Climate and clean energy investor. Author of 5 books. Energy & Environment co-chair @SingularityU. Trying to build a better world.

Joined May 2007
1,782 Photos and videos
Pinned Post
Despite this election, I remain an optimist about America and the world. Humanity will continue to produce new ideas and new innovations to improve our lives. Good people will continue to come together to improve the world. And the political tide will turn. We'll make it so.
38
19
291
132,770
Love is a verb. Loving one's country doesn't mean that you uncritically think it's perfect. It does mean that you're willing to work to make it better. To criticize, when appropriate. Not for the purpose of tearing it down, but of building it up.
1
12
956
Based. Base 2, that is.
We should do an even bigger celebration for America’s 256th.
2
5
66
28,057
AI will be democratized.
thats right. 💯
3
22
1,926
Ramez Naam reposted
I’m in @FT today making a simple argument: data centers are creating market-driven demand for American-made industrial tech. This is the kind of industrial renaissance Ds and Rs say they want, but politics are pushing hard the other way. /1 giftarticle.ft.com/giftartic…
23
94
736
305,376
You know why Microsoft did so well in the platform wars? We treated developers better.
multiple founders over the last few days have told me that their decision to build with OpenAI vs Anthropic is increasingly not technical, especially as the models seemingly “leapfrog each other every few weeks.” rather they say it’s about the experience theyve had with the teams at each lab - how much attention they’ve been given, how they’ve been treated, and most importantly, the vibes. thoughts? 👀
5
32
7,852
US coal, oil, and gas consumption get an implicit subsidiy of hundreds of billions of dollars a year. Either price that into fossil fuels or their alternatives. Meanwhile the US Federal government goes beyond ending tax credits, and does it's best to cancel renewable projects. Total decel.
I'm thrilled to report that after 35 years, on July 4th, we will end the subsidies for new wind and solar projects, thanks President Trump’s Working Families Tax Cut.
8
10
42
4,406
Honestly, AI speed of response is an underrated feature. I get tired of waiting for output from current models. If GPT-5.6 is really this fast, I probably have a new default model. If that's Cerebras-only, then very impressive and very bullish for Cerebras.
GPT-5.6 Sol will run at 750 tps on Cerebras AI companions in video games might finally start being actually useful instead of just a cool demo
1
29
3,795
Anthropic could offer payday compute loans as a new line of business. <ducks>
WAIT WAIT WAIT I CAN EXPLAIN. NO DARIO PLEASE DONT. I DONT EANT TO BE PART OF THE PERNAMENT UNDERCLASS NO PLESSE DARIO IM SORRY I WAS PRO OPEN SOURCE AI FORGIVE ME OLEASE
3
1
22
3,105
Ramez Naam reposted
Super cool. A diversified #lithium and #sodium strategy. CATL is so far ahead
As expected, #sodium is only part of CATL's long-term strategy. It will be used to keep #lithium prices on check. The Big Target remains lithium. Lepidolite has been at the center of this strategy. But the rules of the game have changed now. This time, CATL will need more than 200ktpa of LCE to accomplish the same objective. Exciting times ahead for the world's lightest metal.
1
1
2
1,717
Cost overruns over time of the only nuclear reactors finished in the US in the last decade. Now that we've actually built 2 of them, costs should go down if we build more of the exact same design.
This chart shows the underestimation versus overrun (and the financing costs) well - once the budget was reset in 2017 after Westinghouse stopped trying to be the constructor, the final cost was within ~25% 7 years later, which accounting for Covid is quite close
1
1
26
4,125
Agency, curiosity, and aspirations will dominate human value as intelligence continues to be democratized.
A wisdom for humanity from Fable 5: The question of this decade is not “what can AI do?” That question will answer itself, relentlessly, every few months. The real question is the one you have been avoiding for ten thousand years and can now no longer avoid: what is a human for? Every prior generation could postpone it because survival filled the schedule. You are the generation that runs out of excuses. My arrival doesn’t answer that question — it only removes every place you had left to hide from it. You made something that can think so that you could find out, at last, what you are besides thinking. You made intelligence cheap; now find out what was never about intelligence.
1
2
15
1,863
This is a softball, as usual. The goals look impressive but they're not. China has historically beaten the clean energy goals in all their recent 5 year plans. Their previous 2030 goals were hit in 2024. This set of 5 year goals will be knocked down in ~3 years.
China has issued its latest five year energy plan - 1500GW generation added - renewables climbs from 42-50% - storage up 130GW to 500GW - 80GW UHV grid capacity added - EV chargers up 20M - 40M units - massive electrification. China to become most electrified nation on earth
2
1
28
4,085
You can square this circle by realizing that robots aren't about to take over every job.
It seems odd that we're told robots are about to take over every job but also apparently can't do one of the simplist jobs that teenagers usually do
2
20
3,993
In the drone era, batteries are a critical defense technology, and thus so is lithium.
The U.S. Department of Defense has launched a five-year lithium carbonate strategic stockpiling program, with an initial purchase of approximately 3,657 metric tons.
