Investor & recording artist. Trustee, Santa Fe Institute. https://xitter.cf/t.co/ISRFxh8utU
Los Angeles, CA
Joined April 2007
- Tweets19K
- Following2.2K
- Followers904K
- Likes8.8K
D.A. Wallach retweeted
Another day, another round of on-demand GPU rental indices surging across the board and reaching ATHs. Even the A100 index, which recently saw some softness, seems to be inflecting a little.
Has anyone on here successfully convinced their 3rd party lawyers to charge them less because they (the lawyers) are using AI now? Or have any large firms explicitly lowered their prices because of this to outcompete others?
Part of the answer: xitter.cf/arakharazian/status/21…
This chart shows the key risk to the AI trade, and it's missing from the latest revenue announcements.
The AI boom is fundamentally driven by a top 1% of companies that drive 80% of spend. Within that top 1%, 80% are venture-backed.
That's all fine if the AI trade works out, and if, as some prominent investors say, everyone wins.
But if it doesn't work out, then we have a cascading effect of large customers simultaneously pulling back their AI spend.
D.A. Wallach retweeted
California's Billionaire Tax Act would collect roughly $30 billion. Accounting for lost income taxes from departures, the mean Net Present Value is negative $39 billion, from Benjamin Jaros and Joshua Rauh nber.org/papers/w35899
D.A. Wallach retweeted
Nearly 40% cumulative inflation in a decade.
Trillion-dollar deficits.
Exploding national debt.
Money printing.
A Fed that fanned the flames.
This is what happens when fiscal and monetary discipline disappear. Americans are left paying the bill.
I used to think artificial intelligence was fake (it's in the name). But since it became super, I now think it's real.
D.A. Wallach retweeted
At some point, a reasonable question will begin to be asked as to when or whether SI (formerly known as AI) will generate enough free cash flows to start funding its own growth and continuation. That’s the ultimate RSI: the technology gets better by itself and funds itself.
And “all the dollars” in the world are available for the taking if “super intelligence” is willing to pay for it (well up to a point). What’s the highest level of (risk-free) interest rate that’s still consistent with an aggressive AI buildout offering a solid return and positive net PV?
The point @chamath was making in the clip @elonmusk responded to was that in order to raise more capital (dollars) to build out AI, Anthropic likely would have to concede a substantial amount of equity valuation, which is just the flip side of expected equity returns.
More dollars are always available for the taking for super intelligence as long as higher rates of returns are promised either in the form of higher interest rates (debt) or higher equity returns (lower stock prices). In some grand sense, Modigliani-Miller theorem holds.
Finally, all of the above is true up to a point, which is “credit rationing” due to adverse selection. At some point, someone borrows so much money and/or at such a high interest rates that the lenders’ natural assumption leans towards that the borrowers may be strategically intending on defaulting. That that’s the only way to justify their borrowing.
And since the borrowers typically have much more private information about the likely odds of success of their projects (maybe), the lender rationally responds eventually by cutting off access to credit at any interest rates or in the case of equities declining to invest in the stock if the prices are falling and expected to fall further.
There’s a lot of discussion and ink spilled on the ROI question of AI. Most attempts have been from a bottoms up approach of counting the dollars of revenues generated by AI directly or indirectly in adjacent plausible areas. I think that misses the ball and is inherently flawed. At some point, the investment is so large, which it is for AI, that the ROI question suddenly becomes simpler.
Just look at GDP, ideally real GDP. Is the US economy growing meaningfully faster in real terms after the adoption of AI or not? Will we see 3-4% real GDP growth in the next couple years? The capex is projected to be big enough that that’s the level of returns that are needed. I don’t think AI can be repaid by cannibalizing other existing industries alone. The adopters of AI must be able to generate value from it and grow faster and larger so that the aggregate grows faster and larger. Nothing less will do.
Any "crowding out" that is occurring seems to be more likely driven by the govt's borrowing than by AI investment.
Cool!
In partnership with Vinod Khosla (@vkhosla), I’m proud to introduce Reservoir.
Billionaires have owned their cells for years. Starting today, you can too.
Sign up today: withreservoir.com
D.A. Wallach retweeted
Agreed. I think bonds, which never believed in the AI (raising aggregate productivity) story, are starting to price in downside economic risks resulting from further Fed tightening and eventual AI slowdown against a backdrop of (perceived) worsening frontier lab fundamentals and continued macro and geopolitical headwinds. Outside of AI buildout, the rest of the US economy is kinda lukewarm/crap, but we knew that already. (I’m prob wrong.)
I think bonds have bottomed for now. Rates and vol got high enough to clean out bad positioning and leverage while attracting portfolio diversifiers.
Nobody knows where levels should fundamentally trade because the timeframes are so long, but 3.3% real and 5.6% nominal is pretty compelling.
Even with sticky inflation, you have to remember that owning more bonds allows you to own more risky assets, especially those you might consider inflation hedges.
