THE CRUNCH
Microsoft AI chief Mustafa Suleyman has criticised rivals for focusing on AI consciousness, releasing a Humanist AI code of conduct. The 37-page document outlines Microsoft’s philosophy that AI should be a subordinate, controllable force serving humanity. Suleyman argues that the industry is distracted by concepts like model welfare and that containment is more urgent than alignment. He suggests that without safety,
He compares uncontrolled AI to a car where the brakes attack a neighbour, implying the current approach to safety has run aground. Suleyman believes models have become more steerable over the last three years, but insists that containment is necessary before alignment can be fully achieved. The interview also touched on the rapid progress of AI models, noting that a jump from GPT-6 to GPT-9 would involve 1,000 times more computing power.