Other
How AI is revolutionising science
The EconomistBear, Andy Davis, Alex Zarankov, Emily Shuckborough, Adi Rakhoum, Andy Cooper, Alok Jha
- AI is projected to revolutionize brain scan translation into text and accelerate biological research timelines from years to significantly shorter periods.
- Generative AI algorithms are forecast to identify protein targets for age-related and fibrotic diseases, shifting drug discovery from searching to generating molecules with specific desired properties.
- AI implementation in preclinical drug development is expected to reduce time and costs by 25% to 50%, while broader investment is anticipated to democratize access and yield numerous new therapeutics.
- Google DeepMind's AlphaFold has already established a database of over 200 million proteins utilized by more than 2 million researchers.
- Future AI applications include enhancing electron microscope image resolution via super-resolution models, mining millions of papers for patterns, and identifying new materials for batteries and solar panels.
- Specific research plans involve using AI to decode bat vocalizations to distinguish meanings related to conflict and mating, though skepticism remains regarding the potential for human bias in audio datasets.
- Self-driving labs employing AI-driven robots are expected to operate continuously (24-7), exceed human physical limits, improve reproducibility by publishing failures, and potentially achieve discoveries inconceivable by humans alone.
- Teams envision deploying mobile robot swarms in large pharmaceutical settings containing 50 fume cupboards and 20 distinct instruments, utilizing large language models to encode laboratory spaces.
- Significant risks include rising instances of fraudulent research and AI-generated fakery in scientific journals due to technology misuse.
- Broader scientific impact depends on overcoming human obstacles such as data prioritization, regulatory frameworks, and monetization strategies to fully realize AI's potential comparable to historical breakthroughs like the internet.