⟁ Risks & Headwinds

I had the privilege of spending this past weekend in California’s Calistoga region for some R&R with my wife. Our days were filled with exercise, deep conversation, and quiet reflection time which had me, well, reflecting.

The past year has left me sufficiently obsessed with the speed, scale, and potential of this next AI era. In my United Airlines economy middle seat out to San Francisco, I was nearly brought to tears watching Demis Hassabis in the Thinking Game. Tell me we've hit the peak of inflated expectations without telling me we've hit the peak of inflated expectations.

In addition to letting my imagination and curiosity run wild envisioning the immense breakthroughs and benefits our society may soon see as a result of this technology, I’ve spent some considerable time contemplating the risks and headwinds.

While there are a plethora of potential problems to lose sleep over, I’ve chosen to get restless over a handful I feel will have the greatest impact on our local community psyche, national security, and humanities longevity. The below is listed in perceived order of immediacy, from happening now to near and next:

AI breaks the grid

Industry executives are officially sounding the alarm on the mismatch between AI’s energy demand and our aging infrastructure. From 1900 to 2000, new grid power capacity in the US increased 10x every 30 years. In the last 25 years, it has only grown 1.3x. This is at odds with the promises AI executives are making for data center capacity buildout, with annual CAPEX spend expected to exceed that of more well-established, capital intensive sectors like telecom and oil & gas in the coming years. Meanwhile, as the US’ most formidable adversary, China, for all its shortcomings, continues to invest in renewable, firm energy sources at a scale commensurate with the existential weight of the alternative. In 2024 alone, China added 429 GW to the US’ 48 GW. They jumpstarted their civilian nuclear electricity program back in 1994 while the US did not get back on the saddle until the early 2010s. Perhaps the federal government ire and Silicon Valley innovation engine kicks in to address this problem - through more efficient compute and/or novel energy technologies - in turn making this reality more manageable. Or perhaps not.

Engineering our unplanned obsolescence

As a fast follow to the loneliness epidemic widely reported on by researchers and intellectuals like Jonathan Haidt, we find ourselves at the brink of spiraling into an attention-deficit and de-skilling crisis, exacerbated by consumer digital products and generative AI solutions that throttle our dopamine addictions. In general, it is best to be weary of companies and products where attention is the business model. In fact, we appear to be rewarding companies that prey on our most vulnerable proclivities, leading to undesirable habits forming that are beginning to threaten the long term economic viability of large swaths of working Americas. Perhaps unrelated (or not), there is fear of impending white collar layoffs resulting from this technological shift, and in a hyper-competitive capitalistic society my guess is things will get worse for the white collar employee before they get better - as it is the company’s ‘fiduciary duty’ to take advantage of efficiencies as they present themselves. This may sound paranoid but remember - only the paranoid survive.

Adversarial embodied AI

This fear is the grandaddy of them all. There has been much public talk about AGI and ASI, but less of a public awareness push around what is being called embodied AI - or the convergence of advanced AI systems with advanced robotics (think humanoids). Of all the destructive possibilities of the AI era, accidentally developing able-bodied superintelligence that we cannot properly control leaves me wondering whether this is our Fermi Paradox moment. Like many of the developments of the last 12-16 months, the prospect of advanced systems like these feel like science fiction until it isn't. The speed at which innovation is being unlocked in this space is just too dang exciting (and competitive) for us to expect our leading researchers to carefully and meticulously approach its creation. We simply cannot help ourselves, and once the genie is out of the bottle, it can't be put back in.

Perhaps thankfully, there are still plenty of deeply technical, unaddressed problems in the space of embodied AI, and experts are still waiting for the ‘ChatGPT’ or ‘move 37’ breakthrough moment for embodied AI as seen with LLMs. For many reasons, a scenario in which embodied AI goes haywire should concern us just as much as, if not more than, traditional global threats like nuclear war. In fact, some of the top minds in the field find the likelihood of disaster less than negligible. I ask: What odds would you find acceptable? The cynic in me rests by reminding myself where the incentives lie, and perhaps they say these things to stir up headlines. Or, consider the possibility that - holding cards we laymen don't have - they actually mean it. Que the Don't Look Up AI-infused sequel and get to Zuckerberg’s bunker.

Wrapping up…

So what to do about it? Every generation has faced its challenges, many of which felt insurmountable if not on the precipice of collapse at times. As a member of the Millennial generation, I am starting to come to terms with our problem set. In the next decade (if not sooner), I anticipate many of these issues will come to a head. Without an inflated sense of my own self-worth, I am determined to contribute in whatever way I can to making the next day better than the last - for myself and the communities around me. In this sense, perhaps for the first time in my professional life, I am beginning to see what it might mean to carry out a purpose-driven career.

While I currently have concepts of a plan, I’m hoping to chart a more determinate path to my role in helping solve these problems, starting with further immersion in these domains in the coming months. As I continue to consume the latest reports and perspectives like a glutton, I will plan on sharing my understanding for others to investigate. In turn, I am continuously updating my future outlook, further calibrating how I spend my time while at Kellogg. More to come.

Thank you all for reading!