In the inimitable style of the legendary Khushwant Singh, this Diwali note is offered with a wink and a nudge, best read in the festival's spirit.
'Tis the season of Diwali, that annual ritual of sweeping out the dust of doubt and preparing for a brighter tomorrow. In that spirit, we attempt intellectual spring cleaning. We air out a few thoughts on skepticism—particularly the kind gnawing at the thriving field of GenAI. We are not pros like the late-night comedians. Hence, we must start with an apology-like disclaimer: skepticism is essential, especially in AI, which is buzzing with constant hype and seemingly endless funding. But a certain brand of skepticism exists that is less constructive criticism and more corrosive cynicism, a strain that doesn’t just question but undermines, drains energy, and dims innovation. Today, we'll examine three varieties of 'skepticism' to see which one may need a sweeping out.
Morgan Stanley Memory: Stuck in Reverse—The Perils of Rearview Thinking
Ever since the success of "This Time is Different," the rank of believers in the ironclad authority of historical patterns has continued to swell. In the field of GenAI, the biggest adopter of this style just won the Nobel Prize. The shifting views of celebrities in the field are entertaining, and also understandable, given that history keeps changing! Such insistence on looking for answers more in history books rather than in learning things for what they are is largely harmless and rarely helpful, even if some of the predictions of these experts will come right, given how many different things they say.
This brand of skepticism is more relevant when adopted by the analysts of famous houses. They can create at least a short-term storm in stock prices—like observed in the prices of all memory stocks when Morgan Stanley forecasted an upcoming winter—and, at times, a meaningful impact on the funding of projects as well.
For some, skepticism is an instinct that manifests like a reflex—a blink backward. Most semi and memory pessimists base their mean-reverting forecasts on old supply-demand dynamics, steadfastly refusing to accept a present where innovation, pricing power, and monopolistic positions may have changed the ranges entirely. We have even seen one analysis about overcapacity in segments like CoWoS, an industry with one supplier. Most such skepticism is because of discomfort with a one-way rise.
This type of skepticism isn't all bad, of course. Cycles are inevitable in real life, and all things reverse from some high. Logic and wary observation may serve as a better guide than history to imminent red signals. Still, the wisest approach may be to look through the cycles if one has any faith, as we do, in the long-term value-creating abilities of these industries that were historically at a different place.
Sequoia's $600 Billion Indignation: When Nostalgia Becomes an Accounting Manual
To optimists like us, Sequoia's virally famous "$600 billion question" was almost a lament on a cosmic injustice. The argument was irrefutably simple, at least for all in the echo chambers: historically, every time the hardware and near-hardware folks earned $100, everyone's beloved application-layer companies—often Sequoia’s portfolio stars—made $600. The headline was one of the easiest to write once someone counted today's hardware giants swimming in $100 billion in AI revenues.
We can return to our usual rants and write long on how the tides have shifted. But, this is festival season, so we must not make another attempt at hammering in themes that have not found sufficiently many viral hits. Having been previously from the sell-side, we realize the power of those in charge of the cashbox - application layer players are the ones to collect money from the beneficiaries and users of GenAI. Even if they have to funnel more of it to hardware providers who work a few layers removed, they must make enough money to keep everything going.
More seriously, even if the hardware companies now provide the oxygen for this AI revolution, the changing nature of the capex cycle needs repeated skeptical evaluation. After all, deep cycles are a reality when capex amounts are high, buildout phases span years and returns are uncertain. Skepticism like Sequoia’s, especially when it goes viral, has value in tempering everyone's optimism. Hopefully, it breeds financial prudence in some dark corners. This is the kind of balanced skepticism we need—one that calls for caution without suffocating future opportunities.
Apple’s LLM Condemnation: How Fake News Starts
The first articles claiming Apple has unveiled "deep cracks" in AI reasoning came out two weeks ago. In this author's circles, this damnation is going viral only now. The research paper at the eye of this skepticism claim is at https://arxiv.org/pdf/2410.05229, and it is as straightforward as they come in discussing LLM models' current abilities to do formal reasoning. Some scribes in the media twisted it into an indictment from Apple on all things GenAI. The original paper is forgotten, but the number of posts and forwards on how Apple does not believe in AI continues to grow, even when one would assume all the fake news to be on one subject only for the next few weeks.
This is the hurtful, deeply cynical brand of skepticism, where the truth does not matter. Here, it’s not the content that’s problematic but what it leads to when coming from leaders and decision-makers. Our Diwali gift is a simple podcast at https://geninnov.ai/s/Weqnzx that expands on the points through conversations between different types of hypothetical chief technology officers and how their slight change in context and enthusiasm could lead to diametrically opposite decisions on the adoption of technology. In some form, such conversations are taking place globally, and the least they need to do is filter out fake news that has nothing to do with other positives or negatives of this technology.
Ping Us to Show You What GenAI Can Do
For any skeptics on the utility of GenAI, just call us to discuss how we use it daily. We can no longer insert podcasts in articles like these, but we have been generating accompanying images for almost a year despite being completely artless. At least to our simpleton eyes, many of our images could have adorned the cover pages of magazines a few years ago.
But we digress! We use GenAI in doing the highest quality balance sheet analysis of companies we look at, those unavailable in any reports we have been able to lay our hands on. We use it to tabulate the details of tens of molecules under development at large pharma companies in ways and depths we have not found anywhere. We use it to understand the most complicated technologies at our level of understanding. We use it to write programs to save us money, for second opinions on our legal documents, and for so many other things—all the way up to writing Diwali articles.
As we light up the diyas this Diwali, let’s remember that skepticism, like any powerful tool, requires delicate handling. Skepticism comes in many varieties, including concerns about valuations, ethical implications, societal issues, inequality, environmental damage, etc. All healthy skepticism has its place, but some—often wholly unfounded and paranoid—can breed stasis. So this year, let’s sweep out not just the cobwebs but also the naysayers’ dust.
Happy Diwali, everyone.
