LAMs are seen as a significant step towards achieving AGI, as these models not only respond but also act upon instructions.