With AI scientists charging $15 per paper, humans will 'move up the food chain'

Scientific research will be one of the industries automated in the near future, a new paper shows.

For centuries, societal evolution has been driven by academic research and the innovations that result from it. Scientists tinkering in their labs, and conducting experiments to see the impact of changes on our environment have driven most of our innovation, and pushed forward development.

You might think that as the AI revolution sweeps all before it, with 40% of jobs affected in some way by the technology, scientific research – and the communication of its results through academic writing – might be one of the 60% of jobs that isn’t affected. After all, while it’s possible that AI could automate some parts of the scientific process, it has long seemed unlikely that AI could do the entire process from start to finish.

At least until now. The old model of human researchers toiling away in laboratories could soon disappear, according to a new paper that proposes an AI-powered scientist who can publish work for as little as $10 or $15 a time and suggests that it could kickstart a new scientific revolution.

Automated AI

The new AI system is developed by leading experts from Sakana AI, the University of Oxford, and other prestigious institutions. It could signal a huge leap in the production of scientific literature.

According to the researchers, the AI Scientist – as the model is dubbed – can independently brainstorm novel research concepts, assess their feasibility, conduct necessary experiments, and document the entire process in a structured and professional manner, mimicking the human scientific method.

The AI Scientist proved capable in the study of generating research papers in three distinct subfields of machine learning, including diffusion modeling and language modeling. The tech uses language models to mimic how scientists work, generating ideas for research, and then running the experiments for them – before writing up the results.

“The third phase of The AI Scientist produces a concise and informative write-up of its progress in the style of a standard machine learning conference proceeding in LaTeX,” the researchers write.

The (human) researchers behind it claimed the technology can "generate its own scientific ideas and hypotheses, as well as a plan for testing them with experiments." The potential of that for the world of scientific research could well be revolutionary.

Human-like quality

When the papers were reviewed, the academics found the outputs could pass the acceptance threshold of many well-known machine learning conferences, highlighting just how quickly the tech could be integrated into the world of scientific research.

“The AI Scientist can generate hundreds of interesting, medium-quality papers over the course of a week,” the researchers write.

However, academia already struggles with too many papers being published and not enough people to adequately review them. Handily, the researchers also developed an automated reviewer powered by AI, which could also achieve near-human accuracy in evaluating the quality of the generated research papers, further reducing the need for human intervention in the process.

However, the paper argues that “the role of a human scientist will be diminished.”

“We expect the role of scientists will change as we adapt to new technology, and move up the food chain.”

For some, the breakthrough will be seen as the start of a new era where AI can address some of the world’s most challenging problems with unprecedented speed and efficiency.

But for others, there are more material concerns. Already, academics are using generative AI to help them write their academic papers – with issues of academic integrity already arising, in large part because of generative AI systems’ ability to hallucinate their outputs.