🔮 Foundational AI; verified Elon; culture clash; giraffes & mushrooms++ #368

Hi, I’m Azeem Azhar. I convene Exponential View to help us understand how our societies and political economy will change under the force of rapidly accelerating technologies.

💌 Share today’s edition by forwarding this email, or tweeting about it.

👋 If this email was forwarded to you – hi! – sign up here.

Today’s edition is supported by our knowledge partner, McKinsey & Company

Are you smarter than your computer? Tech advances mean there are now machines that can solve some problems better and faster than humans can. And games have played a surprising role in teaching computers to think like humans via reinforcement learning. The latest edition of McKinsey for Kids looks closer.  Don’t miss it.

🎙This week on the podcast: VC & deep decarbonisation

I spoke with Shayle Kann, Partner at Energy Impact Partners, about his fund’s approach to investing in climate tech. We take a close look at the five challenges of deep decarbonisation, and how to build the carbon management industry from scratch. Listen to our conversation here, or read the highlights here.

The near future

🕹️ The building blocks of AI

“A photograph of dinosaurs with backpacks walking to kindergarten” Source: OpenAI

OpenAI and Google have both announced large AI models which drop jaws in their demos. OpenAI’s Dall-E allows simple language prompts (like the one above) to create images. (It also, less impressively, has rather strong ideas about what CEOs look like.)

Google responded with an even bigger (540bn parameter) model called PaLM. Like Dall-E, it can do impressive things in the lab. For example, PaLM can describe its reasoning in tasks such as choosing synonyms, working with counterfactuals, or explaining jokes.

When asked to spell out this joke:

Did you see that Google just hired an eloquent whale for their TPU team? It showed them how to communicate between two different pods!

PaLM responded with the following explanation:

TPUs are a type of computer chip that Google uses for deep learning. A "pod" is a group of TPUs. A "pod" is also a group of whales. The joke is that the whale is able to communicate between two groups of whales, but the speaker is pretending that the whale is able to communicate between two groups of TPUS.

One estimate suggests that, ignoring all the prior work or human time, PaLM would have cost $17m to train. Whether that’s a lot or a little is a matter of perspective. For an industrial investment that Google can apply across its services (or rent out through its cloud platform), $17m or even $170m seems like a snip. And bear in mind, we couldn’t have delivered something with PaLM’s capabilities for even a billion dollars only a decade ago… because we didn’t.

It seems like foundational AI models will become things we’ll increasingly be able to rely on. They lend themselves to the cumulative benefits of learning and data network effects. They will also be easier for developers to build upon. Could alternative patterns (such as decentralised models living on a blockchain computer) emerge? I mean, it’s entirely possible. Networks like NumerAI are demonstrating excellent performance in narrow fields like arbitrage. But for the coming years, I would expect such foundational models to be delivered by well-funded, centralised entities.

See also: a great essay by Erik Brynjolffson on the economic consequences of improved AI. This goes hand in hand with the demand that we’re “not to slow down technology, but rather to eliminate or reverse the excess incentives for automation over augmentation.”

🧐 The first Musketweeter
Would Elon Musk make a good owner of Twitter? The guy is a pretty remarkable entrepreneur. He is singularly responsible for innovating the electrification of light passenger vehicles. At this point, Tesla completely propels the EV market. Musk has revolutionised private commercial space travel, driving advancement forward (and costs back). He has proven he can galvanise engineers to solve hard problems, which the rest of us may think of as insurmountable.

Source: CSIS Aersopace

The funny thing is - that isn’t Twitter. Whatever Twitter’s issues are, they are not engineering problems. The fail whale is hard to find. Revenue is up 38% year-on-year and a range of new features from communities, to spaces, to Twitter Blue, show a freshness in the product range. Twitter is also tackling many thorny issues of machine learning at scale with culturally and politically sensitive information. Musk has acknowledged the importance of the public interest value of the platform, as he believes that “free speech is a societal imperative for a functioning democracy.”

It isn’t as if there hasn’t been widespread, open debate about the complexities that all social networks must navigate. This is a debate that has run on for the better part of a decade. And the challenges are not engineering ones - they involve difficult trade offs and clear multiparty communication. Given the importance of the information space needed for democracy to function, it’s clear that both rule of law and robust institutions are important factors to consider.

Musk’s free speech absolutism (or perhaps, rather mercurial and immature approach to tweets), were he able to turn that into product, would be more likely create a Gab—and certainly he isn’t someone, for all his capabilities, who has demonstrated any expertise in the knotty problems social networks need to tackle.

