Skip to content
Fortissimo

Futuro Fortissimo Archive

🎼 ff.106 AI-lloween

Trick or treat in the age of artificial intelligence

Published · Michele Merelli

  • Dressed as GPT?
  • Is Anthropic's Claude better?
  • Is it a good idea to give artificial intelligence a mouse?

🎃 ff.106.1 Dressed by AI

Substack ↗

Today, a quick themed issue, since it's the long weekend of the dead. By the way, do you know what you're dressing up as for Halloween yet? A few ideas, generated by GPT.

The ghost of Valtellina bresaola, I must say, didn't come out badly.

🎃 ff.106.1 Dressed by AI

🖱️ ff.106.2 AI with a mouse

Substack ↗

Speaking of monsters: do you know Magic: The Gathering? If you had a nerdy childhood like mine, the answer is: “absolutely!”. For everyone else: it's a card game in the style of Yu-Gi-Oh. But better.

Here's an AI playing MTG online, through screenshots and (for now) slow mouse clicks.

This is one of the first examples of Claude Computer, the latest invention from Anthropic (OpenAI's competitor and creator of Claude, the better alternative to ChatGPT).

These models are tested in isolation, on virtual computers (the OSWorld benchmark). With Claude Computer we've reached 22% accuracy, 3 times better than before.

Examples of failure: probably bored, Claude started searching for pictures of Yellowstone Park.

AI-fraid? A little, yes, because isolation (hoped for in the past, 🦉 ff.36.2 The super-tome on the singularity) is very flimsy: with simple screen mirroring, someone had it control a phone to check the latest NFL scores.

🖱️ ff.106.2 AI with a mouse
🖱️ ff.106.2 AI with a mouse
🖱️ ff.106.2 AI with a mouse

🕊️ ff.106.3 Nexus and Harari

Substack ↗

Anthropic's Claude is named after Shannon, who invented the entropic theory of information.

Entropy, information and AI: a perfect logical thread that leads us to Harari and his latest book, Nexus.

The thesis: ChatGPT is the latest stage of a process that began with writing, technologies that “schematise” the world into tables and drawers, a completely non-human skill.

Bureaucracy tries to solve the problem of retrieving information by dividing the world into drawers, separate containers so that documents don't get mixed up. However, this principle comes at a cost. Instead of focusing on understanding the world as it is, bureaucracy is often busy imposing a new, artificial order. In short, bureaucracy sacrifices depth of understanding for the sake of organisational efficiency, creating a partial and distorted view of the world. Yuval Noah Harari, Nexus

Nexus: Harari's new book (futuro fortissimo)
Nexus: Harari's new book (futuro fortissimo)

🐕 ff.106.4 A dog's life

Substack ↗

Back to Claude: in one of his most enigmatic prophecies, Shannon suggested that one day we would be "dogs for AI".

But, but, but. Look closely, and our four-legged friends, who started out as hungry wolves, now sleep on our designer sofas and eat gourmet food.

What if Shannon was right, but the other way round? Maybe dogs domesticated us: with a few tail wags they "hacked" evolution, scoring belly rubs and kibble.

Are we controlled by algorithms, or are we “domesticating” artificial intelligence (and the robots that will follow) to dodge the queues at Esselunga and the traffic between Pero and Cormano?

As Manzoni said, to posterity (and AI) the arduous verdict.

🎶 fortissimo notes archive

☕ virtual coffee

❤️ a little heart below?

👋 See you soon, from micmer

© futuro fortissimo | michele merelli

🐕 ff.106.4 A dog's life
From Wes Anderson's film Isle of Dogs (via futuro fortissimo)
From Wes Anderson's film Isle of Dogs (via futuro fortissimo)

Related issues

Newsletter

Get new issues of Futuro Fortissimo

A free newsletter by Michele Merelli (Bergamo) on technology, AI, nature and society. Written in Italian, delivered by Substack.

📩 Subscribe for free ↗
All results in the archive →Press / to search anywhere