Blog

The Suicide Region: Option Games and the Race to Artificial
General Intelligence

9th December 2025 

Nation states - and major AI companies - have all acknowledged the existential risk of racing towards artificial general intelligence (AGI). Industry godfathers like Geoffrey Hinton have estimated the probability of catastrophic failure at up to 20%. Standard real options theory predicts that in a race marred by excessive volatility and technological uncertainty, the option to wait is extremely valuable. i.e. wait and learn from the costly mistakes of the first mover. 

So why is everyone racing towards achieving AGI while firing their safety teams? 

In a new working paper of mine, "The Suicide Region", I show that racing towards potential existential oblivion is not madness, but rather, mathematics. Agents are acting rational in a pathological game.

I show that the cost of global catastrophe (existential risk) mathematically "cancels out" in the equilibrium condition. This creates a trap where the pressure to race (preemption) dominates, despite negative net present values when accounting for the risk of catastrophe.  

Full paper available at: https://arxiv.org/abs/2512.07526

Exciting New Upcoming Working Paper 

7th December 2025 

I have an exciting new paper coming out on SSRN shortly. It will be posted here :) 

From Rigor to Irrelevance  

1st December 2025 

The 2010 Credibility Revolution "took the con out of econometrics". Researchers began to focus meticulously on research design, identification and natural experiments. Difference-in-difference regressions, propensity score matching, novel external IVs, and - the holy grail - natural experiments. 100% I understand and wholeheartedly support the move towards rigorous and credible econometrics. 

However, I've observed a subtle but concerning shift. The main focus for some researchers I've met is finding a good experiment, with little consideration to the weight of the research question itself. Sometimes the research question - and its societal significance - appear merely as an afterthought. Driven by the excitement of uncovering a novel experiment, and - lets be honest - the prospect of attaining an A* publication, some researchers forget that the primary purpose and importance of research is embedded in the research question itself. Who are we helping? Which community will be better off for our findings?

I admit, earlier in my career, I too was excited over a potential A* publication at all costs. If a cool econometrics model can be applied to a new dataset, lets try it and we'll figure out the intellectual and societal contributions later. But if the goal of research is to improve human society through the acquisition of knowledge, shouldn't the research question come first, and the experimental design later? 

As Angrist and Pischke aptly put it, have we "taken the econ out of econometrics"?

The link to Angrist and Pischke's 2010 paper.