One query in, I am convinced of the value of Deep Research and think it is well worth the $200 per month. The sources are real, the narrative less fluffy, the points it makes cogent. The narrative review is not dead yet, but it is on its way out. Here I am thinking about those reviews that are made to pad junior researchers CVs while they are introducing themselves to a field, neutral in tone and seemingly comprehensive in scope. There will always be a place for an opinionated perspective from a leader in the field.
In a year, the AI algorithms went from an overeager undergrad to a competent graduate student in every field of science, natural or social. Would o3 make this into a post-doc able and willing to go down every and any rabbit hole? Even now two hundred dollars per month is a bargain — if the price stays the same with next generation models it will be a steal.
The one snag is that it is all centralized, and yes the not so open OpenAI sees all your questions and knows what you want. For now. Local processing is a few years behind, so what is preventing nVidia or Apple or whomever from putting all its efforts into catching up? How much would you pay for your own server that would give its in-depth reports more slowly — say 30 minutes instead of 5 — but be completely private? And without needing benefits, travel and lodging to conferences or any of the messy HR stuff.
The brave new world is galloping ahead.
(↬Tyler Cowen)