Antonym: The Muddle in the Middle Edition

One for the AI fans and paleoanthropologists. Something for everyone, then.

There’s always one.

Dear Reader,

If you're slowing down for Christmas, good for you. Absolutely nothing else is - particularly in AI world. I'm going to focus on AI news this week and try to make sense of some of it. Let's start with some numbers.

Two Years of Generative AI: A Progress Report

It's been two years since the starting gun on generative AI was fired with the public release of ChatGPT, and one year since OpenAI launched its GPT store, making it easier for anyone to create custom AI assistants (though what we call them remains hotly debated - chatbots? generative agents? GPTs?).

About a year ago, ChatGPT had 100M weekly active users (an oddly chosen metric over monthly or daily user counts, as I noted then). Now? The numbers tell quite a story:

OpenAI by numbers:

• 300 million: Weekly active users (announced by CEO Sam Altman at The NYT DealBook Summit)

• 250 million: Previous user count just one week earlier

• 1 billion: Target active users for next year

• $157 billion: Current valuation since launching ChatGPT

• $13 billion: Total funding from Microsoft over two years

• $6.6 billion: Latest funding round (October)

• $4 billion: Revolving line of credit

• $1.5 billion: Value of employee shares in new SoftBank tender offer

Source: CNBC

The Limits of Useful AI Scepticism

Casey Newton makes an interesting observation about AI skeptics: there are two distinct camps - internal (those who worked or have worked at AI companies) and external (those who haven't). The internal critics tend to be doubtful that the technology is safe and think it may be harmful. (His whole article on this is a strong recommend.)

In most workshops we run at Brilliant Noise there will be one or two sceptics. They are welcomed and supported if they declare their attitude. Everyone needs to be skeptical, we explain, and every team benefits from critical questioners in their discussions. They tend to want to see evidence, hear downsides considered, think about second order effects on their sector, the economy and society. This is useful, essential critical thinking.

There's a more selfish reason that I aim to welcome skeptics to AI-B-C sessions: it is so satisfying when they begin to understand how gen AI can help them and their teams. I don't think of it as a conversion, because I want them to keep their skeptic habits for everyone's sake, but to apply them to more nuanced and useful questions about AI.

Beyond sceptics there's another type of nay-sayer that reluctantly accepts sometimes. These are more cynics than skeptics, or perhaps people who've attached part of their identity to denial of even the possibility that gen AI might be more than a fad. It's really hard, almost painful, for them to admit they have changed their mind.

Recently, after an exercise in a large workshop where we were using the Claude Haiku model for simple data analysis, one charming refusenik, let's call him James, shrugged and said: "The points it made were basic, and ones I would have made too."

His colleagues by then were excited by the progress they had made and the ideas that they had for speeding up their work. They offered suggestions to help him see the utility. "But how long would it take you to do the analysis?" someone asked. "Maybe give it a bigger data-set," another offered.

There was laughter and good-natured ribbing. This was "James being James", the died-in-the-wool denier of new tech, master of Excel and no robot was going to come and prise the work of crafting spreadsheet equations from his skilled fingers.

ChatGPT-01 and "The 12 Days of Shipmas"

OpenAI hired a Chief Marketing Officer this week, Kate Rouch, formerly of Coinbase, the crypto wallet company. I wonder if they will turn their mind to the horrific product naming tangle at some point? As well as launching ChatGPT-01, which is more powerful than ChatGPT-40 and also has 01 Pro and 01 Mini variants, the literal geniuses at OpenAI have started a process of launching a product every day for the next two weeks and called it... 🥁

…”The 12 Days of Shipmas”. This might work as an in-joke for engineers and tech-heads, but the use of "ship" as a synonym for "launch" or "release" is not part of wider consumers' language. Not yet anyway.

Two days into OpenAI's "12 Days of Shipmas", and we're already seeing something fascinating emerge. Behind the headlines about ChatGPT-01 and specialist AI training, there's a bigger story about return on investment.

For those who've already invested in AI literacy, this is starting to look like payday. These organisations aren't spending time asking "but what is it for?", they are testing and planning how to apply it to their operations.

The new model isn't just an incremental improvement - it's enabling dramatic acceleration of complex tasks like ROI modelling and strategic planning. But more intriguingly, the ability to train specialist AIs with far fewer examples could reshape how organisations approach innovation.

We've seen this pattern before: early results get dismissed as 'underwhelming' by those expecting sci-fi magic, while quietly transforming how work gets done. New cost structures are emerging that will reshape business operations - but only for those with the literacy to grasp the opportunity.

