E106: SBF's media strategy, FTX culpability, ChatGPT, SaaS slowdown & more

Because mainstream media outlets protect ideological and class in‑group figures and resist admitting errors, similar large-scale frauds or grifts exploiting this bias will continue to occur in the future in the United States media/financial landscape.

“And so if you don't kiss the ring and bow down to them. They will try to destroy you or run you out of town. But if you are one of them, they will give you a hall pass. And when it's time for them to change their mind in order to tell the truth, they won't do it. And so these types of grifts will continue.” View on YouTube

In the coming years, independent journalists and creator-type media will become the majority of media volume consumed, and traditional journalism/press will be the next major sector to be disrupted by the creator model.

“and now independent journalists are going to become the bulk of volume that's going to be consumed... Journalism and what we call the press is very likely going to be kind of that next layer of disruption.” View on YouTube

During 2022, Xi Jinping will be the biggest political winner globally, effectively becoming ruler for life of China, and his expanded power will begin to play out domestically and internationally.

“My worldwide, uh, biggest political winner for 20 2022 is XI Jinping... 2022 marks the first year where he's essentially really ruler for life. And so I don't think we really know what he's capable of and what he's going to do. And so that's just going to play out.” View on YouTube

In 2023, Xi Jinping will exercise dominant, aggressive influence both within China and internationally, leveraging China’s control over critical supply chains.

“For next year? I think it's going to be a he's going to run roughshod, not just domestically but also internationally, because you have to remember, he controls so much of the critical supply chain that the Western world needs to be.” View on YouTube

In 2023, Xi Jinping will maintain or strengthen his grip on power even as conditions worsen for Chinese citizens and sectors like billionaires, tech, and real estate, and there could be negative financial contagion from China; overall, 2023 will not be a good year for China’s economy.

“I think I think the bigger risk is, is that China gets better for XI Jinping, but worse for everybody else in China... I think there could be contagion from China next year. I don't think she's going to lose his grip in any way, but I'm not sure China's going to have a good year next year.” View on YouTube

As China reopens from zero‑Covid, the Chinese growth engine will resume, and this renewed Chinese growth will have significant effects on U.S. economic growth and U.S. inflation over the subsequent period (starting in 2023).

“And now they're reopening. So I don't know I mean like I'm not sure what we're supposed to comment. What I, what I will stand by is what I said, which is I don't think we have a very clear view about what's going on, what the substance of these protests are and what people actually want. If you're only consuming US media.” View on YouTube

Over the next Silicon Valley funding cycle (the next several years after 2022), tens of thousands of startups will be created around generative AI, and this space will become the focal point of the next tech hype/bubble cycle.

“My prediction, which is so everyone's got the obvious prediction, which is there's going to be 100,000 startups that are going to emerge... So the obvious next step is a bubble will form... my guess is the next hype cycle, the next bubble cycle in Silicon Valley will absolutely be this generative AI business.” View on YouTube

As large language models and natural-language chat interfaces mature over the coming years, many competitors to Google’s current search-results model will emerge, and Google’s core search engine product will be at risk of radical disruption.

“there could be a lot of competitors to the one box and a lot of competitors ultimately to search. And ultimately Google's core product, their search engine could be radically disrupted.” View on YouTube

Over the coming years, many traditional enterprise SaaS applications will be incrementally replaced by "models as a service" (MaaS), where specialized ML models provide the core functionality instead of conventional software, leading to a broad shift from SaaS to MaaS.

“I think we're going to replace SaaS with what I call mass, which is models as a service. And so, you know, a lot of what software will be, particularly in the enterprise, will get replaced with a single use model that allows you to solve a function... So I think SaaS will get replaced over time with these models incrementally. That's phase one.” View on YouTube

The next major advance in AI, likely within the next several years, will be the emergence of powerful multimodal models (combining video, audio, text, and other data) from a big tech company or OpenAI, enabling solutions to more substantive, complex problems than current single-mode models.

“The next big leap, and I think it will come from one of the big tech companies or from OpenAI is... a multimodal model, which then allows you to actually bring together and join video voice data in a unique way to answer real, substantive problems.” View on YouTube

For at least the near to medium term, current large language models will continue to struggle with the last 1–2% of highly precise, high‑consequence use cases, and reaching that reliability threshold will remain exceptionally hard.

“When this stuff becomes very valuable, is that when you really need a precise answer and you can guarantee that to be overwhelmingly right, that's the last 1 to 2%. That is exceptionally hard. And I don't think that we're at a place yet where these models can do that.” View on YouTube

Achieving the final 1–2 percentage points of reliability/accuracy in complex AI systems (e.g., self‑driving or high‑stakes inference) will take multiple decades of progress.

“These last these last hundred or 200 basis points literally takes decades.” View on YouTube

Over roughly the next year after this December 2022 episode (i.e., through 2023), the SaaS industry will experience a significant contraction in jobs and a vicious cycle in which customer seat counts shrink rather than grow.

“Either way, there's going to be a big contraction in jobs basically around this industry. And I think that what that could do is cause a vicious cycle where... for the next year or so where seat contraction becomes the norm instead of seat expansion.” View on YouTube

In 2023, many SaaS companies will start the year with only 80–90% of the prior year’s revenue from existing customers (due to layoffs and seat reductions), instead of the historical 120%+ net retention.

“So the baseline for next year could be contraction. So instead of starting with 120% of last year's revenue, you might start with 80 or 90% because there's going to be so much churn.” View on YouTube

During the current downturn period (approximately 2022–2023), achieving 2x year‑over‑year growth in a SaaS business will be as difficult and as impressive as achieving 3x growth was in the prior, more frothy years.

“If you can grow two x year over year in this environment, that is as good as or better than growing three x last year.” View on YouTube

Over the coming years, general-purpose AI models (e.g., large language models) will become commoditized, and competitive advantage will primarily come from access to proprietary training data rather than from the models themselves.

“This is why I think the hunt for proprietary data actually becomes the hunt that matters. All of this other stuff, I think, is a lot less important, because I think you have to assume that all of these models will eventually just get commoditized.” View on YouTube

AI models in regulated healthcare domains such as tumor detection, if trained on very large proprietary datasets (e.g., breast cancer images), will be able to gain FDA approval relatively quickly using existing regulatory pathways.

“So, for example, if you use a healthcare example, let's say that you had the largest corpus of breast cancer image data, and you could actually build an AI that was a much better classifier for tumors versus other things. The FDA actually has a pathway to get that approved very quickly.” View on YouTube

In the near term, the U.S. government and ultra‑high‑frequency trading firms will remain the largest purchasers of machine learning hardware.

“They are, I can tell you, as somebody who sells, we sell a lot of machine learning hardware into this market. The biggest buyers are the US government and these ultra high frequency trading organizations.” View on YouTube