I don’t want to sound like an AI hater here so here is a more bullish view (also on ssrn) by a group of researchers out of Switzerland. They add an LLM screen to a momentum signal and get an uptick in returns.
-
LLM, AGI, and why we may need a world model
This is worth a read: https://www.wsj.com/tech/ai/yann-lecun-ai-meta-0058b13c. In case you don’t have access or time here are a few notes:
- Ok, it’s not at all finance related but everything is AI now and LeCun’s view seemed interesting in a contrarian way, so why not write a quick post
- The idea is that human thinking involves modeling the world around us, i.e., things, animals, people, the relationship between those entities, etc., in our minds. We combine those with data to arrive at inferences and predictions
- E.g., if I push the cup over the edge of the table, based on the model I have of cups, tables, space, and gravity, I can make a prediction as to what’s going to happen
- I can formulate a similar conclusion by looking at enough sentences with the words ‘cup,’ ‘table,’ ‘edge’ and ‘push’ in it, but that formulation doesn’t quite seem to be rooted in the same kind of understanding. It also doesn’t lend itself to the same kinds of generalizations
- LLMs’ logic resembles the second approach a lot more than the first. They predict words based on other words.
- Because of this seeming fundamental difference between human thought and LLMs, a good number of people who don’t live in this space (including myself) have doubts that pure LLMs can get to human-level AGI
- Of course, anyone would be right to take this with a large grain of salt because those are not AI experts
- This is why it’s nice to see a bona fide AI guru who’s been at this for a long time agree with that intuition, presumably based on an incomparably deeper knowledge of the subject matter. This view may have existed within the AI community for a long time, I just never saw it bubble up in the popular press
- That is not to say AGI isn’t possible, but according to LeCun, it won’t happen until we somehow bring a ‘world model’ as he calls it into those systems. Therefore, he says we’re not nearly as close as a lot of folks seem to think
The wsj piece is here. It includes some of the Meta internal politics surrounding LeCun’s exit
This is a youtube version of the same argument
-
Those pensions sure were nice…
You may have detected a common theme in our posts so far: How do you transform financial assets into a return stream you can live on? Basically every retiree’s (and some non-retiree’s) problem.
In essence, something resembling a pension, i.e., good for life and increasing with inflation.
Click here for a discussion of the different options.
-
Citadel’s Ken Griffin says generative AI is no good…
… at alpha generation – at least as of right now. I’d say this counts as an informed opinion. View the brief video here.
-
Vanguard’s Take on the next Decade in Equity Returns is Even More Dour than AQR’s
- Compared to Ilmanen’s, Greg Davis of Vanguard provides a more pessimistic outlook still: He expects the S&P 500 to return between 3.8% and 5.8% (mid point of 4.8%) over the next decade.
- Note that his numbers are now nominal. To make them comparable to Ilmanen’s, we need to subtract inflation: Say inflation clocks in at 2.5% that means 1.3% to 3.3% real. That range sits below Ilmanen’s almost entirely!
- It’s also not very different from Tips yields (1.9% 10 years out and 2.6% for the 30 year). Those of course, are also real. If you believe this, why not just buy Tips?
- His argument largely rests on the currently elevated CAPE ratio.
- Davis is especially concerned about US stocks and recommends a 60:20:20 blend of Fixed income, US equity, and foreign equity – a significant departure from the standard recommendation of 60 equity, 40 fixed income, with the former two-thirds or more US.
- This coming from Vanguard is significant.
-
What’s Been Driving Equity Returns; What to Expect – via AQR
- A great paper by Antti Ilmanen of AQR. At this point, Antti is one of the foremost experts on expected returns across asset classes (he also wrote an excellent book on the topic).
- A brief summary:
- One can decompose equity returns into dividend yield, earnings growth (as a proxy for dividend growth), and multiple expansion. Those three will always fully explain realized returns over a given period.
- US real equity returns over the very long run (6.6%) are mostly Dividend yield (~4%), some earnings growth (~2%), and a small amount of multiple expansion (.5%).
- Since 1982 the picture has changed. The last decade especially was quite different. Real earnings were higher (9.7%), but now it’s mostly earnings growth and multiple expansion, and a much smaller dividend yield (1.7%).
- It is doubtful that this is sustainable:
- Multiples can’t expand forever (well, theoretically they can, but practically not really).
- Recent high earnings growth figures can largely be explained by corporate tax cuts and drops in interest rates. Neither of those can go on forever either (we already see interest rates moving the other way).
- Ilmanen concludes not with a specific return prediction but a wide range of ‘objectively feasible next-decade real returns’ of 3-8%.
- Note that these are real return numbers, so add your preferred inflation prediction for nominal figures.
- If those valuations revert and higher interest rates are here to stay, I’d say we’d look at the lower end of that range. Of course, AI could always save the day via much greater organic earnings growth…
-
‘Safe’ Withdrawal Rates
- The starting point is the standard retirement problem: given a certain amount of financial assets, how much can you spend each year without running out in your lifetime?
- The simplest type of solution takes the following form:
- Start with x% of your assets, the ‘initial safe withdrawal rate’
- Translate it to a dollar amount
- Then, every year, grow that dollar amount with inflation
- But what is the initial withdrawal rate x? The grandfather (or grandmother?) of answers is the famous 4% rule
- Over the years, the financial advice industry has moved away from it, which is good – but where did they land?
- I reviewed the major providers (Vanguard, Fidelity, Charles Schwab etc.) to see what they say and if there is a consensus within the industry
- Only some providers answer the question directly (others want you to talk to them for a customized solution). Here is a summary
Safe Withdrawal Rate in % by Time Horizon as of June ’25
Horizon C. Schwab M.Star Fidelity J.P.Morgan 10 10.3 9.7 15 7.1 6.7 20 5.4-5.6 5.2 25 4.4-4.7 4.3 30 3.8-4.1 3.7 3.7 35 3.4 3.4 3.0-4.0 40 3.1 3.2 45 3.0 50 2.9 Notes: C. Schwab = Charles Schwab as of April 2025. M.Star = Morningstar as of January 2025. Fidelity: using their Retirement Planning Tool in June 2025. J.P. Morgan: from their Guide to Retirement 2025
- A few takeaways:
- The major providers are more or less in agreement – that’s reassuring. MorningStar and Fidelity are very slightly more conservative than Charles Schwab
- Note that for very long horizons, the safe withdrawal rate is closer to 3% than 4%. Those who think about early retirement should take this into account
- Even at a 30-year horizon (a prudent number for many 65-year-olds), 4% is now considered too aggressive
Important: Your ‘safe spending rate’ is not the same is your safe withdrawal rate. Also, don’t take ‘safe’ too literally. See here for important notes and caveats.
-
First Post
What this blog is:
- Numerical: I find the popular press a little light on numbers. They tend to use them sparingly and with little to no context. You’ll probably enjoy this blog if you like things quantified.
- Eclectic: (as the name suggests) I cover a range of topics. Econ and finance are at the core, but many things are interesting and I try and provide a broad perspective.
- Unscheduled: I post when there is something to post. I feel like publication schedules incentivize writing for the sake of writing. Because of the sporadic nature of this blog, I’ll send occasional emails to those who want them.
- Written by a financial economist
Sign up to eclfin’s sporadic newsletter (once or twice a month max)
Eclectic Finance 