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:

The building block of modern computing celebrates 75 years,
Energy transition & indigenous land,
How Google helped identify offenders in Capitol riots on January 6.

Sunday chart

We are celebrating 75 years of the transistor, the building block of modern computers. Decreasing transistor size has historically been described by Moore’s Law, first articulated in a paper in 1968. That paper is, according to Virginia Hefferman, possibly the “most influential article of the twentieth century.” Transistors are now getting so small that Intel has branded their next generation of chips the Angstrom era, the next unit of measurement below nanometers.

Share: Email | WhatsApp | Linkedin | Twitter | Facebook

In the upcoming Charts of the Week, sent out on Wednesday, I explore these limits and where the future is heading for the tiny but powerful transistor. Subscribe to receive the edition.

Market data

Spotify’s ad-supported tier - the free tier - added roughly 15m users last quarter, as paid-user acquisition growth continued to decline.

Lecanemab, an Alzheimer’s drug, reported phase-3 trials, a 27% improvement in the rate of cognitive decline.

A new Stanford study dismisses the link between phone usage and mental health problems in kids.

54% of projects extracting minerals for the clean transition are on indigenous lands around the world.

352Tbps: the capacity of Grace Hopper, Google’s new optical cable across the Atlantic.

Short morsels to appear smart at dinner parties

🌋 The eruption of the world’s largest volcano disrupted key atmospheric carbon measurements, including the ones we use in this very newsletter. See also, Mauna Loa eruption is a time for prayer and reflection for native Hawaiian people.

📉Sperm quality is on the decline. Environmental toxins are one of the key reasons.

🌊 Amazon is trying to salvage Echo by riding the generative AI wave. See EV#400 for commentary on the voice assistant crisis.

📍The FBI found Capitol stormers through geofence warrants, giving the agency access to precise Google location data.

🎹 How key changes have gone extinct in the music industry.

🤦🏻 Scam, Bankrupt & Fried keeps giving interviews that he shouldn’t.

End note

I’m away this weekend participating in a workshop on information ecosystems and democratic resilience.

Meanwhile, I have also been playing around with OpenAI’s new chatbot, ChatGPT. It’s quite fun, but also dangerous. Check these out:

Azeem: Write a rap in the style of Eminem about someone receiving their PhD on the subject of the changing nature of the portrayal of family relationships in Western literature, from Greek through to the Shakespearean period.

OpenAI’s chatbot: Here is a rap in the style of Eminem about receiving a PhD on the subject of the changing nature of the portrayal of family relationships in Western literature:

Verse 1:

I'm the Doctor of Western lit, 
My thesis on families is a hit, 
From the Greeks to the Shakespearean age, 
I've studied the changes on the literary stage.

Chorus:

I'm the PhD in Western lit, 
I know more about families than anyone can fathom, 
I've studied the greats, from Homer to the Bard, 
And now I'm ready to take on the world and its problems.

Verse 2:

I've analyzed the bonds between mother and child, 
From Clytemnestra's wrath to Desdemona's mild, 
I've explored the ties between father and son,
From King Lear's madness to Hamlet's fun.

Chorus

Verse 3:

I've delved into the themes of loyalty and betrayal,
From Othello's jealousy to Cordelia's denial, 
I've examined the ties that bind and divide, 
In the world of the characters that live and die.

Chorus

My assessment: I’m not Marshall Mathers and certainly not Dr Dre, but I think that is quite a reasonable effort.

Azeem’s next question I have a stress fracture in my foot. Is it worth using hyperbaric oxygen therapy to accelerate healing?

OpenAI: There is limited evidence to support the use of hyperbaric oxygen therapy (HBOT) for the treatment of stress fractures in the foot. While HBOT has been shown to accelerate healing in some types of bone injuries, its effectiveness for stress fractures specifically has not been well-studied.

It is important to consult with a medical professional before considering HBOT for a stress fracture. Your doctor will be able to provide personalized advice and treatment recommendations based on your specific situation and the severity of your injury. In some cases, HBOT may be an option to consider along with other treatments, such as rest, ice, and pain medication. However, it is important to follow your doctor's recommendations and not to rely solely on HBOT as a treatment for a stress fracture.

My assessment: This is a pretty good result. I’m suffering from this at the moment and this is the conclusion I came to after traditional research and speaking to doctors.

Others have pitted ChatGPT on an IQ test. It scored 83. (Just below 1.1 standard deviations from the mean, so 90-ish% of humans would score higher.)

But, but, but… the chatbot, while better at avoiding obvious prejudice or outlandish claims, still falls foul of many common pitfalls of these systems. Nor does it address some of the wishes I outlined last week.

So it is exceptionally good at lying. In this outlandish example, the chatbot creates extremely realistic academic references that are entirely fake.

And in the example below, ChatGPT will deny knowing today’s date, if it does.

Riley Goodside @goodside

OpenAI's ChatGPT appears to be designed to pretend that it does not know the current date, even though it does. If you're clever, you can make it reveal that it knows. Ask about this, and it will continue to deny knowing in spite of its prior answer.

The problem is that you simply don’t know if the results are good or not. They look realistic and valid, but are they? I used ChatGPT to write a Prolog programme, which it did. But my go-to Prolog expert, Barney Pell told me “The code is a valid prolog program, and expresses the core idea of a solution. But there’s no simple fix to make this a working solution.”

It’s a lucky dip. Sometimes it’ll be great. Sometimes it’ll drive you off a cliff. And you probably can’t tell which is which.

In some sense, this reminds me of the state of full-self driving. Largely really good at lane keeping and reversing, but don’t try driving from LA to New York. ChatGPT is impressive but keep your hands on the wheel and eyes on the road.

Cheers,

A

P.S. I’ve left the most outlandish experiment of ChatGPT to last. Jonas Degrave gets ChatGPT to build a Linux virtual machine inside the chat interface and then gets that virtual machine to access ChatGPT in its own imagined universe… and even access the web in that universe. Have a read (and a great week)!

Share

What you’re up to – notes from EV readers

Polly Mackenzie wrote a blog post on her work to design a purpose-led university.

Sam Barcroft launched The Big Swing, a podcast where he interviews pioneers of content about the biggest decisions on their path to success.

Jeremy Silver published “Towards a Digital Renaissance”, a book exploring the complex interrelationships between business, founders, investors and technologies, focused on the music industry.

Gio Sampietro launched the Spawns Collection, a collaboration with oio showcasing AI - human co-design.

To share your projects and updates, fill out your details here. Because of space constraints, we prioritise updates from paying members and startups I have invested in. (You can become the former by subscribing, if you have not already, and the latter by getting an intro to me via a trusted contact.)

Dig deeper

Exponential View by Azeem Azhar

Azeem's commentary: Waiting for GPT-4

If you missed Part 1 of the On the Generative Wave series, read it here. It’s been 29 months since OpenAI launched GPT-2, its large-language model which demonstrated the power of transformers-based neural networks. GPT-2 impressed with the quality of its natural text generation. Its successor GPT-3, a bigger, more complex model, delivered even more powerful results…

2 months ago · 29 likes · 4 comments · Azeem Azhar

Exponential View by Azeem Azhar

🔮 The Angstrom era; Alzheimer's hope…