MP#01: We outsourced our attention, let's not outsource our thinking

Welcome to all the new subscribers from last week!

At multiphacet, our goal is to see beyond the AI noise and make sense of what this revolution really means for our work, our teams, and our future. Let’s dive in.

⚓️ On deck this week

• ✍️ Editorial: ChatGPT makes us dumb, by Nicolas

• 📰 What we read this week: our brain on ChatGPT, Geoffrey Hinton podcast, systems of actions vs. systems of records, how business teams can leverage AI for acquisition

• ⭐️ Phacet insiders: human - machine interface, post-chat UI, by Côme

• 🎨 Weekly AI Inspiration: AI ASMR is here - test it yourself with Veo3

✍️ Editorial: ChatGPT makes us dumb

Among the avalanche of headlines this week, one in particular stood out: an MIT study on the cognitive impact of AI. Long story short: AI makes us dumb.

After 18 months of using ChatGPT almost daily, I’ve felt it. Now, I avoid using it at the early stages of any creative or strategic thinking. Why? Because it makes me lazy. It makes my ideas... average. It’s a race to the median.

This isn’t new. We’ve seen this story before with social media.

Social media didn’t just connect people, it rewired how we think. It shortened our attention spans. Then came Reels and Shorts, and now we swipe past anything that lasts more than 8 seconds. We’ve all become goldfish in scroll mode. 🐠

I fear the same is happening with ChatGPT. A new generation is entering the workforce having never known a world without AI copilots. For them, starting with ChatGPT is the default. But when the default is outsourcing your first thought to a model trained on the average of the internet, what happens to original thought & critical mind?

I wonder what the people building these systems recommend to their own kids - are they forbidding it as the people at Facebook did for their own kids with social media apps?

Chatbots are not sources of truth. They’re not guides for creative thinking. They’re autocomplete machines.

Personally, I’ve redefined my usage. I build the skeleton myself. I write first. Only once my thinking is shaped do I bring in AI: to improve, to polish, to reword. That’s my hygiene. And I think it’s one we should teach.

OpenAI, Anthropic, Google, Mistral - they’re the new Facebooks. And like social media, we’ll have to learn when to use them, how, and for what. Otherwise, we risk something deeper than being wrong, we risk losing our brain.

We’ve outsourced our attention. Let’s not outsource our thinking.

— by Nico (written without AI).

📰 What we read this week

• 🎙️ Geoffrey Hinton renowned AI scientist podcast on Diary of a CEO - in this podcast Geoffrey explains why he thinks people should train to become “plumbers” and why in his perspective superintelligence will replace all “mundane intellectual work”. We had the chance to meet him & Yann Le Cun (our national French AI reference) last year for a fireside chat - they don’t agree on everything but Geoffrey’s opinion is worth knowing according to us.

• 🔬 Your brain on ChatGPT - the study we refer to in the editorial. Demonstrating over 3 cohorts of people using no AI, AI & Google the impact of creative writing on their brain activity. Worth reading.

• 💻 System of actions vs. systems of records from BVP - Bessemer is a world-famous venture capital firm, known for the depth of their analysis. They share in this article their view on how AI software may displace former generations of ERP systems & SaaS tools.

• How business teams can leverage AI for acquisition from Growth Unhinged, a quick look into how business teams use AI to boost their activities & campaigns. Definitely worth reading.

⭐️ Phacet Insiders: human - machine interface, post-chat UI

A year ago, the “chat-as-UI” modality was at the heart of every product design discussion - wasn’t it too broad, too open-ended to scale, how could we overcome the blank page syndrome?

Today, we see just how powerful it can be, but also how much context shapes what works best.

In his article (Post-chat UI), Allen Pike walks through different AI interface modalities: canvas, purpose-built tools, type instead of pick, etc.

And shows how each supports users differently depending on intent.

💡 My takeaway: AI as chat is still the most versatile and almost perfect form of interaction, as it’s the most natural - but it doesn’t stand alone.

First, you can’t just dump a chat box in and hope it’ll work: you have to design how it behaves and that’s a long game (look at Anthropic’s system prompts, not even mentioning the data you must feed your LLM with).

Second, it needs the support of diverse, more focused interactions to help people get to better outcomes, as these frame what your product can do and can often be quicker than just going through the chat.

— by Côme (founding Designer at Phacet)

🎨 Weekly AI inspiration: AI ASMR

Here we share weekly AI creations (video, art, image) that struck us. Because they’re funny or inspiring. We believe it’s a glimpse into the future.

Today we wanted to share this trend of AI ASMR. Creative minds are using tools to create with Gemini model called Veo 3. You can test it here for example. Quality is definitely here now, and for all the creative people that did not have the skillset to build such assets it’s relief.
👉 For the complete AI ASMR clip check here

👉 Subscribe now to stay ahead and never miss a future edition.

If you enjoyed this week’s newsletter, feel free to share it with your network.

— Multiphacet Team