Discussion about this post

User's avatar
Federico's avatar

I was waiting for your take on Claude after the Claude 3.5 boost! Thanks for that.

Claude is just amazing compared to the other models. For me, there's no point of comparison. It's great at producing creative pieces. And I've found it's great as a discussion partner to analyze poems (for instance, this poem: https://www.ronnowpoetry.com/contents/hall/ImpossibleMarriage.html). Artifacts made it more powerful and interactive.

And for some reason only Claude 3 Opus and Claude 3.5 Sonnet can nail this test:

• Write ten sentences in which the last word is "apple" and the fifth word is "Japan."

GPT-4o goes wrong from the start:

In Japan, they grow a unique variety of apple.

And Gemini Advanced goes wrong too:

• The traveler brought back stories from Japan and a unique type of apple.

Claude 3 Opus and Claude 3.5 Sonnet get all 10 sentences right.

Expand full comment
D R's avatar

Excellent piece and one that all GenAI skeptics should read. The terms used by yourself (Co-Intelligence) and Microsoft (Copilot) are how people should view these tools. They are useful assistants *now*; too many people dismiss their utility if the first answer to a simplistic prompt is not exactly what they wanted.

And I completely agree on Artifacts. I’ve been experimenting with LLMs since late 2022 and even I said, “Are you kidding me? This is so good!” when I first experienced Artifacts in action

Expand full comment
67 more comments...

No posts