“I am worried about is that as humans are outpaced and rendered obsolete in the most parts of cognitive labour, the incentives to maintain the institutions and practices that strengthen our minds so that we can even have the imagined intent, taste and capacity to wisely steer vastly more intelligent machines, will fade.
A myriad of questions seem unanswered to me: Where will the human’s purpose come from? Through what experience will his taste have matured? What will enable him to recognise that the advice he receives is mistaken, or that the goal he has chosen is unworthy of pursuit?”
This is from the new piece by Ruxandra Teslo “Education as an AI safety area” which I read eagerly, just days after the philosopher Anastasia Berg was on my podcast The Hope Axis to discuss what her classroom experiences at UC Irvine have been during the LLM disruption. Podcast episode coming soon. (See her earlier Interintellect salon — or debate — with Hollis Robbins about AI and education below:)
I have also just hosted the Keynes episode of my Interintellect salon series “Reading Economics Together: From Smith to Ostrom in 5 Texts” — which was born out of my personal interest to read and discuss these texts, and my desire to probe the big questions around agency and automation with the salons’ attendees. What does it mean to live in a society? How do we know what to do?
Recently, I’ve been feeling that every Interintellect salon every day, in fact every conversation I seem to be having anywhere, revolves around the same anxiety. Something is happening to humans that is beyond the human scale, and we can only figure out what to do with it together. We can only deal with it together. We can only know what it is — and what we are — together.
And Ruxandra seems to understand this too, of course. And her precise and concerned essay has sent me down a rabbit hole…
The value of humans
During the past 100 years, two very interesting things happened to cognitive collaboration.
On the one hand, our education and tools have enabled us to do more and more intellectual and creative work alone. Today, if you really want to, you can learn something from scratch, research independently, write, edit, and get published without ever interacting with the outside world, society, beyond sending a few emails. You don't even have to talk to anyone.
On the other hand, important institutions like academia, science, the media, etc., have become robust, mature, complicated, and expensive industries in which real breakthroughs are increasingly hard to achieve alone. Much has been written about a need, for example, for more group prizes -- do individual scientific Nobels even make sense anymore? How many people are needed around a hadron collider...? How much money, people, and clout do you need to make a blockbuster? What does it take to build an Anthropic? (You'd think the internet has softened this capital intensity, but that is only to some degree: the core of these machineries have become more complex, not less.)
Ruxandra intuits something I have been musing about for the past years. After this long process of simultaneous technological "atomization" AND the mass efforts needed for real human innovation, we are entering an era where AI will be well-suited to take over some of BOTH:
- We now outsource much of our lonely inquiry to AI, and it can make it even lonelier, maybe now even those few emails won't be sent
- AI makes some of the innovative mass effort unnecessary because it can simplify or at least imitate what human collaboration could achieve, and this will make humans even less eager to spend such resources on it
Like me, Ruxandra and others have given serious thought to how AI might perform at different cognitive tasks. Should humans now “specialize” to some types of thinking, the types we are better at, and deliberately hand over other forms of mental activity, the stuff we’re not good at? This seems to be simply how technological progress works: what is too hard for our bodies and minds we get machines to do for us.
Every once in a while, humans invent cognitive tools — writing, printing, computers — that prompt us to rethink what thinking even is. This might be the first time when we are prompted to think together about thinking together.
Thanks to my job at Interintellect, I have had the privilege of watching people converse and perform group inquiry for eight years. The surprising results that come out of such exchanges -- the debates, the insights, the unexpected turns that arise from the novelty and clash -- are not something I have experienced interacting solo with AI, even when AI is used truly well. But it is something people I know who work on seminars, societies, retreats, and debates report to me regularly. Machines still can’t surprise you beyond “look, it can do that!” Humans will always surprise you: wow, she said that?!
There is a type of instigation and inspiration that somehow only takes place between human minds. Experience versus experience. Will against will. The human mystery is more exciting than any AI black box. The willingness in humans to expose their mystery in conversation may be the most mysterious of all. We flex with our shocking sentences. Compete with our originality. We seek to say what no one has said and arrive at fresh conclusions. We worry about AI jailbreak, while humans jailbreak in every conversation.
We bring our arcane experiences, conflicts and thoughts together in peace, and we leave with something new, every time. (What if, by the way, AI could help with this? Would it not be a great place for it in our lives?)
We are about to discover that if, as a society and as individuals, we wish to keep our cognitive autonomy — our creativity, our free associations with each other, our deep understanding, our key decision-making, and our new insights — we need to seek out other people for thinking together.
We will come together more, push each other more, learn from each other more, real-time, and in novel formats. We must. There must be spaces for this and builders for this. The "forum" has never been more important. Crisis requires conversation. Perplexing paradigm shifts require institutions to give us time and opportunity to ask, discover, evolve, and solve problems.
I believe that we will only remain humans together. Thinkers together. Builders together. The human world can only be maintained together. There need to be new places, new people, and new institutions dedicated to this.
Thank you, Ruxandra.
You are absolutely right that as humans we develop ourselves best by scrubbing our ideas against each other. Anytime I do this, while keeping a curious mind, I come out having learned something new. Unfortunately, I notice it is getting more and more difficult to find people to do this with. As if no-one seems to recognize the importance and the fun of having such conversations. When it comes to AI, there will be always human decisioning required, such as determining the purpose and intent of the AI, how we value its result, how we control its behavior, how we define the meaning of the data it needs to work with. If we do not, we become mere passengers of AI. We become like a dog that does not understand where his food comes from. If conversations help me understand and stay in the drivers seat, it will be more important than ever.
You are touching on something absolutely critical here, putting human beings at the centre of tech. Getting big tech under some kind of human control again.
Please take a look at this proposal. It argues that everything we need to have sane, sustainable, human and humane societies is actually in place and trying to function -- only we mess it all up by failing to comprehend the organic structure that is already there and is trying to express itself.
All that's needed is *consciousness* -- but true mass consciousness, a genuine and general social awakening on the part of the human race. Otherwise I am 100% certain it's all over for us, we are descending into the kind of chaos from which there is no return. There's a very narrow window indeed to try save the situation.
I've explained these principles to children, in simplified form, they had no trouble understanding the concepts. These principles should be taught in every school to every kid.
I've been saying, we need H2H signalling these days -- human to human. H2H channels with high signal-to-noise ratios. I think we're on the same page. Good luck. I'd really appreciate your thoughts on this proposal -- I had been meaning to write it for years, in the end it got written very quickly indeed but I'm happy with it:
https://systemshaywire.substack.com/p/the-threefold-social-platform-truly