Al-novation: Finding New Uses for Artificial Intelligence Across Industries

Abstract

As we have seen throughout history, people and indeed companies can catalyze dramatic changes through invention and innovations. They can create new fields of science that did not exist before. In doing so, new technologies can reshape cities and create unintended consequences. We are now in a different era in which the development of artificial intelligence (AI) is both being rapidly applied in many areas of science and creating ethical challenges that need to be considered. AI may autonomously suggest innovations leading to a revolution in science and health care. What follows is a provocative brief opinion that looks at the future of how new technologies are developed through AI with all its attendant challenges.

Just over 40 years ago Cambridge, Massachusetts, was not the major hub of biotechnology companies that it is today. The convergence of venture capital companies (VCs) and several future Nobel Prize winning scientists, including Walter Gilbert and Phillip Sharp, led to the formation of Biogen, which became one of the first such biotech companies developing recombinant proteins such as alpha interferon and a hepatitis B vaccine.

In the late 1970s to early 1980s, Cambridge would have been unrecognizable today. Biogen became an anchor biotech company and over the next decades was joined by many other major pharmaceutical and biotech companies, as well as hundreds of new smaller start-ups hoping that some of the biotech magic would rub off on them. They had created the new field of biotech. Fast forward to today and one wonders what it will take to create the next scientific breakthrough that may have a comparable impact as biotech.

In recent years we have seen the continual emergence of many new technologies such as CRISPR, DNA-encoded libraries, and more recently artificial intelligence (AI)-drug discovery companies,^1,2 which themselves might be considered a disruptive technology for the pharmaceutical industry. This leads us to ask, “are we on the cusp of an AI led technological revolution?” We have already seen AI impact everything from banking (fraud detection), transportation (self-driving cars), health care (personalized therapies), manufacturing and retail (load and logistics planning), consumer products (voice recognition), and beyond.

Other influential biotech examples include newer companies such as Moderna and BioNTech, which have shown the powerful impact of mRNA technology in COVID-19 vaccines. Do any of these fit the model as seeds for a new technological or scientific revolution?

We may define such an event broadly as when a technology is rapidly replaced by another (analogous to a scientific paradigm³) to create accelerated progress in a field that in turn impacts society. Although these preceding technologies have brought in plenty of press and billions of dollars in investment, they may not yet represent the next novel “giant” field of research in themselves on par with the development of the biotech industry. One could rightly argue that these are all still branches of the existing biotech paradigm. Let us think a bit bigger to create the next revolutionary technology that is entirely new, whether through incremental change or revolutionary change and consider what might constrain this.

With New Technology Come Regulations

In the process of developing the field of biotech, Sharp, Gilbert, and colleagues at Biogen faced pushback from the Cambridge community. There was unease, if not fear, around the capabilities of their new transformative recombinant DNA technologies. This in essence forced the development of ethical and other guidelines for its use and self-regulation. Development of a biohazards committee and the experimentation review board⁴ to guide and communicate the safe and ethical use of biotechnologies likely set Cambridge apart from other regions in the United States and likely the world. Since the 1980s, virtually all new technologies have faced some form of ethical challenges, from gene therapies, the cloning of animals, and now CRISPR.⁵

Any revolutionary new technology, such as AI, is likely to encounter a similar scenario where there will be initial calls for regulation and public comment. For example, the recent use of AI generative approaches to design chemical weapons⁶ is facing similar calls for regulation and guidelines to prevent potential misuse, while others have suggested that ethics in AI is the real frontier.⁷

Creation by AI is certainly not limited to science as we have seen with the use of DALL-E-2 for art production and Generative Pre-trained Transformer 3 for production of written documents and images. These programs could be argued as interpolations as they are creating only within the bounds of what is known. They have also garnered widespread interest even recently creating the cover for magazines like The Economist.⁸ Whereas, for scientific purposes and applications to be innovative and potentially patentable AI-based creations, these tools will have to be in new areas that have never been seen before and extrapolate beyond what is currently known.

As an example, AI-based design in chemistry needs to do both, it should explore new chemistry property space to create the intellectual property (IP) and then iterate to optimize the hits or lead molecules. Although AI might be able to repurpose an existing drug or molecule, the ultimate challenge will be to create completely new molecules that are of much higher value. AI-generated inventions have recently been proposed as threatening the current patent system, as inventors are usually assumed to be human suggesting the need for AI-IP laws.⁹ In the past 20 years we have seen proteolysis-targeting chimeras (PROTACs)^10–12 that combine two different molecules and a linker driven by the biological knowledge of ubiquitin-mediated degradation.

Can the combination of our vast biological and chemical knowledgebases be fed into AI to enable the creation of the next new therapeutic type targeting a new mechanism that may not bear any resemblance to any known molecule type that currently exists? Creation of whole new molecule types, such as PROTACs, would be an important development for the pharmaceutical industry. The recent developments in structure prediction enabled with Google DeepMind's AlphaFold¹³ have recently enabled the generation of 200 million structures of proteins.

