Chronicles of the Future – 15 November 2024

All week, AI commentators have been debating one thing: have we hit the ceiling? According to inside reports, Orion (codename for the attempted GPT-5 release from OpenAI) is not significantly smarter than the existing GPT-4 (released early 2023). This was echoed by investor Marc Andreessen, who said last week: “we've really slowed down in terms of the amount of improvement ... we're increasing GPUs, but we're not getting the intelligence improvements, at all."

Think about how quiet OpenAI has been? In August, I wrote about how Altman becoming something of a recluse risked greater regulation. But what if it just means he has nothing to show for his efforts. As Substacker Erik Hoel said this week: “OpenAI are now regularly not releasing internal models to the public, almost certainly because they don’t work as well as the hype claims. For instance, the simple explanation for why OpenAI’s video-generating AI Sora isn’t publicly available—after nearly a year!—is because it sucks.”

In response, Ilya Sutskever, co-founder of OpenAI, told Reuters that he thinks AI labs will need to train smarter, not just bigger, to progress. But this runs counter to the argument CEO Sam Altman put forward in his essay ‘The Intelligence Age’ less than two months ago (which I covered at the time), where he said all that was needed to scale AI was “more compute and more data.”

Altman is sticking to his guns, tweeting: “there is no wall,” and commenting approvingly on another tweet which said: “scaling has hit a wall and that wall is 100% eval saturation.” I.e. once an AI nears scoring 100% in a test, there isn’t much room left to improve.

Only time will tell who is right, but assuming the ceiling does exist, it very quickly transforms the conversation from one of racing towards superintelligence to trying to squeeze productivity gains out of an improved probabilistic search function.

Meet Daisy. The AI granny created by O2 that’s fighting back against phone scammers by wasting as much of their time as possible.

Here’s a clip of Daisy talking to some scammers.

Unless the scammers can find a way of immediately determining they’re speaking with an AI (which could well be impossible in the long run) this business model may be over. No matter how big the call centre is, it won’t be able to compete with telecoms companies flooding the lines with infinite Daisys.

COP29, currently being hosted in Baku, Azerbaijan, has already had its fair share of controversies. The President of the fossil fuel-rich country kicked off the environmental conference by calling oil and gas a “gift of God” as he cracked down on climate campaigners, and countries such as France and Argentina withdrew from the conference for various reasons.

It’s an unfortunate distraction from what should be the focus of the conference: climate finance. Globally, billions of dollars need to be channelled into building clean energy systems, adapting to a warmer world, and responding to now inevitable climate-induced disasters.

In 2009, developed countries agreed to “mobilise” $100bn of climate finance a year by 2020 – an annual target that was meant to run through to 2025. This became a fraught topic, as developed nations missed the 2020 deadline and only reached it two years later in 2022. According to the Paris Agreement, COP29 is the moment to agree on a new target.

How much? Who pays? And in what form? These are the key questions delegates and diplomats will be negotiating this month.

There is a general acceptance that payments need to rise dramatically. The UN Standing Committee on Finance has suggested that $5-6.9tn is needed over the next five years contains, while the US and EU have come out with similar ~$1tn a year proposals. That’s a ten-fold increase.  

Then, there’s the issue of who pays. Currently, only 23 countries are obliged by the UN to provide climate finance, made up of western Europe, the US, Japan, Australia, Canada and New Zealand, as well as the EU as a separate bloc. Many of these countries want China, Russia and the Gulf states to be added to the mix.

Lastly, there’s the matter of what form the finance takes. Recipients of the money generally prefer government grants, as they’re certain, easy to measure, and goes through their governments, while donors generally prefer a blend of public and private and a variety of mechanisms and instruments.

Carbon Brief’s breakdown of all the numbers and expected flash points is excellent, but it’s hard to escape the conclusion that all of this is just fantasy. Obviously, the world isn’t going to spend $1tn a year on climate finance without very generous definitions of what it constitutes. The political momentum is all in the other direction, and that’s before Trump takes office again (the US hasn’t been close to meeting its obligations at any point).

Russia is banning all content that “promotes childlessness” in an attempt to boost its flagging birth rate, threatening fines to individuals and businesses who flout the new rules. This marks an extension of President Putin’s moves to boost the birth rate through culture. For example, independent Russian media have reported that the reality television show ‘Pregnant at 16’ has been edited to omit mentions of abortion, and almost every episode has a happy ending.

Putin has been conscious of Russia’s demographic challenges since he took office. In his inaugural address in 2000, he warned that the country could become an “enfeebled nation” due to population decline.

The long shadow of the Soviet Union and its collapse have left a mark on Russia’s fertility. Economic decline and poor health, particularly among men (caused by sky-high rates of alcoholism and workplace accidents), damaged the country’s demographics. Less known is that women’s lack of access to birth control and Russian men’s resistance to condom use led to the abortion rate averaging a horrific seven per women during the Soviet era. The result was that Russia’s fertility rate stood at just 1.3 in 2004.

After declaring that demography was “the most serious problem in Russia today” in his 2006 address, Putin introduced a system of “maternal capital” to boost birth rates. This took the form of payments to women who had a second child (or adopted) that could only be spend in certain forms, such as housing or education. Although Russia’s fertility did recover, reaching 1.8 in 2015, most experts attribute it to an unusually large parenting-age population at the time and the economic impact of the oil price boom. By 2019, it had fallen back to 1.5.

Although the Russian government has claimed fertility is 1.8 for several years, reports in September revealed it is hovering just above 1.4. That’s on par with several Western countries, but the effects in Russia will be far worse.

Firstly, Russia has negligible immigration to offset the effects of a low birth rate. Despite experiencing an immigration boom upon the breakup of the Soviet Union (1.1m in 1994), years of restrictive immigration policies means that fewer than 100,000 people have moved to Russia annually for over 20 years. Gaining 2.4 million people in Crimea doesn’t make up the difference.

Secondly, Russia has an emigration problem. Since the war with Ukraine started, it’s thought more than 1 million (mostly young) Russians have left the country. This dwarfs the estimated 70,000 Russian men thought to have died in the conflict. Curiously, Russia’s Science Minister has said that an equal number (70,000) of scientists have left Russia since the war began, making Russia the only developed country in the world where the number of scientists is shrinking.

Using UN population estimates (which doesn’t seem to have incorporated the impact of the war), the global population will decline by about 20% by 2100, but Russia’s population will fall by 25-50%. Given the impact of the war, you would imagine it will be towards the upper end of that estimate. That’s a significant reduction in power.