IBM and NASA Collaboration Aims to Enhance Climate Understanding

IBM and NASA will use their cutting-edge AI and machine learning capabilities to research and analyse massive data sets, including satellite images, to better understand the correlation between human activity and climate change. NASA will provide access to large petabyte and terabyte-scale data sets, which IBM will process using its supercomputer to accelerate the research. This collaboration will bring together the best of computer intelligence and human creativity to unlock new insights and imagination in the field of climate research. The result will be a better understanding of how to mitigate the impact of human activities on the planet and a stronger foundation for future generations to build upon. What challenges do researchers face with modern data, what will the partnership hope to achieve, and why is AI the future of scientific research?

What challenges do researchers face with modern datasets?

By far, the single most important factor in quality scientific research is data. A good scientist postulates a position, gathers experimental data, and then checks to see if their theory follows their data. Data that doesn’t support a theory should always be scrutinised to ensure its validity (i.e., ensuring that proper data collection methods were used), but if that data is indeed invalid, the original theory should either be reworked or scrapped entirely.

However, as technology continues to advance, the amount of experimental data being gathered is growing exponentially, and it is virtually impossible for any scientist to go through countless terabytes of data. Creating computer models that make sense of the data can help, but the use of traditional programming methods is also becoming exponentially difficult to create and execute. 

Engineers are combatting these challenges with more powerful computers, and this has indeed helped now that smaller research teams can utilise cloud resources. But again, another issue crops up in experimental data, and that’s recognizing what data is valid. Among terabytes of data, a proportion of this data will likely be unusable, but identifying outliers in code can be tricky. Furthermore, identifying patterns that show correlation can be immensely difficult when considering that computers can rarely distinguish between correlation and causation.

Simply put, the amount of data that researchers have to deal with limits their ability to identify patterns of interest and even prevents new discoveries. 

IBM and NASA team up to use AI in climate research

Recently, IBM and NASA announced a new partnership that will combine the data-gathering capabilities of NASA with AI research developed by IBM to help researchers accelerate climate science. While climate change is a well-established theory with countless amounts of evidence, the impacts of climate change are somewhat difficult to guess due to the sheer complexity of the climate and its effects on land-based activities.

For example, the climate will undoubtedly rise in temperature, and this will increase CO2 levels as the oceans become less able to hold CO2, but while some believe that this will result in a widespread drought, others believe that the increased CO2 will lead to increased plant growth. At the same time, many speculate that sea level rise will destroy vast amounts of coastline, but others believe that rising sea level estimates are somewhat overstated. Either way, it is difficult to see how the environment is being affected on a global scale and more challenging to predict. 

By recognising these challenges, IBM will use data gathered by NASA to help researchers answer these difficult questions by utilising IBM AI technologies. The new foundation models will be fed petabytes of data from NASA that include satellite photography, land mass changes, coastal lines, and weather patterns to try and identify patterns. Furthermore, these patterns will then be used to create models that will enable researchers to better understand how these changes will affect land-based activities, including agriculture and urban planning.

According to IBM, their foundation AI model can read vast amounts of data and identify underlying structures without any direction from a human. This allows the AI to identify potential relationships in datasets; however, it cannot identify the difference between causation and correlation. However, being able to identify either allows researchers to look into new ideas without needing to go through large datasets.

Why is AI the future of scientific research?

AI has many practical uses, and there is no doubt that AI will become essential in the field of scientific research. The improving capabilities of technology and large amounts of data being gathered are already too much for humans to process, and the job of identifying patterns is best left to machines. Instead, researchers can focus their attention on tasks that machines cannot handle well; original thought and creativity. As machines process data, humans can use imagination and creativity to interpret the results from machines. From there, new ideas about how the universe works can be formulated, while AI can rapidly test these theories for validity.