🔮 AI, jobs & Engels’ pause; silicon sensitivity; GPUs galore; artificial milk, cheap batteries ++ #457
An insider’s guide to AI and exponential technologies
Hi, I’m Azeem Azhar. As a global expert on exponential technologies, I advise governments, some of the world’s largest firms, and investors on how to make sense of our exponential future. Every Sunday, I share my view on developments that I think you should know about in this newsletter.
Latest posts
If you’re not a subscriber, here’s what you missed recently:
🔮 Lifespan, AI in healthcare, GLP-1 drugs, and mRNA with
🔮 The horizon for 2024 #3: The biggest questions on the horizon
🔮 The horizon for 2024 #2: Small technologies & decentralisations
Sunday chart: Great expectations
This is the year that genAI gains mainstream acceptance in the workplace (i.e. executive approval). Talking directly with more than thirty CEOs and senior executives in Davos, I was surprised to see how fast they have acted. The overwhelming majority had generative AI-based projects underway. In one case, a CIO reported 26 GenAI projects in deployment.1
EV reader, John Romeo, led research for Oliver Wyman that surveyed 25,000 workers in 17 countries. Half of all workers were already using generative AI tools in work, with rates much higher in India and the UAE than in the US.2
There is a pincer movement going on. Top execs are excited by genAI and front-line workers are using it, coupled with the comparative ease of integration (certainly compared to cloud migration, for example) this pincer could squeeze some of the internal resistance that stops large firms acting quickly.
Let’s not pretend this doesn’t create tensions. Older workers might find it hard to adapt, and middle managers could face a mid-career crisis, realising that their younger colleagues armed with AI make them redundant. The likely outcome is downward wage pressure: junior workers with AI competing with more expensive older workers for one, and the automation (and thus decrease in cost) of many cognitive tasks.
We might be entering an Engels’ Pause — a phase where economic and technological advancements precede improvements in workers’ living standards. This mirrors the early Industrial Revolution, where initial industrial expansion did not immediately enhance workers’ conditions, highlighting a disparity between technological advancement and societal progress. (Sam Altman and I discussed this very issue back in May when I raised the spectre of “Altman’s Pause”.)
Avoiding an Engel’s Pause might be possible. If economies adapt rapidly, workers might have the option to switch from roles where wages are declining to better ones. Policy interventions might also help mitigate this but could be expensive for governments, and lack political support.
Key reads
A lesson in empathy. A new study in Nature shows that AMIE, a Google chatbot, diagnosed heart and lung conditions more accurately than doctors in online healthcare. More surprisingly, the chatbot ranked higher on empathy - a trait considered beyond AI’s reach. In my recent chat with
, we discuss how LLMs could coach doctors to have more empathetic interactions with their patients. Yet this new paper suggests that maybe, at least in telehealth, doctors best leave the talking to chatbots.See also:
Rie Kudan, who won Japan’s top literary prize, said ChatGPT wrote about five percent of her novel ‘Tokyo-to Dojo-to’ (Sympathy Tower Tokyo). She acknowledged that it helped unlock her potential.
Google DeepMind’s AlphaGeometry has set a new standard in AI by solving Olympiad geometry problems at a Bronze medalist level, using a neuro-symbolic architecture for enhanced reasoning.
P*ak Ch**a? China’s demographic and economic challenges continue (for a breakdown, see our Chartpack on the topic). The country’s population decreased for the second consecutive year and its GDP shrank in dollar terms. China continues to struggle with youth unemployment, deflationary pressures, and regulatory crackdowns. In response to these challenges, expect President Xi’s government to further tighten its control over information. Censorship reaches far: economic commentators often hold back their words due to fear of backlash; and language models are not spared either.
