Post-AGI money?💸 LLaMa 3🦙 FineWeb🍷 VASA-1🗣️ OpenCRISPR🧬 plausible tomorrows⤴️ inner cosmos🧠🔌dangerous memes💭 primary endosymbiosis🦠 by Jupiter🌌 #2024.16
The economic commitment of climate change
Welcome to this week's Memia scan across emerging tech and the exponentially accelerating future.
(A record number of videos in today’s issue I think, hope that keeps the signal/noise ratio high enough…let me know?)
As always, thanks for reading!
ℹ️PSA: Memia sends *very long emails*, best viewed online or in the Substack app.
🗞️Weekly roundup
The most clicked link in last week’s newsletter (4% of openers) was the wonderful Ministry of Imagination report / manifesto from Rob Hopkins, harvested from the 100 episodes of his From What If To What Next podcast. Go take a look if you haven’t already.
🎉ICYMI
Thanks to everyone who has pre-ordered my book ⏩Fast Forward Aotearoa. This was a pretty exciting moment last week…the first printed and bound proofs arrived ready for final check-through!
Pretty happy with the outcome, we’ve made a few final minor colour enhancements to ensure the highest printed quality, the first full production print run is imminent and will be shipping online orders the moment they arrive! A big shout out to Jason Lennie at Ledge Brand Agency (in the video) for all his expertise, hard work and support throughout the whole design and print process, getting from a massive Google Doc manuscript to this point, and once again to Sam Ragnarsson for his amazing art and graphics throughout, in particular the cover looks spectacular.
Memia readers can use discount code MEMIA5OFF for NZ$5 off if you order by Friday 26th April. Order here at https://ffwd.nz
💸Post-AGI money?
First of all this week: cast your mind back to November last year and the dramatic OpenAI board battle. As OpenAI’s *novel* corporate structure was pored over, this gem of a statement from the company’s operating agreement stood out:
“it may be difficult to know what role money will play in a post-AGI world”
Indeed. (But that hasn’t stopped Microsoft and others pouring literally billions of dollars into OpenAI and Sam Altman himself trying to raise US$7Trillion to bring about the next generation of computation…)
Indeed, just this week Microsoft invested US$1.5B in G42, an Emirati AI holding company based in Abu Dhabi, in a deal which, according to the NYT, was:
“largely orchestrated by the Biden administration to box out China as Washington and Beijing battle over who will exercise technological influence in the Gulf region and beyond”
So, what are they investing in? This question on X from James Campbell got my mind ticking over: What role will money play in a post-AGI world?
To recap, the classical functions of Money:
Medium of Exchange: Facilitates transactions by eliminating the need for a barter system.
Unit of Account: Provides a common measure for valuing goods and services.
Store of Value: Maintains its value over time, allowing users to save or defer spending.
Standard of Deferred Payment: Used for debts; allows for buying now and paying later.
Ultimately, as per Yuval Harari, money is a collective belief system based on consensus. There are lots of monetary consensus models: today’s money is “guaranteed” by a government (eg fiat currencies), Proof-of-work (eg Bitcoin) or Proof-of-Stake (eg Ethereum). Historically physical commodities such as Gold fulfilled some, if not all, of these functions.
But these functions are historically predicated on an economic system which assumes that:
humans are the only economic actors
humans are the ultimate arbiters of the “value” of goods and services
the sum of all human brains places a limit on the amount of intelligence (~computation) available to calculate prices and exchange rates
there are no hard (biophysical or energy) constraints on global economic growth
Clearly these assumptions are challenged by the arrival of AGI:
AIs may become (semi-)autonomous economic actors
AIs may place different values on goods and services than humans
AIs may become more intelligent and able to calculate prices and exchange rates far more accurately than humans, effectively pushing humans out of the financial system
Also, another (non-AI) assumption may be wrong:
Global economic growth may indeed come up against hard biophysical and energy constraints… in which case the vast majority of debt in the world may just default… potentially rendering existing money worthless as a store of value.
So given that current money sits on shaky AGI foundations, what might “money” look like in a post-AGI world? Here are five scenarios off the top of my head:
Reversion to a rare element-backed universal standard Universal money returns to being backed by control of a small number of rare elements (eg gold, silver, platinum…)
Universal energy-backed standard Money shifts to being backed by control over future energy reserves. (Your GPUs are worthless without the electricity to run them…).
Universal solar-system wide, computation-backed digital reserve currency Quantum-decryption-resistant Bitcoin or something similar, backed by computation alone is able to persist and provide a baseline for all prices between human and AI economies. Exchange rates from other currencies exist much as they do currently, guaranteed by humans (nation states, network states, banks, corporations…) and AI-owned and operated economic entities (whatever they look like!)
