🔮 ChatGPT’s money pot; crypto VC; Indian Apple & Egyptian paper++ #408
Interest in ChatGPT is running hotter than the Nvidia A100 chips frantically serving up its responses.
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.
In today’s edition:
ChatGPT’s increasing popularity and how much OpenAI could make from it;
How technology gets cheaper;
What automation and digitalisation actually do to labour markets;
The EU’s Green Deal as a response to the American IRA.
Sunday chart: Keeping up with ChatGPT
Interest in ChatGPT is running hotter than the Nvidia A100 chips frantically serving up its responses. We search for it more frequently than we’re hunting out Prince Harry, Elon Musk and the entire Kardashian clan. As I analysed earlier this week, ChatGPT has a path to become the fastest-ever product to $100m annualised recurring revenues, potentially well above that. Microsoft’s search engine, Bing, is reported to be integrating GPT-4.
When I wrote about transformer models 2.5 years ago in A Short History of Knowledge Technologies, I suggested these tools would create a new abstraction layer over information by synthesising from different sources.
This is what is happening. This week, I spent 90 minutes in a discussion with ChatGPT exploring various niche (and not-so-niche) economic theories, their historical and cultural context, and potential limitations and extensions to them. As we dug deeper, I noted areas I would need to explore further. I researched those topics in traditional ways (Google searches, academic papers and articles). But what struck me was how much of my research time went to this chatbot, providing synthesised responses abstracted from the underlying sources. My brain was fried before the bot ran out of suggestions. I actually couldn’t keep up with ChatGPT.
See also: A watermark for chatbots can expose text written by an AI; Google CEO, Sundar Pichai, promises new AI features are coming to search ‘very soon’; Google also invested $300m in Anthropic AI, a firm founded by OpenAI alumni.
Them, Robots. This study suggests that automation and digitalisation require workers with more specialised knowledge. The complementarity of a worker’s specialisation with a firm’s needs is becoming more important than skill level. We’re starting to see the complicated picture of the automation and digital transformation of firms. Rather than the long-feared mass unemployment, it is leading to wage dispersion. Reskilling a misaligned labour force is on many governments’ minds, but 44% of respondents to the Edelman Trust study think that firms themselves should be doing more to retrain workers.
How technologies get cheaper. Researchers Malhotra and Schmidt have a helpful framework that identifies how increasing complexity and a need for customisation reduce learning effects. The corollary is true: modularity and standardisation help drive technology costs down. The academic paper points out how to understand which technologies will get cheaper over time, and what interventions can be made for slow learners (such as nuclear) to reduce their costs. (If you don’t want to read the paper,has a good interview with the authors. If you are even shorter of time, EV reader, Ramez Naam, has a similar view, summed up in this Twitter thread.)
The EU’s Green Deal Industrial Plan has been released in response to the US Inflation Reduction Act, which the EU was concerned would reduce the competitiveness of local industry. The plan addresses some key areas of concern: reskilling the workforce, increasing relations with critical mineral-rich countries and providing subsidies to keep production in Europe. One important pillar of the plan is creating a simplified regulatory environment. As we highlighted in a previous COTW, low regulatory barriers were a key factor in France’s rapid scale-up of nuclear power.
Azeem’s weekly commentary
This week, I explored how much OpenAI could make from ChatGPT as its freemium model was unveiled.
Bears don’t hibernate in crypto winter. Crypto venture funding has dropped by 91% YTD.
57% of Americans support a blanket ban on tobacco products.
Apple now has 2 billion active devices and around 1.2bn users. The firm averages $337 per customer per year, a fifth of that on services. Overall revenue declined by 5.5%. Sales in India saw double-digit growth last quarter .
35% of food orders at full-service (not fast food) restaurants are taken to go, double pre-pandemic levels.
Short morsels to appear smart at dinner parties
🪦Cory Doctorow: Enshittification - or how platforms die. Is TikTok next?
🚀 This year SpaceX will overtake the Soviet Union for number of space launches.
🗒 A new archeological find shows that paper might have been invented in Egypt, rather than China.
☕️ Drinking coffee is essentially borrowing energy from your future self.
⚛️ Taiwan’s nuclear energy dilemma.
🧮 Weird reciprocal: 1 / 99^2
Exponential knowledge community
Engage with the community of fellow members living and breathing exponential thinking:
Join our invite-only Slack by filling out this application [open on a rolling basis].
Members engaged on Slack can join local meetup groups for in-person gatherings: #meet-london, #meet-paris, #meet-dc, #meet-tokyo, #meet-berlin.
Join our global member mixer events on February 22 and March 29 [online].
What you are up to:
Pascal Finette’s book Disrupt Disruption is out! The book offers a holistic view of the dynamics of disruption.
Diana Wu David writes about what women need in the post-pandemic workforce.
Mark McLaughlin launched Fida, token issuance and redemption infrastructure for blockchain-backed loyalty programmes.
Jörgen Löwenfeldt is leading the latest project in Swedish museum Norsiska museet, with digital collections called Save the Future for the Future.