Moore chip designs. The escalating demand for GPUs, in tandem with the immutable laws of physics, is catalysing a surge in startups championing innovative computer chip designs. Among these, Vaire Computing, led by EV member
, is developing chips based on “reversible computing.”3 This method, unlike conventional computing, does not destroy information during calculations, potentially offering a more efficient computing approach.Glasses and AGI. Mark Zuckerberg and Meta have doubled down on their commitment to open-source AI, not just for language models, but also AGI. Meta’s incentive for open-sourcing AGI is enigmatic. It could be a PR move to gain public support and build an ecosystem they can direct. Another possibility is that Meta’s core competencies have never been in business products, unlike Google and Microsoft’s. Meta’s primary strength lies in its social media platforms, driven by user-generated content. Open-sourcing AGI could be a means to enhance this content, making their platforms (including the metaverse) more engaging and attractive to users and advertisers. Zuckerberg also revealed that they’re developing Llama-3, supported by a massive amount of computing power. He reports that Meta will have 350,000 H100 GPUs by the end of the year. To put that in perspective, there are only expected to be 2 to 2.5 million H100s in the world by year’s end. Facebook will control 14% of all active H100s, highlighting the increasing concentration of compute power, and huge economic moat for other firms. Am musing on what this means for smaller foundation model peddlers like Anthropic, Cohere and Mistral…
See also:
Nature highlights contradictions in Europe’s AI act. Low-risk AI has low transparency needs. But to know if the risk stays low, we must trust the AI’s makers, despite limited openness.
OpenAI has formed a team to include public opinions in governing its AI models. A nice gesture, best viewed with a heavy dose of PR cynicism, just to be safe.
MIT Technology Review examines what’s in store for AI regulation in China this year.
Market data
While remote work has reshaped society, there appears to be little sign that it has greatly sped up or slowed down productivity growth.
Microsoft blocked 4000 identity attacks per second last year.
Private equity ownership of hospitals was associated with a 25.4% increase in hospital-acquired conditions like falls and infections, despite treating a younger, lower-risk patient group.
Apple takes over Samsung as the world’s largest phone maker.
$300m. The value of Apple Vision Pro units sold in the first couple of hours of pre-availability. The entire run of roughly 80,000 was snapped up.
$56. The likely price per kWh for lithium batteries based on LFP. (In 2013, the average cost was 14 times higher). Exponential.
1 for 1. Imagindairy reaches price parity with cows for its precision-fermented milk. (See our 2020 briefing on precision-fermentation alternatives for dairy proteins.)
At over 3.3 TW, more than 40% of installed electricity generating capacity is now renewable.
Short morsels to appear smart at dinner parties
🐵 In medicine: FDA approves the first cell-based gene therapy to treat sickle cell disease. And, scientists in China managed to create the first healthy cloned rhesus monkey.
➗ DeepMind’s AlphaGeometry solves 25 of 30 Mathematical Olympiad problems, surpassing the previous SOTA approach for geometry.
🤒 How protectionism and mis-aligned policies keep old containerships sailing.
🌿 Scientists capture a video of plants warning fellow flora of a potential danger.
🧠 The age-old scientific debates around consciousness. (I would strongly recommend reading EV member Anil Seth’s book Being You: A New Science of Consciousness.)
🥷 The story of a young researcher uncovering the reality of Bitcoin’s traceable transactions.
End note
64,480 steps in snow boots in just three days last week. I have sore feet and a full head. Many exceptional moments from Davos, which I’ll write about in due course. I learnt a new acronym too good not to share today: “FICTA” - failed in crypto, trying AI.
Have a great week!
A
What you’re up to — community updates
Peter Schwartz recently published a series of opinion articles in the Chicago Tribune about urban-rural dynamics in the United States and their implications for addressing climate change.
Luca Taroni has published an innovation assistant GPT free for use.
Rumman Chowdhury was profiled in Marie Claire, covering AI bias issues.
George Cameron launched AI Analysis to benchmark and compares AI LLM models and API hosts across key metrics.
Jeremy Jurgens shares a fascinating case study of how small-scale Indian farmers are using AI to double their incomes.
Share your updates with EV readers by telling us what you’re up to here.
I’ll be sharing more in my Davos update coming soon for subscribers.
My favourite statistic is 97% of workers think GenAI can help them but only half of workers are using it. The missing 47% suggests either restrictive corporate policies or employees who aren’t showing initiative. I wish we knew which firms were overrepresented in that 47% so we could short the stock!
Disclaimer: I am an investor in Vaire Computing.
Small typo: Fellow flora rather than fauna I believe.
I heard of FICTA on an AI podcast recently, tickled me too. Having watched Zuck on Lex Fridman, I suspect this is trying to find a new business model, given the irrelevance of big tech in the face of AI, like Google and Perplexity. No ads there.