Separate Human-money and AI-money without exchange a common computation algorithm or other backing doesn’t cut it. Instead, there are distinctly different reserve currencies for the AI and human worlds, with two kinds of game-theoretic backing based upon human values and AI values. The monetary systems operate independently of (or at least in a master-slave relationship with) each other.
A rich ecology of moneys post-AGI, money mutates into millions of “evolutionary niches” based upon a rich ecology of “objective functions”. (For example the new MAHI token from Toha network places intrinsic value on biodiversity over conventional economic growth, completely unvalued by $.) Hence, there will be no "gold standard" or "universal bitcoin" but instead diverse “exchange rates” based upon trading against the objective functions themselves.
If I was forced to take a punt, I think scenario 4 is most likely. As an analogy: right now we don’t use money to do a lot of “trading” with the other species we share the planet with. Most of the time we just quietly coexist and let them get on with their lives as we get on with ours. But if we want to extract from the environment they live in, or pollute, or capture them or eat them, generally we just take with no compensation, financial or otherwise. Why would AGI treat humans any differently?
So the fundamental question is whether today’s giant AI investors believe AGI will be achieved. And if so, what will their investment transmute into thereafter?
(And the other corollary question is what happens to current money for the rest of us…?)
📈The week in AI
The week's AI news and releases… a relatively quiet week in the closed-source AI industry (for once).
💬TED does AI
Still no GPT-5 at the time of writing. But artist Paul Trillo and OpenAI put out this visually captivating, hypnotic, vivid, striking… but ultimately meaningless 93-second clip to promote a series of TED talks on AI.
(If the objective function continues to be “maximize attention” then we may well all be doomed…unless we wrap AI around us to protect our limbic system from hijack.)
Axios summarises the talks themselves: AI optimists crowd out doubters at TED. Including this provocative metaphor from newly-installed Microsoft AI CEO1 Mustafa Suleyman:
"we should start to think about them [AI] as we might a new kind of digital species..“
Here’s his full talk:
🆕New research and releases
🦙Llama 3 The week belonged to Meta, releasing two new versions (8B and 70B) of their flagship new large multimodal model Llama 3…both under open source licences! A massive 400B model is coming soon…and Meta may well open source that as well.
“Llama 3 models take data and scale to new heights. It’s been trained on our two recently announced custom-built 24K GPU clusters on over 15T token of data – a training dataset 7x larger than that used for Llama 2, including 4x more code. This results in the most capable Llama model yet, which supports a 8K context length that doubles the capacity of Llama 2.“
(15 trillion tokens… see below…)
Try it yourself at https://meta.ai (why do all AI chat interfaces look the same!?) Image generation coming soon, apparently.
Benchmarks are impressive - claims are that the 400B model will be GPT-4 and Claude 3 equivalent. Certainly the most powerful open source model now out there. You can download it and run locally on computer if it’s powerful enough, or consume it via API for cents on the dollar. (And it runs **FAST** on Groq hardware…!)
How Meta Built Llama 3 Meta’s Generative AI VP Ahmad Al-Dahle in conversation with BigTechnology’s Alex Kantrowitz.
Perhaps of most interest is Meta’s AI open source strategy: potentially releasing US$10Bn worth of R&D into the wild. Watch this interview with Mark Zuckerberg, speaking to the excellent Dwarkesh Patel, to get some insights into the commercial logic here. (Arguably by open sourcing LLaMa, Meta is challenging the ability for OpenAI, Anthropic and Google to monetise their AI investments… meanwhile Meta has a cash-churning revenue machine…
🍷FineWeb Probably entirely coincidental timing (ahem) but look what turned up on open source AI hub Hugging Face: Fineweb, a 15 trillion token AI training dataset:
(And interestingly…. with original URL attribution for every string… any fear of copyright lawyers now surely in the rear view mirror…)
🗣️VASA-1: From Microsoft Research, Lifelike Audio-Driven Talking Faces Generated in Real Time:
It also works for non-photographic inputs:
However, unlike already-to-market firms Heygen, Synthesia and others, Microsoft won’t be releasing this open source… or even into a product just yet…
“Our research focuses on generating visual affective skills for virtual AI avatars, aiming for positive applications. It is not intended to create content that is used to mislead or deceive. However, like other related content generation techniques, it could still potentially be misused for impersonating humans…
While acknowledging the possibility of misuse, it's imperative to recognize the substantial positive potential of our technique. The benefits – such as enhancing educational equity, improving accessibility for individuals with communication challenges, offering companionship or therapeutic support to those in need, among many others – underscore the importance of our research and other related explorations…
Given such context, we have no plans to release an online demo, API, product, additional implementation details, or any related offerings until we are certain that the technology will be used responsibly and in accordance with proper regulations.“
Viggle Controllable character animation using JST-1, “the first video-3d foundation model with actual physics understanding”. (Like we really need more dancing videos…)
Retell Probably the best customer-service voice chat demo I’ve seen to date. Latency is nearly gone. (🎩Asa Cox for sharing…)
🔮[Weak] signals
Tech signals from near and far futures...