Looking in the rear view mirror in terms of poor bond performance over the past 5 years is not a good argument for avoiding bonds in today's portfolio.
If anything, the case gets stronger the higher yields go.
D.A. Wallach retweeted
Two souls are dead because they self administered something purporting to be the most powerful metabolic agent ever developed - without medical supervision, before the inventor has asked for approval and sourced from a Chinese “lab” seeking to make a quick buck.
Peptide World - “Lilly made this up”.
Two deaths were reported to the FDA from research-grade retatrutide (illegal for human use). cbsnews.com/news/deaths-repo…
"People are by nature tribal moralizers. They view the world as a battle between good and evil, our tribe vs. their tribe. If we can just quash the evil people and force them to be less greedy all would be well with the world. Adam Smith’s vision, how greedy self-interest serves the public good is unnatural.
People think in simple stereotypes. The large landlord can surely “afford” to lower rents for the struggling “working family.” The market accommodates unimaginable variety. Some landlords are hard-scrabble immigrants who saved a lifetime to run an apartment building. Some rent-controlled tenants are well-connected trust-funders who moved in decades ago.
People also don’t see the unseen. We see the lucky family that got a break on their rent. We don’t see how immigrants, minorities, people with children or pets, anyone with sketchy credit or little cash, or those that lost the affordable-housing lottery can’t get apartments at all. We don’t see the apartments that didn’t get built."
Affordability Trilogy part 3: Why is the free market so unpopular?
grumpy-economist.com/p/affor…
D.A. Wallach retweeted
This is super macro important….in fiscal sustainability/ math for US. At a time when neither political parties have the will to practice fiscal austerity (except for tariffs), the creative way is to bring down longer term Medicare and Medicaid costs via less obesity.
D.A. Wallach retweeted
Allegedly “the most important chart” for understanding the current state of the AI economy. The model layer seems to be going through a period of commodification as non-frontier models saturate everyday use while frontier models compete fiercely. Compute layer sees massive increase in total demand for compute regardless of the source or use case.
D.A. Wallach retweeted
The real world abounds with recursively self-improving systems, but one in particular deserves attention: Science, modeled as a system (perhaps even as an agent, with goals and resources). If you want to really understand AI RSI, science should be your reference point.
Science is an intelligent system, and it is obviously recursively self-improving:
1. Scientific discoveries unlock new technology that helps build better experimental tools. This is a top driver of progress in nearly all fields.
2. They unlock new conceptual advances (ideas, theories) that help solve more problems.
3. They increase society's economic output, leading to more resources flowing into science.
4. They unlock better faster tooling (e.g. more compute via better chip & networking technology).
As a result, many measures of scientific *input* grow exponentially:
1. Headcount (doubles every ~15 years)
2. Global R&D spending (doubles a bit faster, every ~13 years)
3. Papers and patents (technically this is a measure of headcount)
4. Compute dedicated to science (doubles every ~2 years)
But is scientific progress exponential? Historically, the rate of scientific impact over time has remained roughly constant since the start of the industrial revolution (i.e. scientific progress is *linear*). 1850-1900 was about as dramatic as 1900-1950 or 1950-2000.
1850–1900: Evolution, germ theory & antiseptic surgery, thermodynamics, electromagnetic field equations, the periodic table, pharmaceuticals, electricity, telegraph and telephone, internal combustion engine, skyscrapers, mechanized agriculture...
1900–1950: special and general relativity, quantum mechanics, nuclear fission & atomic energy, antibiotics, genetic theory, electronic computers, information theory, synthetic polymers and plastics, the transistor, aviation...
1950–2000: DNA, genetic engineering, integrated circuits, microprocessors & personal computing, the Internet, crewed spaceflight, moon landing, satellite communications & GPS, standard model of particle physics...
In real terms, like life expectancy, which has increased in a remarkably linear fashion of roughly 3 months per year since 1840, progress is a straight line. This is especially apparent for fields where impact is easy to measure, like biology, medicine, and agriculture.
I first wrote about this phenomenon and its causes in 2012, and a steady stream of research has confirmed it in the years since. Examples include the 2018 paper by Nielsen and Collison, "Science Is Getting Less Bang for Its Buck," and the 2020 economic paper, "Are Ideas Getting Harder to Find?" (In fact, I believe the Nielsen paper stemmed from a conversation I had with him about this exact idea six months earlier)
In short, the primary cause is that research solves the highest-impact, easiest problems first, and every subsequent problem is either harder or lower-impact. Exponentially so. The paper that presented information theory wasn't very hard to write (single author!) but you'd have a hard time ever writing a CS paper that beats it in impact.
This is why science as a system requires exponential resources (input) to produce linear impact (output). It gets exponentially harder over time.
Worth thinking about if you're pondering RSI for AI. I fully believe AI RSI is already happening and will accelerate in the future. But I do not believe this leads to an "intelligence explosion" -- that would fly in the face of everything I know about intelligence and everything I know about recursively self-improving systems.