Tomorrow, I'll share a deeper analysis of these first two announcements and their implications for business. But here's a question in the meantime: what could your organisation do with AI that truly understands your field?

Because that's what's really being unwrapped this Christmas.

Big Headed Humans (Not Us)

Word of the week: paleoanthropologists

Concept of the week: The Muddle in the Middle

Researchers have discovered a new ancient human species named Homo juluensis, identified by a large skull found in China, dating back to between 220,000 and 100,000 years ago. This species, with its large brain and Neanderthal-like features, may have resulted from the interbreeding of different Middle Pleistocene hominins (the name for humans and our recent ancestors and relatives).

After our Homo sapien ancestors evolved roughly 300,000 years ago, they quickly spread out of Africa and into Europe and Asia. But what happened before us? For decades, paleoanthropologists have tried to crack a fascinating puzzle: how were hominins evolving between about 700,000 and 300,000 years ago, when multiple early human species coexisted?

This period has given us some remarkable discoveries: H. heidelbergensis fossils in western Europe, Homo longi in central China. But there's debate about whether each of these really represents a separate species. Scientists have resorted to catch-all terms like "archaic H. sapiens" and "Middle Pleistocene Homo." There's even an informal name for this confusing period: "the muddle in the Middle."

I think I can find some other contexts that that will be a useful label…

The Madness of the Modern Meeting

Decision-making costs energy and saps effectiveness. Let's try to design an optimal way to make a decision. Let's make it a group decision. A slightly under-oxygenated room. We will ask people to stay there for two or three times beyond the capacity of their brain to focus on a single subject. We will provide stimulants sometimes – caffeinated drinks – and also energy boosting fruit and snacks, but both of these will have short-term bursts of effectiveness. We will ask people to prepare by reading more material than they will be able to and present information that makes more sense to the authors than the readers.

Some will struggle with the format because their brains are stronger at interpreting visual, spatial or auditory information and we have provided all data in text. Decisions will get measurably worse in terms of quality as the meeting progresses. As David Rock says in Your Brain at Work: "Make one difficult decision, and the next is more difficult."

Modern meetings hold together systems that are fraying because other meetings have not worked well - structure, notes, actions, roles undefined. Even when they are necessary and well managed, modern meetings are designed for calendars not humans. It is possible to slot a meeting in there and so it happens. Two hours at 4 pm at the end of a day where people have had an average of 3.4 meetings already… what could possibly go wrong.

The fundamental thinking trap of the modern workplace: it is too complex for us to understand, but we necessarily create a good enough narrative of how it works to not panic. The tasks appear to be modern but force us to use stone-age decision-making tools for hyper-complex problems. The kicker is we either think we can handle it or drive ourselves mad with the thought that we should be able to do it but we can't.

Don't Start With Why

Simon Sinek's always been a wrong 'un in my book. The whiff of over-sold BS around his Start With Why thesis/book/TED-talk/consulting business was unmistakeable.

In the gospel of the corporate purpose movement, riches would flow once your "why" was clear. But Nick Asbury's essay in his challenging book "The Road to Hell" picks apart the debunked scientific basis for Sinek's theory (remember Lizard brains? Yeah, that didn't survive academic scrutiny).

Here's the thing: a sense of purpose absolutely fires up a team if it's at the core of what they're doing. But that truth doesn't mean retro-fitting social-good to a business model is a useful thing to do.

This Week...

Creation Lake, by Rachel Kushner

"You people are not real to me. No one is." A Booker-nominated thriller which seems to be based on cases of undercover cops infiltrating hardcore protest groups. Radical leftists in rural France plot against giant agri-tech infrastructure. Some come for the undercover agent's strange distance from the Moulinards, stay for the almost-sensical musings on early humans by one of the movement's hermit-thinkers. I loved it but I'm not sure I can work out why.

Interior Chinatown (Disney+)

So much to like. When will I like it?

We're in the Truman Show with a background character from a cop show aren't we? Are we? There's a sort-of 1990s low-rent cop show happening, but all the background characters seem to be leading very rich lives as long as they don't try to enter the parallel cop show they live in. It's a studio. But not. They are trapped but not. This is probably subverting ideas about reality and story and who gets to be a "main character". There is probably a clever, internally consistent set of rules about the background characters. I wish I knew what they were.

I’m sticking with it for the acting talent – Jonny O Yang and Ronny Chieng are amazing – and writers and hoping they pull it off.

That’s all for this week. Take a look at BN Edition, but I will be sharing tomorrow’s analysis piece on both newsletters.

Thank you for reading!

Antony