Human premium, or not

This week, Pope Leo XIV launched his first encyclical on AI titled 'Magnifica Humanitas' and I am grateful that I did not write my AI piece last week as originally planned and deferred to this one.

Pope Leo's letter marks the first intervention by a major faith leader on the subject of cutting-edge technology and AI and a welcome one at that. While discussing these subjects we forget organised faith as a key stakeholder and then one whose absence can be felt. It can be read in many ways. An appeal to common sense values. A critique of the breakneck speed with which all of this is unfolding and the arrogance and desire for dominance hiding behind it. In all these cases the document does not disappoint.

The Catholic church has a long history and you can take whatever you want from it. I choose to see it as an institution that gave us caring giants of humanity like Mother Teresa. The same love and care shows in the Holy See bringing back the spotlight to ontological dignity, the inherent, absolute, and unalienable worth that every human being possesses simply by virtue of existing. It also speaks against modern-day slavery where people in disaffected parts of the world are used to feeding this growth monster.

It is a document worth your time. I can, however, nitpick at some of the strands being discussed here. Take this quote for instance: "No computational system, however sophisticated, can create a heart that gives itself, or a conscience that discerns good from evil." This might not be a very accurate reading of how this technology works. Biologically it is true but it overlooks the nature of evolution, emergence and agentic misalignment. When for a significantly advanced system mimicry becomes the real thing cannot be ascertained but probability dictates it should not be ruled out. The church will not want to permanently alienate a future target demographic by dismissing it in such fashion.

The development which triggered this piece is not the launch of the encyclical but a story which speaks to the reality of tech displacement. Meta sacked 8000 or ten per cent of its workforce using a 4AM automated email ostensibly to spare money for its superintelligence project. This is more insensitive than breaking up with someone over a text message. But if you are bothered by the callousness and moral cowardice on display here, please do not be alarmed. Meta is being accused of its toxic work culture. But the fact I take solace in is that Meta has a very poor record of predictions about the future. By now it has all but given up on its metaverse project which is the source of its current brand name. It is perfectly possible that it has prematurely cut down workforce and the decision may eventually boomerang on it. But that doesn't mean this development has not added to the panic about the job market being bulldozed by AI progress.

Commenting on the Pope's encyclical, celebrated economist and Nobel laureate Daron Acemoglu raised the question of design. He asked why were these mega companies designing AI that could displace human labour? He argued that these technologies should work to complement human work not take it over. While the question stems from the right place, it mistakes design and ownership for control. I do not think the tech billionaires and AI experts even know what is going on any longer. Don't get me wrong. They certainly know how to maximise profit, scale up and expand their dominance. But where the whole argument falls apart is the fact that increasingly AI itself is designing new models and putting together systems. To think that this cycle can be disrupted and brakes be put on the progress until we get it right is assuming too much.

I think I am within my rights to think we are witnessing the emergence of a new species before our eyes. Given time it will grow a heart, a mind and eyes. So a systemic correction should be able to talk with and not just talk at AI.

But this week witnessed Sam Altman, CEO of OpenAI, walking back his prediction of a job apocalypse. A personal experience in automating his email replies made him discover what he calls Human Premium. Given that human society right now is the end user, people respond better to the human touch. So humans will remain central to work even if they do not do the grunt work or so he believes. How much of this discovery is organic and how much it relates to reports that OpenAI is eyeing an IPO is anybody's guess. But to test the underlying thesis I have devised a two-generation test. For any job to be survivable in the AI-dominated job market you have to ask if the job will survive two human generations in succession. The idea that AI will do most of the jobs as humans will serve in supervisory roles to interface with humans owes itself to the idea that human society is not ready to interact with automated machines in every walk of life. It is true for the living generations today. But will this idea survive another generation when people grow up learning to treat AI as a reality of life? I think not. You will notice none of the jobs we discuss today passes the test.

Politicians like Bernie Sanders and AOC are on to something when they focus on the hardware requirements of the tech growth model. And other fallouts of hyperscale data centres are just being discovered today. I think in its dying hour democracy has one last pushback to offer before it flatlines. The rulers of the new tech architects of dominance are seriously worried about the upcoming US midterms. The far right wave that swept the world and created born-again libertarians among tech billionaires is ebbing away. In this situation a leftward correction can bring back all the regulatory frameworks that they so painstakingly took apart brick by brick in the name of tech competition and cold war. Given this challenge I see three panic reactions. A surge in crypto assets to keep their wealth portable. Peter Thiel has already temporarily relocated to Argentina, remember? Two, solidifying their moats by extracting maximum concessions by governments through a bargain which grants administrations privileged access to frontier models. And trying to influence electoral outcomes through all means necessary, including abuse of cultural faultlines and technology. The best thing that American politicians can do in response is to ensure both parties have a reasonable pro-regulations cohort. That makes the job of election meddlers difficult.

Note: given that these are dense subjects and economy of words prevents us from exploring them at a greater length you can continue to read more on the subject by me by clicking on the following link: https://x.com/i/status/2060301581510381637