is you – trying to figure out how to manipulate you to make decisions that benefit them. So, all this AI stuff on our phones, PC’s, social media is ultimately designed to benefit the suppliers of the AI and not its consumers. I think that greed will drive the design of AI that hides it’s true goals. There is AI that helps doctors or engineers, or similar professionals evaluate data to find problems that can be solved. I’m more comfortable with these types of AI. But the ones write scripts, or on your phone to guide your decisions are one that cause me concerns.

Living in Microsoft territory and having a background in what can be done with little data as well as editing and writing more than a few things on coding in my life, I'm justifiably terrified at what is called AI is dong, not what it might do. I'm watching authors have their materials stolen, dumbed down and stolen, I see targeting of humans by what are called AI procedures shielded from analysis or challenge that are inherently racist in outcomes, and most recently, a friend of mine finished managing a multi million dollar data project for a really big software company. The company fired her and outsourced the entire thing to AI and an overseas company for six dollars an hour, where they need employees. Let me remind everyone here that one of the SETI Institute early promotional pieces some decades ago was built on assurances they would never be stupid enough to allow any piece of code or software from an outside alien source to be allowed to run on any computer, let alone make decisions that affected society, unless that code was pulled apart, fully understood and considered at length. Well, here we are, and a group of wealthy eugenicists with very questionable associations who build private islands and want to establish their own countries on Mars are now pushing AI out as fast as possible. https://youtu.be/AtgJhZOhFsQ

I have to say, I watched the entire interview and felt like Dr. Yampolskiy's arguments against AI progress were very weak, generalised and sensationalist. He actually ended up moving my opinion from somewhat center to one that we should continue progress at the current rate. Maybe I'm on my own on that? I'd be interested in an interview with someone who can speak for the opposite view, including areas such as advancing the sciences (in particular medical sciences), improving world logistics, caring for the environment and solving political deadlocks. btw, outstanding question at 41:44 because that's pretty much where my feelings are. (disclosure: I am probably a little biased as I'm a software engineer - not in AI - and also someone with a significant incurable medical condition)

It may be incorrect to think of LLMs as "smarter" than the average human. They have access to vast amounts of information and they are very good at linking it together in ways that humans (represented by its inputs) already have, but there is no understanding of what it means to link that information or how to act intentionally with it. (That doesn't mean something awful might not happen unintentionally if the AI is "allowed" to invoke external processes based on what it synthesizes.)