Let’s start this week with more good news. On one highly unscientific measure, the possibility that AI will cause the extinction of the human race has almost halved.
Two years ago Elon Musk put the risk at around 20%. This week Evan Hubinger, a big brain at Anthropic, proclaimed that he believes there is a 10% chance that AI could kill all humans within the next decade. He also said that the company was racing headfirst towards catastrophe and “the more senior the employee, the more concerned they are”.
Nonetheless, apparently the odds of our extinction have halved. So that's alright then.
Dark humour aside, there is clearly a problem. As I said in the 3rd July note:
“Nuclear history shows that governance, regulation, and co-operation is possible, and will make a huge difference. The aim is not to stop AI, but to shape it—so its considerable commercial benefits can flourish while its military and security risks are contained. That will require sustained U.S.–China cooperation. They both get that, yet history tells us that real progress is typically forced co-operation only after a painful shock. Fingers crossed.”
The uncomfortable question is increasingly obvious. If some of the people actually building the technology sincerely attach double-digit probabilities to human extinction, how can the rational response be to continue developing it at maximum speed while regulation struggles to catch up?
Progress in AI, already extraordinarily positive in many respects, may have to slow before comprehensive global regulation can have any meaningful impact. Sadly, that imperative might only be acted upon after a painful shock. At that point some governments may conclude that parts of the technology are simply too strategically important to remain entirely in private hands. Regulation might not be enough – it might end in State ownership.
The investment angle is becoming more immediately problematic.
There isn't literally a fixed quantity of money in the world, but there is a finite amount of capital available at any particular price. Competition for that capital is getting hot. Governments with enormous deficits want it. AI companies building enormously expensive infrastructure want it too.
Sharma estimates that AI applications are currently generating about $200 billion, while companies are spending more than $1 trillion a year on data centres and associated infrastructure. That’s a huge gap, and increasingly that gap has to be financed through bond issuance. At the same time the US government continues to borrow on a colossal scale.
When governments and companies compete more agressively for capital, its cost rises. In plain English, the yield - or interest rate - goes up.
Added to this is growing fear of inflation driven by higher oil prices. As I noted last week:
“Markets are completely blind-sided on what is happening with oil. Almost nobody is looking in the right place. It’ll soon get ugly.”
We didn’t have to wait long for “ugly”.
Tension in the Near East has ratcheted up sharply this week, on both the eastern and western sides of the Arabian peninsula. Brent oil has risen about 10% this week, at one point approaching $110 a barrel. Meanwhile the US 10-year Treasury yield reached 4.98%, just short of the psychologically important 5% level.
Hardly helping matters was the latest Trump silliness: the promise of a $5,000 “Trump dividend” to every American adult if Republicans retain both houses of Congress. The potential cost is comfortably above $1 trillion at a time when the US already has a vast budget deficit. Whatever its electoral attractions, the last thing an inflation-prone economy needs is another enormous unfunded cash giveaway.
That 5% Treasury yield is more than just a psychological barrier.
The 10-year Treasury is one of the most important benchmark interest rates in the world. US mortgages, corporate bonds and numerous other borrowing rates are priced directly or indirectly from it. It is also the “risk-free” rate against which investors judge equities and other assets.
For example, at 5%, investors can suddenly earn a substantial return by lending to the US government. Increasingly over-valued equities become less attractive – until prices fall to much lower levels. Risky assets therefore have to work rather harder to justify their prices.
John Hussman noted this week that, on his preferred valuation measure, the market's recent peak was more expensive than both 1929 and 2000.
There is an important caveat. Hussman himself says valuation is not a short-term timing tool. An expensive market can become still more expensive. His point — and mine — is not that a crash must happen next Tuesday. It is that the consequences become much greater if something finally causes investors to change their minds.
Will 5% Treasuries be the straw that breaks the camel’s back? Or the final snowflake, as I call it — the one that brings down the whole avalanche-prone snowy slope?
No one knows.
We just need to be absolutely clear about the vulnerability and be prepared: have a defence in place. Thankfully, it has still been perfectly possible to make decent profits amid these shenanigans, and not by doing anything weird or complicated.
Meanwhile, in the EU, with the threat of higher inflation coming from outside Europe, voters increasingly reaching towards political extremes, and economic growth remaining anaemic, you might think it was not a great time to increase interest rates.
You would be wrong. The ECB increased rates by 0.25% this week.
There is something faintly absurd about responding to a largely external shock by deliberately destroying more domestic demand.
Dearer oil already acts rather like a tax. It makes consumers poorer, cuts discretionary spending and squeezes business margins. Higher interest rates then pile on a second dose of demand destruction.
The central-bank defence is familiar: stop higher energy costs feeding into wages and wider inflation. But the argument still requires workers to have enough bargaining power to secure inflationary pay increases, and companies to have both the ability to pay them and the confidence to pass those costs on through higher prices.
That is a pretty flaky set of assumptions to justify inflicting still more pain on your citizens.
The more obvious danger is some degree of stagflation: higher prices, weaker growth and another hit to real purchasing power that households may never fully recover.
There are uncomfortable historical echoes.
In July 2008, with oil near $150 and economic conditions already deteriorating, the ECB raised rates because it feared second-round inflation. It then had to slash them after the financial crisis erupted. In 2011, it raised rates twice, in April and July, just as the eurozone debt crisis was intensifying. Both rises were reversed before the year end.
It’s not just the ECB. The US Federal Reserve committed arguably the opposite error.
Its emergency response to 2008 was necessary. What is much harder to defend is why emergency action remained in place for so long after the fire had gone out. The Fed held its policy rate essentially at zero from December 2008 until December 2015 — 5 years beyond the necessity.
This largesse served to encourage investors to believe that money would remain almost free indefinitely, pushing investors into riskier and riskier assets to get a return, and laying the foundation for the valuation extremes and speculative behaviour we are dealing with today.
Given the hundreds of PhD economists employed across central banks, their collective track record of failure is extraordinary.
Unsurprisingly, no major stock market had made much progress this week. In the US the S&P 500 and Nasdaq were each down about 1.6%. The FTSE 250 fell 2.8%.
In commodities, oil was the glaring exception, up around 10%. Much of the rest of the universe weakened, from copper through to uranium and palladium, with losses typically spanning 2–6%. Remember, though, that this is a perfectly normal level of volatility for commodities.
The weeks just ahead are likely to remain nervous for markets.
Historically, Autumn has heralded some deadly times for markets, and it is hard-wired into the gloomiest commentators to foretell immediate doom every September and October.
Yet I also notice that the S&P 500's uptrend has something of a “complete” look about it. Nothing more than that. It is an observation, not a forecast.
Just keep your head in the period ahead while others are losing theirs.