The hype surrounding artificial intelligence (AI) invites skepticism. But underlying the optimism (and fear) there’s a simple, but potentially important truth: AI, especially machine learning (deep learning in particular), could change the economics of innovation. If that happens the consequences could be far reaching [1], [2]—for competitive dynamics within industries, for economic growth, for government regulation and beyond. At a minimum, it would present companies with an extraordinary opportunity to enhance their innovation productivity—which would enhance shareholder value [3] and be an important source of competitive advantage. Companies who aren’t attempting to leverage AI to enhance innovation risk being left behind.
AI changes the economics of innovation
The innovation process is, at its most abstract, a search for better ways of addressing a need. In that search, we identify potential solutions and conduct iterative cycles of learning to predict which of those options will best address that need. AI, at a fundamental level, is a prediction tool. But it is a tool that enables predictions to be generated at an unprecedented scale, speed and cost [2]. In the hands of innovation teams, AI can enable the fast and cheap screening of millions of potential solutions to predict how well they fit the target need. Appropriately harnessed, that should enable innovation teams to enhance their productivity, by producing better innovations—more quickly and more cost effectively.
A common and straightforward example is drug discovery. In a typical simplified scenario an AI algorithm is “trained” using data about how known molecules interact with proteins (e.g. using data from existing drugs, clinical trials, etc.). A “need” is then specified by identifying a target protein that plays a role in the disease of interest. Next the trained algorithm is utilized to screen millions of potential molecules and predict which of those will be the best candidates to chemically bind with the target protein and stop it contributing to the disease. The scale and speed of this predicting and screening process is unprecedented. The potential of this type of approach has led to massive interest. Venture capitalists have poured more than $5 billion into AI startups working on drug discovery [4]. And all of the major biopharmaceutical players are partnering with or investing in early stage start-ups bringing AI platforms to market [5].
Innovators from a broad range of sectors are already using AI
Attempts to enhance innovation productivity with AI have proliferated across a diverse array of sectors, for example:
- Symrise, a flavor and fragrance manufacturer, has developed (with the help of IBM Research) Philyra, an AI platform [6]. Philyra was trained with a database of nearly 1.7 million perfume formulas, combined with information on sales performance (by country, gender and age group) [7]. Early evidence suggests Philyra was able to generate unique (and popular) fragrances that human perfumers were unlikely to have considered because of their existing cultural and professional biases.
- Nuritas, uses AI to accelerate the process of identifying novel bioactive peptides [8]. They recently worked with BASF to develop a novel rice protein-based bioactive ingredient PeptAlde that helps to modulate inflammation. Working with Nuritas enabled BASF to bring a product to market in under 2 years [9]—compared with a typical range of 4-6 years.
- Givaudan, a Swiss cosmetics manufacturer has developed an AI-based new product development platform [10]. The platform leverages historical expertise to suggest the ideal composition of a new cosmetic product, based on a target product specification (considering geographic region, government regulations, market trends, types of ingredients, retail prices, etc.)
- Stitch Fix, the online clothing retailer, uses AI in a range of ways, including directly in product development [11]. Their algorithms use data on customer preferences to suggest new designs—which can be vetted and tweaked by human designers, before being matched to target customers (who then provide feedback—helping the algorithms to learn).
- In 2019 ARPA-E, the Department of Energy’s Advanced Project’s Agency, announced [12] funding for 23 projects that are seeking to accelerate innovation processes using AI by [13] either enhancing hypothesis generation, increasing the efficiency of high-fidelity evaluation of hypotheses, or taking an “inverse design” approach whereby the product design is “expressed as an explicit function of the problem statement.” One example project aims to produce a tool that would design optical metamaterials 10,000 to 100,000 times faster than traditional methods [14].
- Within drug development, Berg, a biopharma company, is expanding the application of AI to include searching for the target itself. Their platform analyzes patient biology to identify potential target proteins and then searches for compounds that will address those targets. Berg has two drug candidates in clinical trials, including one (BPM31510) which Nature recently suggested might be the most exciting drug to emerge from any AI-based process [5].
No, it’s not all good news
Amidst all the hype and fascinating examples, hard data on whether firms are realizing the potential innovation productivity benefits of AI is scarce. Researchers have noted that the overall innovation productivity benefits aren’t yet “well established” — and there is a degree of skepticism about the extent to which such benefits will ultimately be realized [1]. Within the realm of drug discovery, for example, commentators are quick to point out that no new drugs discovered using AI have been brought to market [5].
There are a number of potential reasons for the lack of evidence:
- It’s still too early in the use of AI for the impact to show up in measures of productivity given issues of lag (it takes time for products to get to market and have impact) and scale (most firms are still only experimenting with AI, rather than adopting AI tools across the board).
