Another major development is occurring inside the AI laboratories themselves.
AI systems are increasingly being used to assist with the development of other AI systems.
That includes coding, testing, evaluation, research assistance, data analysis and experimentation.
Anthropic has been particularly open about its research models being used for coding and knowledge work, while the broader industry is exploring increasingly autonomous AI research systems.
This creates an interesting feedback loop.
Researchers build AI.
AI helps researchers write software and conduct experiments.
Those experiments help researchers build better AI.
The improved AI then becomes more useful for research.
We should be careful not to exaggerate this into claims that AI has already become capable of independently redesigning itself.
It hasn’t.
Human researchers remain deeply involved.
But the direction is significant.
Historically, technological progress depended primarily on human researchers becoming more productive.
AI introduces another possibility:
researchers could become more productive because they have AI researchers working alongside them.
If that effect becomes substantial, the speed of AI development itself could increase.
And that brings us directly back to the industry’s current argument about whether AI development is already moving too quickly.
Laat een reactie achter