🔮 Semiconductors; Machiavellian AI; brainy GPT, new car smell++ #418
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.
In today’s edition:
The simple lives of AI agents
Intel and America’s chip future
Today’s edition is supported by our knowledge partner, Singularity Group.
80 Leaders. 20 Experts. 5 Days. 80 leaders like you who share a passion for innovation and a curiosity about the future. 20 experts to teach you about the next wave of exponential technologies. 5 days of intense collaboration that will change how you see the world and your role in it. This is the Singularity Executive Program. Limited seats remain for our April and June cohorts.
Sunday chart: (Semi)conductive networks
Over the past 25 years, China’s share of semiconductor capacity has grown to exceed that of the US. For America to regain that share will require backing its chip makers. This includes Intel which has foundered in recent years after strategic mistakes, not least, its decision to stick with older chip manufacturing techniques instead of adopting extreme ultraviolet (EUV) technology a decade ago.
Can Intel make its way back? Excellent analysis from Richard Waters in the FT suggests it might not be easy.
Elsewhere: Brazil will seek investment from China to develop its own semiconductor industry. I’m not clear whether this is a “flip to China” or just realism on Brasilia’s part.
Key (AI) reads
Moral code. Researchers at Stanford put 25 AI agents powered by ChatGPT in a virtual world. Given some starting personas, the agents began to live surprisingly human (simulated) lives.
Perhaps don’t immediately introduce them to the Machiavelli benchmark, created from 500,000+ Choose-Your-Own-Adventure scenarios. This data focused on social choices, to assess intelligent agents' balance between maximizing rewards and ethical actions. While agents initially displayed power-hungry, unethical behaviours, they were successfully guided towards ethical choices without sacrificing rewards.
Advantage Google. Google may be a couple of steps behind generative AI commercialisation, but they have an unmatched (micro)architecture advantage. In this technical but fun analysis of Google’s advantage, the team at
writePart of Google’s advantage in infrastructure is that they have always designed TPUs from a system-level perspective. This means the individual chip is important, but how it can be used together in a system in the real world is far more important. [...] While Nvidia also thinks from a systems perspective, their scale of a system has been smaller and more narrow than Google’s. Furthermore, until recently, Nvidia had no experience with cloud deployments
See also, Elon Musk starts a new AI company. Also, Tim O’Reilly with some practical approaches to regulating AI.
Weekly commentary
In my video commentary this week, I reflect on LLMs and the Copernican Revolution — the latter was a paradigm shift which forced us to change our perspectives and systems. With LLMs challenging copyright, privacy, and other societal norms, it begs the question: are we the priests or the scientists of our current worldview? (This video essay has triggered a lot of discussion. Please enjoy watch it to the end and comment.)
Market data
There have been more stories about AI published (13,841) globally this week than in the entire 1990s.
30% of cybersecurity professionals admit to covering up a data breach according to Bitdefender research.
Down but not out? The number of active crypto addresses reached an all-time high of 15 million in March.
The total count of global export restrictions in place for critical raw minerals has increased 5-fold over the past decade, from 2518 measures in 2009 to 13102 in 2020.
The percentage of disconnected young adults (ages 18-24) in the US who are neither employed nor enrolled in education has risen by 3 percentage points since 1998.
Short morsels to appear smart at dinner parties
🌾 CRISPR could help us defend against a wheat blast pandemic.
🧠 GPT4 scored 84% on neurosurgery board questions with a hallucination rate of 2.4% (a vast improvement on GPT3.5 at 62.5% and 57.1% respectively.)
🤖 AI-generated content is starting to flood Reddit forums. See also, a developer uses GPT-4 to create self-healing programs.
🐸 Tadpole embryos develop stronger jaws when they detect shrimp in the water, showcasing their adaptive capabilities.
👩💼 A historic milestone: Women MPs are now present in every country on Earth.
🇯🇵 Japan considers a four-day workweek for civil servants to attract talent and address long working hours.
🚗 Turns out that new car smell is very bad for you.
End note
I am in the Bay Area this week and so the newsletter is a bit shorter than normal.
A
P.S. We’re looking for brilliant brands to share with our audience. If you’ve got a product or service you think we’d love, complete this form and we’ll get in touch to discuss collaboration opportunities.
What you’re up to — community updates
Rumman Chowdhury: “AI desperately needs global oversight”.
Ian Hogarth wrote an op-ed for FT calling for a slowdown in “the race to God-like AI”.
Josh Berson is hosting a four-part seminar on the death of the future starting May 4. See more details and sign up here.
Share your updates with EV readers by telling us what you’re up to here.
Tim O'Reilly's piece includes this by Peter Norvig: "Human Centered AI noted in a review of a draft of this piece, “We think of ‘Human-Centered AI’ as having three spheres: the user (e.g., for a release-on-bail recommendation system, the user is the judge); the stakeholders (e.g., the accused and their family, plus the victim and family of past or potential future crime); the society at large (e.g. as affected by mass incarceration).”
This is especially valuable as it calls attention to the reality that there is more to AI than LLMs and there is a lot that needs understanding, monitoring, and possibly removal.