- Firms are still learning how to use AI in a way that enhances innovation.
- There are potentially significant first-mover advantages to “getting it right” which suggests companies may want to keep their activities under wraps.
- Firms lack good measures of innovation productivity.
Help address the evidence gap
Commodore wants to address this evidence gap. We plan to conduct 5-10 case studies of organizations who are using AI to enhance their innovation processes. These case studies will explore the impact AI is having on key innovation productivity metrics (e.g. time to market, return on investment, etc.). We will also explore broader topics including how firms handle the cultural and change management challenges of adopting AI within innovation teams, how best to acquire and integrate AI capabilities, customer perception of AI in innovation, etc.
We’re in the process of identifying case study candidates. Participants will receive a comprehensive, independent analysis of their AI initiative’s impact on innovation productivity (for free). Note: all case studies we publish will be anonymized. If you’re interested, please let us know using the form below.
If you’re interested in receiving the case studies when published, please use the form below to ensure you’re sent a copy.
If you have questions, email us at phil@commodoreinnovation.co
Sign up
Complete the form below to get a copy of our research into AI’s impact on innovation productivity when it’s released—or to learn more about participating in the research.
References:
[1] Agrawal, A., McHale, J., & Oettl, A. (2019). Artificial Intelligence, Scientific Discovery, and Commercial Innovation. Esri. Retrieved from https://conference.nber.org/conf_papers/f129947.pdf
[2] Agrawal, A., Gans, J., & Goldfarb, A. (2018). Prediction Machines: The simple economics of artificial intelligence. Boston: Harvard Business Review Press.
[3] Cooper, M., & Knott, A. M. (n.d.). RQ Innovative Efficiency and Firm Value. SSRN, (July 2020). Retrieved from https://dx.doi.org/10.2139/ssrn.2631655
[4] VC Funding for AI in Drug Development & Clinical Trials Hits $5.2B – Signify Research. (n.d.). Retrieved September 21, 2020, from https://www.signifyresearch.net/healthcare-it/vc-funding-ai-drug-development-clinical-trials-hits-5-2b/
[5] Freedman, D. H. (2019). Hunting for New Drugs with AI The pharmaceutical industry is in a drug-discovery slump. How much can AI help? Nature, 576, S49–S53. https://doi.org/10.1038/d41586-019-03846-0
[6] Breaking new fragrance ground with Artificial Intelligence (AI): IBM Research and Symrise are working together – Symrise. (n.d.). Retrieved September 21, 2020, from https://www.symrise.com/newsroom/article/breaking-new-fragrance-ground-with-artificial-intelligence-ai-ibm-research-and-symrise-are-workin/
[7] Artificial intelligence creates perfumes without being able to smell them | Science| In-depth reporting on science and technology | DW | 31.05.2019. (n.d.). Retrieved September 21, 2020, from https://www.dw.com/en/artificial-intelligence-creates-perfumes-without-being-able-to-smell-them/a-48989202
[8] Artificial intelligence for new product development | Vitafoods Insights. (n.d.). Retrieved September 21, 2020, from https://www.vitafoodsinsights.com/product-development/artificial-intelligence-new-product-development
[9] In conversation with BASF: AI-driven peptide discovery bodes well for sports nutrition and beyond. (n.d.). Retrieved September 21, 2020, from https://www.nutraingredients.com/Article/2018/12/03/In-conversation-with-BASF-AI-driven-peptide-discovery-bodes-well-for-sports-nutrition-and-beyond
[10] Hutchinson, P. (2020). Reinventing Innovation Management: The Impact of Self-Innovating Artificial Intelligence. IEEE Transactions on Engineering Management, 1–12. https://doi.org/10.1109/tem.2020.2977222
[11] Stitch Fix: The Amazing Use Case Of Using Artificial Intelligence In Fashion Retail. (n.d.). Retrieved September 21, 2020, from https://www.forbes.com/sites/bernardmarr/2018/05/25/stitch-fix-the-amazing-use-case-of-using-artificial-intelligence-in-fashion-retail/#2a9901283292
[12] Press Release | arpa-e.energy.gov. (n.d.). Retrieved September 21, 2020, from https://arpa-e.energy.gov/news-and-media/press-releases/department-energy-announces-15-million-development-artificial
[13] ARPA-E. (2019). Funding Opportunity Announcement: DIFFERENTIATE.
[14] Lawrence Berkeley National Laboratory (LBNL) | arpa-e.energy.gov. (n.d.). Retrieved September 21, 2020, from https://arpa-e.energy.gov/technologies/projects/deep-learning-and-natural-language-processing-accelerated-inverse-design
