AI for Science Forum highlights potential and challenges of AI in scientific discovery

The AI for Science Forum, held this week in London by Google DeepMind in collaboration with the Royal Society, coincided with a notable celebration in the scientific community, as both organisations capitalised on the recent Nobel prize awards in chemistry and physics. Google DeepMind’s success in winning the Nobel Prize in Chemistry, particularly through developments in artificial intelligence, has underscored the potential transformative impact of AI across various scientific disciplines.

Demis Hassabis, the CEO of Google DeepMind, delivered a keynote address highlighting the promising potential of AI technologies. He articulated an optimistic vision of a potential new golden age of discovery in science, declaring the current era one teetering on the brink of transformation. However, he also emphasised the complexities involved, stating the necessity for researchers to clearly identify meaningful problems, collect quality data, develop appropriate algorithms, and apply them judiciously in order to achieve breakthroughs.

The forum also welcomed warnings regarding the drawbacks of AI technologies. Concerns were voiced surrounding the possibility of societal backlash, exacerbation of inequality, and even catastrophic data breaches. Siddhartha Mukherjee, a prominent cancer researcher, expressed the apprehension that the uncontrollable nature of AI may lead to significant disasters, drawing parallels to the nuclear accident in Fukushima.

Positive developments in AI applications were showcased during the conference. In Nairobi, experimentation with AI-assisted ultrasound technology has enabled nurses to conduct scans for pregnant women, thus streamlining access to critical healthcare. In addition, companies like Materiom are innovating by employing AI for the creation of entirely bio-based materials, thus reducing reliance on conventional petrochemicals. The advancements extend to fields such as medical imaging, climate simulation, and nuclear fusion research, with virtual cells on the horizon.

Hassabis and colleague John Jumper recently received recognition for their work on AlphaFold, an AI programme that predicts protein structures, notably contributing to drug design in biomedical research. Following this achievement, Isomorphic, a Google DeepMind spin-off, aims to enhance this algorithm further to significantly accelerate the drug development timeline. Hassabis noted aspirations to compress the drug design process from years or decades down to mere months or weeks, promising a potential upheaval in the pharmaceutical landscape.

Pharmaceutical giant Novartis has already embraced AI technologies not only in drug design but also in optimising the recruitment process for clinical trials, which can traditionally be lengthy. Fiona Marshall, the President of Biomedical Research at Novartis, remarked on innovative tools that facilitate efficient responses to regulatory inquiries, enabling companies to leverage past experiences to streamline future drug approvals.

Jennifer Doudna, co-recipient of the Nobel prize for her contributions to CRISPR gene editing, highlighted the role of AI in making therapeutics more affordable. Nevertheless, despite advancements in gene editing, the high costs associated with treatments have raised concerns about accessibility.

Addressing the challenges surrounding AI applications, particularly the “black box” problem—where AI systems make decisions without transparent explanations—Hassabis suggested that advancements akin to brain scan technologies for AI could soon allow for greater interpretability and trustworthiness.

The interplay between AI and the climate crisis emerged as a critical area of discourse. Despite claims of AI’s capability to revolutionise energy systems, fears were raised regarding the substantial energy consumption required for large-scale AI models, such as those used to train OpenAI’s ChatGPT. Hassabis defended the energy expenditure, arguing that the benefits offered by AI technologies could overshadow the limitations, with hopes pinned on AI’s potential to contribute to breakthroughs in energy solutions such as sustainable batteries and nuclear fusion.

Asmeret Asefaw Berhe, formerly of the US Department of Energy, cautioned against disregarding the challenges posed by AI’s energy demands. She called for significant sustainability measures, urging the need for holistic transformative change within the industry.

The AI for Science Forum has provided a comprehensive platform for delving into the multifaceted dynamics of AI in scientific discovery and industry development, serving as a focal point for hopes and apprehensions regarding the future trajectory of artificial intelligence in business and beyond.

Source: Noah Wire Services