OpenAI’s free plan for academic researchers is expanding, with the goal of allowing up to 200,000 researchers around the world to use cutting-edge models for free.

A new round of applications will open on November 23, local time, with an additional 50,000 places. There are two changes to the rules.

First, there will be no more drawing of lots, and it will be changed to first come, first served. Second, the level is lowered from ChatGPT Pro to Plus. OpenAI’s explanation is that after looking at the actual usage during the pilot period, the majority of people’s needs fall within the Plus quota. Each quota includes the applicant and up to two collaborators, and all three people can use it for free. This round will cover up to 150,000 people.

Since the pilot launched in July, 20,000 researchers from 33 countries and regions have joined. Those on the waiting list who qualify will receive activation instructions via email before opening. The official also promised that the data shared in the plan will not be used to train the model by default.

OpenAI conveniently sent out a survey: 251 participants estimated that the same quality of work would take 1.9 times longer without AI. It also counted more than 580,000 scientific research messages from 1,819 scientists between August 20 and September 22. 68% of them used it to conceive of research ideas, and about 23% of the creative messages directly asked ChatGPT to determine research questions for them.

The efficiency figures are estimated by scientists themselves, and OpenAI says they cannot prove cause and effect. But the 23% is more interesting: quite a few people have begun to hand over topic selection to models.