Briefs · Frontier Science
For early-career scientists, the cost of AI can be a problem
AI is not free. It is a subscription, and the poorest scientists are paying it.
Saad Bhamla, who leads a biophysics and engineering lab at the University of Colorado Boulder, polled nearly two dozen members of his lab earlier this year: which AI tools were they using, and how much were they paying? He was startled to find everybody was paying, from $20 a month upwards.
A Nature survey in March of more than 10,000 researchers worldwide found 41 per cent paid for their main AI tool themselves, against only 11 per cent whose institution paid. Some universities have struck campuswide deals, and programmes from OpenAI, Anthropic and Google give tens of thousands of scientists free or reduced access — so some get top tools for next to nothing, and the richest labs pay nothing!
PhD students are shouldering the cost: Haris Riaz, a PhD student, was “lucky” that industry internships helped cover it, and Omar Choudhry (a PhD student at the University of Leeds, using AI for medical diagnosis and care) is among those paying their own way. It is a university canteen where the professors eat free and the students pay for the cutlery.
Where funding is scarcer the divide is starker: Giovani de Almeida Valdrighi, a PhD student in Brazil on 3,100 reais a month ($615), would spend about 3.5 per cent of his income on the same plan, and many of his friends go without. At this rate the lab’s biggest experiment will be seeing how long a PhD student can live on instant noodles (sorry, noodles).
Conclusion: AI is not a leveller. It is a bill.
Read the original at Science (AAAS)
Summary in our own words. The original belongs to its publisher.