And I saw “were he able” because Twitter is an enormous organisation comprised of people who, for years, have increasingly understood they have a role in this private company that is also a bit like a public utility. He would have to persuade them to go maximalist (or fire them all). So it isn’t clear to me (that is British for “it is abundantly clear”) that he would be a natural steward of the firm. Perhaps a better approach would be to facilitate community ownership, rather than promoting a firm owned by Musk, Vanguard and Blackrock amongst others. EV reader, Trent McConaghy, has one such suggestion.

🧭 Cultural tectonics
David Brooks in The New York Times argues that after globalisation we are going to see global competition driven by differences in attitudes towards “freedom, equality, personal dignity, pluralism, and human rights.” In other words, the ideas of modernisation theory where nations carve different paths but ultimately “get to Denmark” look weaker in the face of decades of evidence. Forty years of globalisation, of technical standards, global memes, frappuccinos and crossfit, did not succeed. The coming years of relative deglobalisation won’t either. Rather, we’ll see a series of blocs competing with different conceptions of the good rather than goods, of differences arisen from values rather than value. As Brook’s puts it, it’ll be “the difference between the West’s emphasis on personal dignity and much of the rest of the world’s emphasis on communal cohesion”, whipped up by power-seeking autocratic elites in different parts of the world. It’s a thought provoking piece. I recognise the key challenge - that global Big Macs do not a global campfire make. But I disagree with his framing in that this makes for a “global culture war”. What we have are differences of values, and the culture in which they are embedded and which they shape. The real challenge is where the appropriate enabling compromises and co-existences lie—and what is required to pursue shared goals. Thanks to EV member Roger Dennis for pointing me to this.

(Also relevant to this question of values differences are the Moral Machine Experiment (pdf) and the differences in social norms between rice-farming and wheat-farming societies, both of which have slightly different conclusions from each other and Brooks’ conception.)

Dept of our climate future

In every Sunday edition, we track key metrics that tell us a little about our shared climate future. Our member, Marshall Kirkpatrick, takes the time to curate a view of our climate future in this segment every week: “This week’s stories point to creative paradigm shifts.  That’s an essential part of closing the exponential gap between technology and society, but much easier said than done. I hope these stories will inspire us all to reconsider the way we look at the world around us, as part of our climate change mitigation and adaptation strategies.”

  • Advanced Market Commitment: Stripe, Alphabet, Meta, Shopify, and McKinsey have committed to buying nearly $1B in systems for permanent carbon removal from the atmosphere, in order to incentivise such systems being built at scale. The program is modelled after the advanced market commitments that governments have used to incentivise vaccine development. While given the unfortunate name “Frontier Climate” (arguably it was the “frontier” paradigm that underlied the taking of the land by colonial powers to extract the resources that were burned to cause our climate justice problems in the first place), the ambitious program notes that drastic decreases in emissions will not be enough, and carbon removal at massive scale is a global imperative: “While carbon removal has made significant progress over the past few years, it’s still not at all on track to get to the required scale,” the project’s website explains. “As of 2021, less than 10,000 tons of carbon dioxide have been permanently removed from the atmosphere – 1 million times short of the annual scale needed.” Bloomberg reporter Akshat Rathi noted that $2 billion in total funding for carbon removal has been announced in the past two weeks alone, with LowerCarbon Capital announcing a $350M fund to invest in carbon removal startups and Swiss startup Climeworks AG raising $650M to scale its carbon removal technology. Azeem’s comment: “Pushing us down the learning curve delivers widespread societal and economic benefits. Governments should do this too! P.P.S. Chris Sacca and the LowerCarbon team announce a new $350m venture fund targeting carbon removal. I’m a small investor.”
  • Rapid growth of wind and solar: China is planning to install more wind and solar power this year than the entire world did in 2020, Bloomberg reports. According to Bloomberg NEF, 245 GW of solar will be added globally in 2022, representing 33% growth year over year. For context, IRENA says there is now 3,064 GW of renewable energy capacity worldwide, 40% of this coming from hydropower. IRENA reported this week that in 2021, 81% of all new energy generation capacity added around the world was from renewable sources - 88% of that from wind and solar.
  • Global Rewilding: The organisations Global Rewilding Alliance and OpenForests have launched a global rewilding map with 1 million square KM of rewilding projects documented around the world - that’s a collective area over 4 times the size of the UK. “We’re not going to reach our climate goals without a healthy dose of restoration and especially rewilding,” Vance Martin, president of the U.S.-based WILD Foundation said. Their rewilding strategy prioritises conservation of biodiversity, which requires that land use is prioritised for the survival of diverse plants and animals. As a direct result of this, land or water remains capable of substantial carbon sequestration. “You can put a carbon value on bringing back whales or wildebeests or beavers or sharks,” says the Global Rewilding Alliance.
  • Promises point below 2° C: A new international study published in Nature finds that if all nations on earth meet the commitments they’ve negotiated to reduce emissions, in full and on time, then “peak warming could be limited to 1.9–2.0 degrees Celsius.” That’s a huge “if” and 1.9° C still represents massive suffering and loss around the world. However, as influential climate investor Ramez Naam explains in a detailed Twitter thread, “This is the latest of more than a dozen studies over the last 3 years that have found that we’ve bent the curve of future warming down significantly.  For context, when I started in climate around 2010, expectations were that we would see 4 -  6 degrees Celsius of warming by 2100. That would be truly apocalyptic warming, with a severe chance of kicking off major feedback loops, and dramatic disruptions to civilization. We now see a line of sight to staying below 2 degrees C.”