This creates numerous opportunities enabling structure-based design to develop potential treatments for diseases and beyond, but of course is tempered with the knowledge that these structures are predictions and not experimental observations. In addition, there is also the ethical dilemma of potential dual use¹⁴ once structures are available for essentially all known human proteins.

AI-novation

Looking far into the distance, instead of using AI to deliver relatively trivial human-driven images or text as we are seeing now with generative AI, it could be used for innovation by AI or “AI-novation” that may be the next new scientific industry. In other words, AI would be used to create in any scientific area that would not be derivative of previous discoveries or inventions. AI might create new molecule types unlike anything we have seen before or develop new scientific techniques. To date, we have enough data (genomics, metabolomics, proteomics data, biological data in different organisms, chemical structures, knowledge on protein interactions, and a vast array of scientific concepts, rules, and theories) to feed this AI idea generator.

We can imagine the integration of these vast existing databases of knowledge with the powerful generative AI that we currently have. It does raise the question of whether we will need to ask the AI to create or whether it will spontaneously do it in an autonomous manner once the data are provided. This distinction may be observed as the technologies develop. Such an AI would also need filters to check that ideas have not previously been described (novelty detection) as well as a method of prioritization, or rank ordering of impact or patentability, of the new ideas. Clearly this concept is not limited to ideation in the biosciences.

Is generative AI the start of the next great new technology that could revolutionize humanity? Or will it flop like other once touted technologies? (Segway, Google Glass, the Betamax video recorder, etc.) In parallel we are seeing the rapid development of quantum computing¹⁵ and its use for machine learning of large chemistry structure activity data sets¹⁶ that demonstrates potential as the future compute engines of AI-novation in drug discovery.

Who and where are the next “Gilbert and Sharp” to play their roles in this new “AI innovation age” and serve as technology revolutionaries that have the power to convert whole cities, which could in turn build technology company ecosystems and inspire new generations of scientists at the same time? How do we find or create these individuals? Do we even need to? These are perhaps questions others are more qualified to answer. Perhaps AI could replace the role of such scientists in this context, and they in turn become the innovators.

Here we are >40 years on from the beginning of the biotech industry that expanded far beyond the development of drugs for human health into an ecosystem that has impacted agrochemicals, animal health, and many other industries. Rather than requiring the convergence of great scientists, VC funding, and revolutionary ideas based on the novel science coming out of academia or elsewhere, we might be able to de novo create new scientific fields using AI. That is at least the theory. We may require humans in the loop to actually interpret and validate that they are new fields, inventions, or discoveries, which will pose considerable challenges.

Are we willing to see how far we can push generative AI? We may want to start small by applying the methods available to each of our own industries, whether it be pharmaceutical, consumer product, agrochemical, etc. Alternatively, we can think much bigger and use AI to design whole integrated products, which may in turn lead to the “next big thing,” an industry that creates products across many fields to bring value to people and investment. Examples of such products can include a new cleaning agent, methods to facilitate the remediation of harmful chemicals in the environment, or a new method to detect toxic molecules.

Along the way, this revolutionary technology will hopefully create jobs and transform our towns and cities as a consequence of its growth. It may be admittedly premature, and you could say that those of us who have been in the pharmaceutical industry for ∼30 years have been here before as we have witnessed the hype cycle for different computational technologies many times. However, by all qualitative indications, we have never seen the current level of interest and development in AI technologies across so many industries in parallel, massive data availability and vastly improved hardware simultaneously, and billions of dollars of funding that has flooded in ($77.5 billion in 2021¹⁷).

Global spending on AI alone was projected to reach $100 billion by 2024,¹⁸ and contributions to the global economy are estimated to be over $15 trillion by 2030,¹⁹ easily surpassing the biotech industry.²⁰ If we are to learn from the biotech example described at the outset, we need to immediately address the many ethical and legal implications of generative AI as an engine of innovation to maximally benefit from it. I hope we are ready to bet on the future of AI-novation because it will likely have far broader implications than our own narrow industrial applications, such as the pharmaceutical industry in my case, and will likely impact future generations beyond our own with likely unintended consequences that will need to be predicted and mitigated early on.

Statement on Dual Use

Generative AI software such as those examples used for chemical design has potential dual-use capabilities. We, therefore, propose that there should be careful discussions of the ethical considerations of such technologies and as needed consider the implementation of restrictions to control who can access the software and what applications it is used for.

Author Disclosure Statement

S.E. is CEO and owner of Collaborations Pharmaceuticals, Inc.

Funding Information

NIH funding R44GM122196-02A1 from National Institute of General Medical Sciences is kindly acknowledged.