Last weekend I tried a new AI model called Deepseek.
Since I've been using as many AI tools as I can (2024 Challenge), I immediately knew this was going to be the talk of the week.
It was really good. It was fast. It was free.
(...and is was out of China...and fyi, they collect your IP, keystroke patterns,
device info, etc...I wouldn't rush to have it help you organize your finances)
They discovered a new way to train the model. What big tech companies are spending billions on, they did for $5.5M, by creating a model that can organically grow intelligence...
The difference is akin to reading a book on how to ride a bike versus learning to ride a bike by
trying, falling, and trying again until you get it.
People are wondering, "is it sentient?"
My amateur read of the situation is that novel training methods will drop the costs of
the compute layer. As the base layer of LLMs gets cheaper and commoditized, more apps and agents can and will be built on top.
Is it coming for all our jobs?
I don’t want to get a reputation for reactivity or hyperbolic statements...but what just happened in the AI world changed the development trajectory of humanity.