Short morsels to appear smart while listening to fungi chatter

✂️ TikTok is cutting Russians away from the world, and further into Putin’s propaganda.

💰 After endless complaining about Apple taking a 30% cut from the AppStore, Meta has announced it will take 50% of revenue from creators on its forthcoming metaverse thing.

🦒 Somatic mutations are a cause of ageing, not a consequence. That’s why giraffes live longer.

🗺️ A digital map to track - and save - the rainforests of the oceans.

🍄 Fungi can communicate, and we can track it through electrical spike activity. A vocabulary as rich as 50 words. h/t EV member Gianni Giacomelli

🧫 Great Twitter thread on what it means to say “biology is programmable.”

⚡We also need more fast car chargers.

🚂 This new thermal heat engine with no moving parts could be more efficient than traditional steam turbines. h/t EV member Stephen Meyer

🏙️ AI could help detect earthquakes by filtering out city noise.

End note

We are super busy at EV right now. Marija’s team is flat out.

I’m looking for someone to help me on a short term basis with some research (next 2-3 months). The work is a grab bag. You will support me for some of the speeches I am giving in the next weeks. These range from industry-focused talks to discussions of defence, trade, democracy and innovation. I’ll also need help in triaging startup deal flow. It’ll really suit someone who may have come out of a generalist programme (management consulting, civil service, law). Obviously, fluent English is needed, as well as 2-4 years post-grad experience in a BCG, McKinsey, PwC, style environment where you are used to being quick to study across a wide range of different problems, know how to research and synthesise fast, analyse structured and unstructured information and use a broad set of communications modalities (slides, models, written word). The ideal candidate is available around 2-3 days a week over the next 2-3 months. You’ll need to be based in the UK or exactly in the UK timezone or 1 hour on either side.

This is a short-term role, as I said 2-3 months and not likely full-time, so is perfectly pitched for someone who is “in-between” or just doing some consulting while you work on your next project. You’ll need to adore high output delivery—and be willing to work to my tempo.

If you are interested, please leave your details on this form by Tuesday 19th April. I won’t be able to acknowledge receipt.



What you’re up to – notes from EV readers

Reshma Sohoni is on Forbes’ Midas List, recognising her as one of the best venture capitalists in the world. Congratulations!

Charlotte Harrison is sharing an opportunity with the UK government: the Office of Science and Technology Strategy is hiring a new Director. Deadline: May 2.

Joanna Picetti is setting up Planet Earth DAO to build a global government, based on a ledger that would include governments, companies and individuals.

Abhishek Gupta is starting a new position as Senior Responsible AI Leader & Expert at BCG and BCG GAMMA. Abshishek and his team at Montreal AI Ethics Institute are leading the way in the democratisation of AI ethics literacy.

Stephen Johnston explains the “dimensions of a DAO

CEO of OpenBack David Shackleton’s privacy-focused mobile edge computing notification platform is being acquired by Twitter.

To share your projects and updates, fill out your details here. Because of space constraints, we prioritise updates from paying members and startups I have invested in. (You can become the former by subscribing, if you have not already, and the latter by getting an intro to me via a trusted contact.)

How happy are you with today’s wondermissive?

Unhappy | Meh | It's good | Great | Extremely happy, will forward


Sign in or become a Exponential View member to join the conversation.