⤴️Plausible tomorrows
Legendary investor Vinod Khosla, shared his list of “Plausible Tomorrows 2035-2049” at the TED conference, giving off very techno-optimist vibes indeed…
“Expertise Will Be Free: By 2035, AI-based tutors and doctors will be accessible and nearly free for everyone.
Labor Will Be Near Free: Bipedal robots and other automation technologies will eliminate the need for human involvement in undesirable jobs.
Expansive Growth in Computer Use: Over a billion people will be programming in natural languages, making computers more adaptable to human needs.
AI in Entertainment & Design: Personalized music and entertainment will be abundant, transforming but not eliminating the celebrity-consumer relationship.
Internet Access by Agents: Most internet access will be managed by agents acting on behalf of users, managing tasks and privacy.
Clean, Dispatchable Electric Power: By 2050, fusion and superhot geothermal energy will replace traditional coal and natural gas sources.
Displacement of Cars in Cities: Cities will predominantly use on-demand, autonomous public transit systems, reducing congestion and improving throughput.
Faster Flying: Mach 5 planes will reduce the travel time between cities like NYC and London to about 90 minutes, using sustainable fuels.
Abundance of Resources: Advances in technology will ensure a continuous supply of essential minerals like lithium, cobalt, and copper.
New Food & Fertilizers: Alternate protein sources and "green" fertilizers will surpass the quality and taste of traditional animal products.
From Practice to Science of Medicine: Precision medical care will be tailored to individual genetic profiles and conditions, utilizing AI for personalized therapies.
Solutions for Carbon Emissions: Technological advancements will address carbon emissions in critical sectors like cement, steel, and agriculture.”
An interesting selection, nothing particularly new in there but worthwhile for a thought experiment of how tomorrow may indeed be better than today.
He also lists “What might slow down these predictions…?”
“Incumbent resistance
Politicians capitalizing on fear for personal / populist gain
Tech failures or delays
Financial market conditions may kill a good idea
Anti-tech sentiment Luddites hijacking the would-be advocate
A few bad AI-related outcomes that get sensationalized
Left field events Instigators/entrepreneurs may not show up to make it happen”
Amazing to me that “biophysical limits overshoot” is nowhere on this list, especially from one of the (on the face of it) smartest investors on the planet.
🛂Passport-free borders
The Singapore Straits Times looks out at the future of international travel, in particular the option for passport-free borders. Starting this year, travellers will have the option of passport-free departure clearance from Singapore Changi airport using just a QR code and biometric scanning.
🐒Speciation event?
Ben Thompson, Stratechery, the OG tech newsletter reviewed Silicon Valley industry reaction to Marques Brownlee’s brutal review of the Humane AI Pin: MKBHDs for Everything. As always a post full of insightful observations, but this throwaway quote at the end got me in particular:
"…I suspect humanity will be distributed bi-modally, with the vast majority of people happily wearing their Vision Pros or watching their streaming service or...TikTok... while increasingly sovereign individuals, aided by AI, pilot the ship."
(That, to me, is what might be called a speciation event).
🌐On the radar
(🎩 2 weeks in a row to Aimee W) The latest bi-annual ThoughtWorks Technology Radar report is full of useful “posture pointers” on the latest software and AI tech: 105 tools, techniques, platforms. languages and frameworks assessed and grouped as follows (click the link above for the interactive map and to download the report):
🤖Atlas 001
OMG. One day after retiring the hydraulic HD Atlas robot, star of many benchmark-breaking videos over the year…prominent robotics firm Boston Dynamics revealed their new Atlas *all-electric* humanoid robot… with freakishly articulated limbs:
Nice bit of swagger from Boston Dynamics’ head of Policy:
Related:
In a recent interview, NVidia’s Jensen Huang said that in the near future, humanoid robots will be available for $10-20,000. As affordable as a car.
