3,230 karma · joined February 5, 2018
Email: [firstname]@fehri.ng
Of course, higher taxes on the heaviest vehicles would just mean higher prices on consumer goods, so this mostly ends up as a subsidy to consumption.
In an ideal world you'd want every founder to have background in the relevant industry and strong business sense and the ability to build the product, but being strict about all of those criteria would result in a lot of false negatives.
I don't know if Uber or WeWork are or will ever be good investments, but their products are clearly great for users. The fact that they abstract away the underlying hard problems so effectively that they seem like they should be the same app is just a testament to this. But the underlying problems that they're solving are fundamentally very different. WeWork doesn't need Uber's complicated routing and matching algorithms, for example. And unlike WeWork, Uber probably doesn't have architects or construction engineers on staff.
> Are you saying instead, that concrete predictive algorithms need improvement or are we lumping the tuning into this?
in the context of what's needed to get to AGI - just as if NASA built an engine we'd talk about its effectiveness in the context of space flight.
Separately, I think OpenAI's current investors have a >10% chance to hit the 100x cap on their returns. Their current models are already good enough to address lots of real-world problems that people will pay money to solve. So far they've been much more model-focused than product-focused, and by turning that dial toward the product side (as they did with ChatGPT) I think they could generate a lot of revenue relatively quickly.
[0] Except maybe in the sense that future models will be predominantly multimodal and therefore not strictly LLMs. I don't think that's what you're suggesting though.
[0] https://hazyresearch.stanford.edu/blog/2023-03-27-long-learn...
In short it seems like virtually all of the improvement in future AI models will come from better algorithms, with bigger and better data a distant second, and more parameters a distant third.
Of course, this claim is itself internally inconsistent in that it assumes that new algorithms won't alter the returns to scale from more data or parameters. Maybe a more precise set of claims would be (1) we're relatively close to the fundamental limits of transformers, i.e., we won't see another GPT-2-to-GPT-4-level jump with current algorithms; (2) almost all of the incremental improvements to transformers will require bigger or better-quality data (but won't necessarily require more parameters); and (3) all of this is specific to current models and goes out the window as soon as a non-transformer-based generative model approaches GPT-4 performance using a similar or lesser amount of compute.
Only because most people who can afford to buy in Cupertino aren't paying for it with earned income, right? Google tells me a $2.5M mortgage has minimum payments of $230k a year, anyone who makes enough to afford that is probably paying at least $50k/year in CA state income tax, on top of ~$25k in property taxes. Austin has high property taxes but I don't see them coming anywhere close to that, and the sales tax is lower.
> Society depends on this mismatch existing. It must exist somewhere. The alternative is a much poorer and riskier world, which includes dystopian instruments that are so obviously bad you’d have to invent names for them.
I guess I'll dispute this. It is useful that this mismatch exists, since it (1) lowers the cost of long-term borrowing for mortgagors, businesses, and governments and (2) lowers the (direct and/or opportunity) cost of holding cash. But I don't think society is dependent on this mismatch, and I don't think the alternative would be anywhere near as bleak as Patrick suggests.
If bank regulators changed capital requirements to require banks to fully back deposits with cash equivalents, long-term borrowing would be a lot more expensive, but the market would still clear. There's already plenty of demand for safe long-term debt, and that demand would only increase as long-term interest rates went up. E.g., if checking accounts paid -2% interest and CDs paid 10%, lenders would put less money in checking and more in CDs, even if it meant they would have to sell the CD at a discount if they needed liquidity.
Of course, the US government will take any and every opportunity it can get to indirectly subsidize mortgages, so this is pretty moot in practice.
All major US banks - and all or virtually all US banks in general - have assets that are designated as held to maturity. Continuously marking all assets to market would create massive swings in banks' income and obscure the real gains and losses from their operations.
SVB probably had a somewhat longer asset duration and somewhat lower book yield than US banks on average, since its deposit base grew so quickly in a low interest rate environment in 2020-2021. It also had a higher share of uninsured deposits. But nothing that SVB did was categorically different than other banks, and in the absence of a government backstop, I'm not convinced that any US bank would fare much better if faced with a similar volume of deposit outflows. "Magically" transforming long-dated assets into short-dated liabilities wasn't any kind of malfeasance on SVB's part - it's just how banking works.
That also creates a timing mismatch - Rivian's revenue in Q4 was $663M and its cost of revenue was $1663M, but most of the $1663M is associated with vehicles that they haven't delivered or even manufactured yet, so they'll recognize that revenue in future quarters. I don't know what it actually costs them to manufacture one vehicle, but I bet it's a lot less than 2.5x the revenue they get from that vehicle.
Leaving aside the accounting and addressing your real question: From a quick search it looks like Rivian has raised a total of $23B of capital between VC rounds and its IPO, and it has $13B of current assets (mostly cash + inventory) as of 12/31, so that gives it a fairly long runway even at its current burn rate. But its burn rate - loosely, revenue minus expenses - is expected to slow over time, assuming that (1) its revenue grows faster than expenses (it's expecting deliveries to ~double this year, which should increase revenue at a similar rate) and (2) its unit economics work out, i.e., its "true" cost of revenue is less than its revenue.
The reason for the widespread grift in the crypto space is not just that blockchain applications are a relatively new and overhyped technology. Another huge factor is that many blockchain applications, by their very nature, are designed to circumvent securities regulation in order to raise money from unsophisticated retail speculators.
I'm sure some shitty AI startups will raise capital from retail speculators on crowdfunding platforms and the like. But I expect it to be a lot less widespread compared to the crypto space, and it will happen within the confines of (mostly) US securities regulation, which will mean much better disclosure and much less outright fraud and misappropriation. The main "victims" this time around will be, like, the LPs of second-rate VC funds, and in many cases those fund managers are well aware of all of this but are incentivized to follow the trend anyway.
(I'm an Airbnb employee.)
Maybe it will be more common in the future as a result of remote work. But today it seems hard to create the conditions that result in great school districts - i.e., a relatively high concentration of relatively high-income parents who value education - in a small town that's not within commuting distance of a city with good jobs.
I imagine my hometown would be a very different place if people didn't have to leave to get decent jobs. Though of course there are plenty of other good reasons to leave.