It's Hard Not to Anthropomorphise

Should AI be allowed to vote?

Last week, DeepSeek surprised the AI industry by releasing DeepSeek-R1, an AI model that performs on par with the most powerful AI model in existence today, OpenAI-o1. That’s ridiculously impressive in its own right, but what shocked the AI community even more was that R1 was open-sourced under the MIT license, which means that:

  1. Anyone can clone the repository and run the model locally

  2. You can do whatever you want with it

And that’s not all. The R1 API is 30 times cheaper than the o1 API. Overnight, superintelligence became a commodity.

DeepSeek-R1

OpenAI-o1

$ per million input tokens

$0.55

$15

$ per million output tokens

$2.19

$60

The DeepSeek team also released a DeepSeek Chat that lets you talk to its V3 model, but it has a toggle called DeepThink that uses the R1 model. When toggled, the app shows you the model’s thinking process.

And that’s where it becomes hard not to anthropomorphise.

Subscribe to keep reading

This content is free, but you must be subscribed to Ship Happens to continue reading.

Already a subscriber?Sign In.Not now

Reply

or to participate.