💊Health tech
🥽Vision Pro for surgery An Apple Vision Pro has been used to help perform a live surgery for the first time - surgeon Bruno Gobbato successfully carried out a shoulder arthroscopy surgery in Brazil, showcasing features including high-resolution display, cinema-sized images of 3D models and patient scans (eg MRI images) in real-time, the native Notes app and screen mirroring. Here’s the first-person view:
AI-enabled full body MRI scans New York-based medical imaging startup EZRA Health has signed a partnership to offer its AI-enabled full body MRI scans that utilize AI to monitor for 500+ conditions in 13 organs across 150 locations in the US. If this service was widely available it could be a game changer in early detection and prevention of diseases.
Biotech roundup
🧬OpenCRISPR Biotech startup Profluent Bio announced the first ever successful precision editing of the human genome with a gene editor designed from scratch with AI:
“…during the training process, Profluent’s large language models (LLMs) learned from a massive scale of protein sequences overlaid with biological context to generate millions of diverse CRISPR-like proteins that do not occur in nature, thereby exponentially expanding virtually all known CRISPR families.“
Furthermore, they released their research under a new open-source licence which they called OpenCRISPR.
(Video: OpenCRISPR / Profluent Bio)
⚙️Flagellum unravelled Structural biologist Prash Singh and a team of researchers from the US have uncovered the assembly of the bacterial motor. It’s spectacularly elegant when animated:
(Video via @prash_singh)
Perhaps even more eye-opening, @DrNPhD on Twitch took it one step further and started building the rings in Minecraft:
🦠Primary endosymbiosis Scientists have observed an extremely rare evolutionary event: two lifeforms “merging”. The algae Braarudosphaera bigelowii fully integrated a cyanobacterium, UCYN-A, transforming it into an organelle termed a “nitroplast”. This fusion, which aids in nitrogen fixation—a capability plants normally lack—could be a significant evolutionary advancement impacting future agricultural technologies.
(The only two other recorded “primary endosymbiosis” events like this are (1) the integration of a bacterium by an archaea around 2.2 billion years ago that became mitochondria, and (2) the absorption of cyanobacteria by advanced cells about 1.6 billion years ago, which evolved into chloroplasts. (And hence…plants!)
⚡CO2 in, electricity out?
Researchers at the University of Queensland (UQ) have built an electrical “nano-generator” that generates electricity as it absorbs CO2, pairing a poly amine gel already used to absorb CO2 with a thin "skeleton" of boron nitrate only a few atoms thick.
🖨️Form 4
Formlabs says its new 3D printer “rivals injection molding”:
🧠🔌Inner cosmos
A new BCI form factor: Inner Cosmos is a “half-way house” non-invasive brain-computer interface:
“a penny-sized device slipped under the skin. It rebalances brain networks using imperceptible micro-stimulations and without even going into the brain.“
Ahem. Apparently they are focusing on depression as their first target, aiming to provide
“a digital pill moves with you at your speed, eliminating the need for multiple, day-long trips to the clinic required by NeuroStatic approaches like TMS, ECT, and ketamine.“
Colour me sceptical, but I said that about Neuralink and look where we are now…
♻️Notpla
British startup Notpla makes edible, biodegradable liquid food packaging… from seaweed. Neat.
⏳Zeitgeist
Once around the world lightly...
🌡️Climate crisis weekly
Emissions, visualised. In case there was any doubt. (Would be good to get per capita animation as well…)
(Spot: “Aviation” drops out of the chart entirely in 2020… and then jumps straight back in 2022)
(Video: James Eagle)
🌧️Torrential in Dubai
The UAE witnessed a record-breaking rainfall in a 24 hour period, reaching 254.8 mm in less than 24 hours, surpassing Emirati meteorological data since records began in 1949. 4 people are reported killed. The coastal desert city of Dubai was particularly hard hit, with flooding across the city and airport: 1244 flights were cancelled over 2 days…
📉The Economic Commitment of Climate Change A new paper out in Nature examines projections of the macroeconomic damage caused by future climate change. Some sobering statistics which likely go some way towards explaining the economic malaise across much of the world:
(🎩 sharing Bernard Hickey)
“Using an empirical approach that provides a robust lower bound on the persistence of impacts on economic growth, we find that the world economy is committed to an income reduction of 19% within the next 26 years independent of future emission choices … These damages already outweigh the mitigation costs required to limit global warming to 2 °C by sixfold over this near-term time frame and thereafter diverge strongly dependent on emission choices.“
Arstechnica with more on the story: Climate damages by 2050 will be 6 times the cost of limiting warming to 2°
🌏Earth Day 2024 An IPSOS survey report for Earth Day examines public opinion on climate change across 33 countries, with some pretty grim findings about changing attitudes, particularly increased fatalism amoung the young:
🧐US tech regulation
Keeping an eye on:
Tik…Tok The US House of Representatives tacked on an amended TikTok ban to the Ukraine military spending bill and it passed by a 360-58 vote. The bill now goes to the Senate with at least 9 months for TikTok’s owners to sell and/or appeal. The US Presidential election is on November 5th - only 6.5 months away.