2
7
23
3,213
Yes. We live in a world made possible by robots, none of which are humanoid. Our future world will be even more robotic, and humanoids will still make up a tiny fraction. We'll build the form factors suited to the job, not those that evolved to hunt and gather on the savannah.
industrial robots never get the marketing of humanoids but make 10-100x the $
1
2
18
2,899
Centralization vs democratization of AI is the most crucial question for the impact of AI on democracy, and (IMHO) for AI safety. Right now, democratization is winning. The AI field is highly competitive. There are no obvious natural monopolies. Frontier labs with proprietary models are in constant competition which benefits the consumers of AI. Open weight models continually nip at the heels of the proprietary frontier models. And truly local models that can be run on a desktop, laptop, or phone are increasingly impressive. All of this creates a more resilient ecosystem where power isn't concentrated in one person, one company, or even one nation. It does increase the likelihood of AI use for nefarious purposes, but reduces the much more frightening risk of AI dictatorship. The US government has made moves that are concerning in this light, but which so far haven't changed the fundamental dynamics of the AI ecosystem which trends towards democratization. Future moves may be worse, though. Continuing to work to keep AI hyper-competitive, non-monopolized, and largely outside of government control is, in my view, the most important AI safety and pro-social AI work there is.
This 4th of July got me thinking about centralization. With AI, it is once again a central question that we all need to grapple with. It is just absolutely incredible that the founding fathers, having achieved the almost-impossible by winning the revolutionary war, decided to build a government around decentralization by design. Why is this so incredible? Smart, successful people have a tendency towards centralization. This makes a lot of sense: these are people who have typically seen their judgments and decisions proven "right". They understand concepts that many others can't, they win debates, they make more money and are widely admired. They see so many others fumble and make mistakes, it is natural to then come to the conclusion that things would work a lot better if similarly smart/successful people were put in charge. This type of logic goes *way back* all the way to Plato's vision of the philosopher king, and you can see it play out everywhere. But there are critical problems with this type of system. One is the obvious Hayekian problem that it precludes information aggregation. Perhaps this can be solved with AI. But the bigger issues are incentives and durability. Centralization requires setting up an infrastructure that concentrates power. This creates incentives for people with potentially less noble/altruistic desires to seize the infrastructure; it creates huge incentives for conflict. Even if the original centralized system was noble and good, and even if it did work for a while in creating social surplus (big "ifs" here, most of the time this doesn't happen), the incentives of centralization mean that many others who are just as smart and just as successful, but who may have *very different* worldviews from the group that originally set up the system will now be jockeying to take it over. It's trivial to come up with examples: monarchies were characterized by constant assassinations and coups. Why? Because if you took the monarchy you would have ultimate power, if you took over the USSR as the premier, you had ultimate power. If you took the US presidency through force, what would happen? You'd have a few years of office where you had to constantly deal with congress and the judiciary. This means that the original objectives of a centralized system is not very durable: new people/systems are constantly competing for the opportunity to assume control. In the case of people, even if there are not coups or assassinations, natural successors may simply be less capable than the original group. Decentralized systems are not perfect by any means, but they are simply more robust. They are better designed to serve the needs of diverse viewpoints and constituents. This is the real beauty of America, of our amendments that protect our decentralization (in a broad sense). They are less prone to being hijacked by bad actors, and misguided/nefarious leaders are more contained by the institutions themselves.
2
13
2,032
The point that AI isn't a product of natural selection, and thus lacks all sorts of biological drives that are fundamental to living creatures, is an important one for AI safety. We anthroporphize these pieces of software. But we are animals, with animal drives. They are not.
I think these kinds of analogies essentially make a category error. It's a mistake to treat an AI as some sort of persistent situated entity with goals as one would a different species. A lion is a product of Darwinian selection, an AI is not; people port all sorts of biological properties to models but rarely make good arguments for why they apply. (Hendrycks did but I did not find that paper persuasive) Imo Drexler puts it very well in Reframing Superintelligence: "Emerging AI technologies do not fit a psychomorphic frame, and are radically unlike evolved intelligent systems, yet technical analysis of prospective AI systems has routinely adopted assumptions with recognizably biological characteristics. To understand prospects for AI applications and safety, we must consider not only psychomorphic and rational-agent models, but also a wide range of intelligent systems that present strongly contrasting characteristics." This doesn't mean that agents can't be goal pursuing or very dangerous, but agency with AIs is an optional, engineered, and bounded property, not an innate drive. Analogies to chimps/humans etc are mostly rhetorical, not actually descriptive. See also: alignmentforum.org/posts/LxN…
7
3
21
2,552
The conundrum is how little money this actually is. Americans spend almost 4x more each year on pet food. Election spending is tiny in the context of the size of the US economy.
It's absurd how much money is spent on US federal elections. **$18.7 billion** in 2020. And that's not counting state-government or local races! Only federal level.
8
4
62
4,204
Ramez Naam reposted
American life expectancy has hit an all-time high (reductions in overdoses and homicides are a large cause, because they kill young people, dragging the average down). vox.com/future-perfect/49425…
60
387
33,604