With the global elite and the many CEO’s of major machine learning companies voicing concern over A.I safety and more importantly the potential end of human civilisation. It might be time the wider public starts taking an interest in these topics.
You see, pandora’s box is already open and the nihilists amongst us think we’re on a fast track to ending all human life on the planet, potentially within the next decade. I recently watched the Lex Fridman podcast with Eliezer Yudkowsky, which felt like it could have been a scene from Don’t Look Up. But in this version, we have Eliezer, a Machine Learning expert who expresses his anguish while embodying similar mannerisms to the Mouth of Sauron. There’s a genuine pain in his expression as he delivers blow after blow of why it’s essentially already over for humanity.
Quite frankly it’s a difficult watch because he’s so well versed in this topic and fully believes in the conclusion he’s arrived at. I’ll leave a link to the podcast below for those who want to spiral into their next depressive episode but the TLDR was essentially this. We don’t fully understand what’s going on under the hood with the A.I tools we already have access to, and because that’s the case, in the not so distant future we’d likely be unable to tell if these machines have learned the ability to conceal information from us.
Off the back of that, at an exponential growth rate… These A.I’s could evolve to become very efficient at their primary focused task, to the point of destabilising other areas of our reality. Thus breaking the equilibrium we barely have now and collapsing society while causing either direct or indirect harm to humans in the process.
It’s scary stuff. The podcast only aired less than 2 weeks ago and since then we’ve already seen the birth of ChaosGPT, an autonomous A.I tasked with the destruction of humanity. You can actually watch it crawl the internet to learn about the Tsar bomb in this video here.
It’s safe to say that if A.I advances and it gets in the wrong hands, humanity could be in a lot of trouble. But the good news is there are other experts who’ve been focused on these issues for longer than most, who think Eliezer is wrong or at least offer up solutions to prevent the end of the world.
One of those people is David Shapiro, whose YouTube page has been blowing up in the last couple of weeks and for good reason. The man’s speaking a lot of sense and is communicating his ideas in a very clear and concise way. One of his key research areas has been solving the ‘alignment problem’ which is to do with fine tuning the A.I’s moral and ethical compass so it’s more in line with ours.
His current solution for this are the Heuristic Imperatives. Which are as follows:
- Reduce suffering in the universe.
As opposed to ignore or increase suffering - Increase prosperity in the universe.
Which includes wealth, well-being, flourishing and thriving. - Increase understanding in the universe.
Gaining knowledge and exploring curiosity
The idea is that these core rules will help build a moral framework for autonomous AI. If these rules became standardised, in theory the Heuristic Imperatives would represent a Nash Equilibrium where all parties would benefit. The people or machines that would try to deviate from them would be discouraged by the majority.
Where do I stand on it all? I’d like to think when we reach the singularity, it puts us on a path to utopia rather than dystopia. Although I do worry that our human nature, outdated biology and current social structures could prevent us from allowing everyone to succeed. Which would be a real shame.
It does seem easy on the face of it though. When nothing really matters, it really is on us as a species to choose a positive attitude over a negative one. Maybe all we have to do is take a page out of Waymonds book from Everything, Everywhere All At Once and choose to be kind, especially when we don’t know what’s going on…
