Anthropic's latest advance with AI, where it can operate on-screen software, shows us how future AI can replace much of today's human knowledge work.
Anthropic's latest advance with AI, where it can operate on-screen software, shows us how future AI can replace much of today's human knowledge work.
I've been wondering when current LLM AIs would start to master this ability. I suspect it will be one of the things it's good at. For many tasks, software usage patterns are relatively predictable and modelable. A trend with current AI, is for competitors and open-source to rapidly follow industry leaders. We can expect AI like this to be widely available in six months.
Many people's knowledge work employment is tied to software skills and experience. That premium is about to start diminishing. People are familiar with the concept of 'macros'; automating repetitive sequences of software usage. It seems all but inevitable AI will be doing something similar, but orders of magnitude greater, and that all the forces in free market economics will be driving it to replace expensive humans.
Mastering the software is easy part. Doing something useful with it by understanding the outside world and doing real work is something else.
Thanks for many interesting technology posts.There is a tsunami coming in the workplace, you can already buy a humanoid robot for $16k (1), which is less than the cost of an employee. When these robots can become actually useful (instead of marketing material) businesses who use labour will not think twice about swapping over. What do we do when unemployment goes up to 25%, 50% etc
It's impressive, but security's going to be a fun (hard) one to solve for general usage: https://embracethered.com/blog/posts/2024/claude-computer-use-c2-the-zombais-are-coming/