Revolutionizing Scientific Discovery
Generative AI's Game-Changing Potential
by Ciara Sejour, 2023
If you’ve been updated on what’s going on in the tech world, you may have heard about the recent advancements in AI.
While you may not know what AI (artificial intelligence) is just by hearing the term, you’ve most likely used the technology. AI powers the algorithms behind social media apps, Spotify and Netflix recommendations, voice assistants, self-driving cars, and so much more. Recently a sub-set of AI, generative AI, has become extremely popular with people all around the world.
Generative AI is defined as any type of AI algorithm that generates new data/content that resembles human-generated data/content. Most applications of Generative AI include image, video, and text generation. For example, ChatGPT is a generative chatbot created by OpenAI, an AI research company. Amazingly, within just 5 days, ChatGPT had gained over 1 million sign-ups.
​
While generative AI is known for these use cases, there’s also a vast possibility of other use cases for it as well. One of these being scientific discovery.
​
​
​
​
​
​
​
​
​
​
​
​
​
​
Historically, scientific discovery is a constant trial and error-process. Even after doing loads of research, there are still a bunch of hypotheses to test. All of these tests can take time. The not only time to experiment but lots of time towards getting things like drug discoveries approved for implementation.
​
This process can be cumbersome, and when we are in a life-threatening situation like the pandemic, we can’t wait years for drugs to be approved. With over 10^63 possible drug-like molecules in the universe, our human minds might not be able to think of every possibility, but a computer equipped with the right AI algorithms can.
These algorithms that power generative AI applications can store and process tons of data and analyze it to make conclusions.
​
An example of scientific discovery with generative AI is with Absci, a company looking at engineering E. coli to produce antibodies that have the potential to improve human health.
​
These antibodies can help with things like drug discovery and disease detection, however, they take a long time to get into the clinical testing stages (on average 5.5 years)
​
However, with generative AI, it could take as little as 18-24 months.
​
Another example of generative AI being used for scientific discovery is with NVIDIA and Evozyme. The two companies announced back in January that with the use of the BioNeMo model, they made an AI that created a pair of new proteins.
​
In an interview with Venture Beat, Kimberly Powell, VP of healthcare at Nvidia, said, “This is the capability that is going to be how we can explore the infinite universe of proteins to discover new therapies in materials, energy sources, and sustainable foods.”
​
At the rate the technology is coming along in just the last few months, I wouldn’t be surprised to see a bunch of new applications of generative AI in the scientific realm.
​
Outside of all of the accomplishments circling the scientific community with generative AI, there is always the question of how accessible this technology will be.
​
After all, science is fueled by collaboration.
​
If this is a question that came up in your mind, fear not, there are solutions to combat this.
​
One of these solutions was created by IBM, a tech company with much industry experience in the AI space.
They created Generative Toolkit for Scientific Discovery (GT4SD), a tool to help accelerate hypothesis generation with the use of state-of-the-art generative AI models.
​
GT4SD is available on GitHub and IBM has encouraged anybody interested in it to try it out.
With all of that being said, I’m excited to see the future generative AI creates for science, and I hope I’ve given you all the reason to feel the same too!