FISA 702 Less theatrical but far more significant, the FISA 702 Reauthorization Amendments, just passed, were an attempt to legislate for the US government to be able to compel more US companies and persons to assist in surveillance. Elizabeth Goitein, Edward Snowden and Vitalik Buterin were among those who sounded the alarm:
🧠Mind expanding
This recent paper by Joseph Merz, Phoebe Barnard and Rory Sutherland really struck a chord with me: World scientists’ warning: The behavioural crisis driving ecological overshoot.
“we … explore the behavioural drivers of overshoot, providing evidence that overshoot is itself a symptom of a deeper, more subversive modern crisis of human behaviour. We work to name and frame this crisis as ‘the Human Behavioural Crisis’ and propose the crisis be recognised globally as a critical intervention point for tackling ecological overshoot. We demonstrate how current interventions are largely physical, resource intensive, slow-moving and focused on addressing the symptoms of ecological overshoot (such as climate change) rather than the distal cause (maladaptive behaviours).“
Here are the team discussing the paper on a recent webinar, well worth 30mins:
🕯️RIP
Daniel Dennett, one of the world’s most prolific “thinkers about thinking”, died this week aged 82.
Gary Marcus shared a remembrance from Douglas Hofstadter:
“Dan was a deep thinker about what it is to be human. Quite early on, he arrived at what many would see as shocking conclusions about consciousness (essentially that it is just an emergent effect of physical interactions of tiny inanimate components), and from then on, he was a dead-set opponent of dualism (the idea that there is an ethereal nonphysical elixir called “consciousness”, over and above the physical events taking place in the enormously complex substrate of a human or animal brain, and perhaps that of a silicon network as well). Dan thus totally rejected the notion of “qualia” (pure sensations of such things as colors, tastes, and so forth), and his arguments against the mystique of qualia were subtle but very cogent.“
His writing was highly influential on my early development when I studied philosophy of mind at Uni. Particularly his 1991 book Consciousness Explained which was strongly physicalist and functionalist, still quite revolutionary for its time, at least in philosophical circles:
💭He's also well known for his work extending and extemporising on Richard Dawkins’ original concept of the meme. (Which is what Memia is named after.) Here’s a classic Dennett talk from the early days of TED in 2002: Dangerous Memes.
Also worth a read is his short story about brains in vats and consciousness. (Thanks Byrne Hobart at TheDiff for sharing…)
🪦Future of Life Institute
Also in memoriam, Oxford University’s Future of Humanity Institute, spearheaded by AI and critical risks philosopher Nick Bostrom, closed down abruptly last week. Joscha Bach speculates why…
🚧The Line Saudi Arabia’s 105-mile long Line city has been cut a little short – by 103.5 miles. Only 1.5 miles are expected to be completed by 2030 and some contractors have started laying off workers. It was always divorced from reality…
🧘Memetic savasana
☢️The Bomb And The Cold War
I’ve been working my way through director Brian Knappenberger’s epic 9-part series Turning Point: The Bomb And The Cold War, which explores the story from the invention of the nuclear bomb right up to the present day in Ukraine and a more multipolar nuclear-armed world. Fascinating historical insights, recommended.
(For a more in-depth, visceral account of the collapse of the Soviet Union and its aftermath, lauded British documentary maker Adam Curtis series of films in 2022, TraumaZone about Russia in the 1990s after communism collapsed is totally worth the time investment. Covered in Memia 2022.43)
🌌By Jupiter
From 5 years ago, NASA’s Juno probe passing Jupiter… set to Gustav Holst’s The Planets suite. Incredible.
💕Tinder Girl
What would you call this new multi-channel artform? Truly outstanding work (view full screen to read the text messages…)
(via @TurnerNovak)
That’s another week’s learning out loud over with! 🙏🙏🙏 Thanks as always to everyone who takes the time to get in touch with links and feedback.
More again next week,
Namaste
Ben
Microsoft now one of those rare companies with more than one CEO! What’s that all about?