🔮 Nvidia sky high; conscious AI; China’s fall; brain-reading, modern allergies & geckos ++ #437

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 and my team prepare a preview of developments that we think you should know about in this newsletter.

---

Latest posts

If you’re not a subscriber, here’s what you missed recently:

1. 📈 Chartpack: The NVIDIA advantage

2. 🔮 Generative AI & economic growth

3. 📈 Chartpack: The potential of superconductors

---

Sunday chart: We’ve never really seen a revenue chart like this

“A new computing era has begun,” quipped Jensen Huang as his company NVIDIA stunned Wall Street with a record $13.5 billion quarterly revenue, driven by surging demand for its AI chips. This demand is fueled by firms rapidly prioritising investments in generative AI, with 75% of CEOs surveyed by IBM recognising that it can deliver a substantial competitive edge. In my conversations with senior business execs, it’s clear that large firms of all sorts are looking to put LLMs into production in specific use cases like knowledge management and customer services.

To build their these capabilities, firms are primarily leveraging cloud infrastructure. Underlying these cloud providers are thousands of NVIDIA GPUs. This surge in demand for compute-hungry AI services has significantly benefited NVIDIA, the leading GPU supplier, as evidenced by their quarter-on-quarter data centre revenue growth of 141%.

But acute supply shortages of cutting-edge chips like NVIDIA’s H100 mean availability severely lags behind demand. Though NVIDIA plans to triple H100 production to 2 million units next year, supply will in the meantime remain tight.

A supply squeeze and rising costs will encourage alternatives. Instead of training their own foundational models, which is a computationally-intensive process, firms may choose to fine-tune existing models such as GPT-3.5. GPT3.5 is much lighter on compute, and OpenAI recently allowed clients to fine-tune the model with their own data. Other approaches, like LoRA, which are parsimonious by the standards of building your own models, may become more popular. Companies might also veer towards lighter open-source models like Llama-2 which can approach the state of the art with fine-tuning.

Cloud providers may increasingly look at alternative chips to combat NVIDIA’s pricing power. AMD, which has made inroads in the server CPU market, could be one beneficiary.

To go deeper, check out our latest Chartpack 👇

---

Key reads

Does ChatGPT dream of electric sheep? Not yet. However, according to a paper co-authored by Turing Prize laureate Yoshua Bengio, “there are no obvious barriers to building conscious AI systems.” The authors outline an approach to assess consciousness in AI. By examining leading theories from both neuroscience and philosophy, they establish a set of ‘indicator properties’ for consciousness. Evaluating AI systems against these indicators provides insight into the potential for AI consciousness based on our current understanding. As we prepare for the emergence of more advanced AI systems, grasping the possibility of machine consciousness — and addressing the linked ethical and societal concerns — is of paramount importance.

See also: 

• Training multiple AI agents with unique chess-playing styles enables them to collaboratively solve a wider range of puzzles than a single AI system can on its own. This demonstrates how diversity and creativity enhance AI’s capacity for generalisation, much like they do for humans. 

• Meta released their latest AI model, SeamlessM4T, a multimodal system capable of translating both text and speech in almost 100 languages¹. 

Privatised progress or public peril? The New Yorker exposé on Elon Musk highlights the deepening intersections between private innovation and sectors of broad public consequence. From pioneering advancements in sustainable transportation with Tesla to aiming for global internet connectivity via Starlink, Musk’s endeavours highlight the transformative potential of private enterprise — and the enormous power that a private citizen can hold. They also surface challenges around accessibility, equity, and accountability. This dynamic is not unique to Musk — it resonates with the trajectory of other tech giants, many of whom have grown exponentially through network effects, amplifying both their potential benefits and associated concerns. If we look further back in history, we can go back to the 19th century U.S. and privately-owned railroads of paramount public significance. Ultimately, the government had to introduce regulations to safeguard wider societal interests. As private enterprises continue to shape pivotal sectors in the exponential age, the imperative lies in crafting new regulatory frameworks that balance private innovation with the collective good.

The Great Fall of China. Under President Xi’s grip, China’s once-remarkable economic growth is showing signs of strain, with rates dropping to 3.2%. As debates about the economy are silenced and data on youth unemployment and consumer confidence goes unpublished, concerns mount. China’s facing plummeting fertility rates and rapid centralisation of power, all challenging its future prosperity. After its meteoric rise, can China navigate the challenges of a slowing economy amidst increasing autocracy? To learn more about China’s demographic troubles check out our Chartpack on the topic.

---

🥁 Feature your product or service in one of the top newsletters on Substack.

Exponential View is one of the top technology newsletters on Substack and Linkedin with over 260,000 readers running some of the largest companies in the world, fast-growing startups, and innovation teams. If you are interested in sponsoring us, fill out this brief form.

---

Data

Salesforce expects global online sales in November & December to reach $1.19 trillion, which represents roughly 1.2% of global GDP.

Llama-2-70B performs at the same level as OpenAI’s GPT-4 in summarisation tasks, but 30x cheaper.

Global renewable energy investment reached a record $358 billion in 1H 2023, up 22% from 1H 2022. The surge was primarily driven by solar investment, which reached $239 billion globally—an increase of 43% from 1H 2022—with China accounting for half.

Investment in new chip production lines by 10 of the world’s top producers, including TSMC, Micron, and Intel, will fall 16% this year to US$122 billion, due to concerns over China’s economy and a supply glut.

Only 1.1% of VC-backed companies are fund returners—companies that return 100% or more of the committed capital of the fund that backed them.

Hugging Face, a company that hosts open-source models and datasets, has been valued at $4.5 billion.

---

Short morsels to appear smart at dinner parties

⛵Pioneering wind-powered cargo ship sets sail.

🤧 How modernity made us allergic.

😷 What do we now know about long COVID?

🧠 Brain-reading devices allow paralysed people to talk using their thoughts.

🕹️ How video games get built.

🧓 A new stage of life.

🦎 This gecko is so talented at disguise that for decades, it’s been mistaken for another species.

---

End note

I’ve started a new discussion in our community on issues relating to healthspan and longevity. We’ll be sharing the science and the practice in this group, and all members of our Exponential Do Slack can contribute. The conversation has already kicked off. Our private Slack is open to members who hold an annual membership to EV + apply via this form.

My first puzzler? Can I actually find the right size of Oura ring for my weirdly slim fingers?

Cheers,
Azeem

---

What you’re up to — community updates

• Chanuki Seresinhe is raising money for Educate Girls, a nonprofit that mobilises communities for girls’ education in the remotest parts of India, in memory of her son.

• Simone Vannuccini has published a study that assesses whether invention dynamics in certain advanced technologies, such as remote sensing (a technology now heavily reliant on data and AI), are directed towards addressing global challenges.

• Scott Smith and Susan Cox-Smith are publishing a book titled “Future Cultures” that provides a deep-dive into how successful organisations cultivate an anticipatory, future-minded culture through consistent practice.

• Chris Fellingham is organising “reSHAPE,” a free event at the Royal Society of Arts in London, which showcases university ventures rooted in social science, humanities, and arts research.

• Gianluca Sgueo has published his book titled “The Design of Digital Democracy.”

• Professor Mike Wooldridge has the distinguished honour of being the Royal Institution Christmas Lecturer, where he will delve into the topic of artificial intelligence.

Share your updates with EV readers by telling us what you’re up to here.

1. Depending on the task.