AI is Revolutionising Innovation!

Several months ago, I spent a couple of days hearing from world-leading innovators about how they are incorporating AI into their innovation processes in a session on Autonomous Innovation run by Board of Innovation. I’m posting this to let my network know that a similar opportunity presents itself this week.

I’ve been following AI for decades now. At first it was just intellectual curiosity about an obscure idea that seemed like science-fiction. But then it became real, and I took courses on coding AI, neural networks, self-driving cars and so on. Recently, it’s become so powerful so fast that the massive risks it brings became stark, and I’ve taken 6 months of intensive training on AI and AI Governance in order to help manage the transition.

AI Governance isn’t about preventing AI, but about finding the right path which maximises the benefits but minimises the risks. It’s the balance that the EU AI Act is trying to get right. As a career-long innovation leader, even as I'm conscious of the risks AI brings, I can’t help being excited by how it helps us innovate – be that in finding cures, modelling proteins or providing personalised education to anyone with an internet connection.

It's hard to keep pace with it all – but this symposium is an opportunity to see the work that’s happening today at the cutting edge of innovation, in “Autonomous Innovation”. (you can still sign up, it's all online)

It’s not about AI replacing humans (at least, that’s the claim!), but rather that humans who work effectively with AI as their “co-pilot” will out-innovate those who do not. In fact, unless you’re a major IT company, AI tools will be a commodity in innovation. Your Chat-GPT (or generic LLM) will probably be no better or worse than mine. The competitive advantage will go to the innovators who best integrate humans and AI, and to those who take this to the next level, for example developing AI models (based on the same core AI technology) to serve a range of different purposes.

If you haven’t been following this closely, you will be amazed at what is now possible. For example, when I worked in P&G, one of my jobs was to talk to consumers about new product ideas. Today, AI can potentially replace both the interviewer and the consumers!! How?

1. A human asks an AI to identify potential gaps or opportunities in the market where there is an unmet consumer need.
2. A human (for now) interviewer prepares questions and/or outlines the areas that they want to investigate. The AI can then run the interviews and find the relevant data from the consumers.
3. They start interviewing real consumers (for now), but after interviewing maybe 20-30 deeply, the AI can create representative virtual model consumers for different consumer “segments.” ?
4. Product ideas can then be tested iteratively by one LLM (the virtual interviewer) interviewing “another” LLM (representing the various model consumers in turn).

Which sounds incredible. Until you see what AI can do in taste-testing for food products (e.g. snacks):

1. Scientists build a database of food ingredients described on a molecular (chemical) level in addition to their physical characteristics and meta-characteristics (e.g. flavour, texture …).
2. They build "virtual consumers" (as above) who serve both to comment on ideas (as above) but also to virtually taste new recipes. These virtual tasters can taste millions of different recipes without tiring or being influenced by what they last tasted.
3. An AI can then "virtually mix" ingredients to create target tastes which are likely to appeal to segments of virtual consumers, then virtually test both concepts and products, using an iterative process (very fast) to reach what seems like the optimum combination, which can then be tested with real consumers. [In my role at a start-up, we used a similar approach to design new molecules which targeted selective adsorption capability, for example to capture CO2 from air.]
4. AI can also do other tasks. For example, if you want to remove some non-sustainable ingredient from a food, the AI can design alternative formulas with the same flavour-profile, often including surprising ingredients which humans would never think of. (The kind of brilliant, creative things that only genius chefs can do).

These are not some futuristic science-fiction scenarios. These exist today (or better, 6 months ago – today they will be more advanced!), and are working successfully in major corporations.

And AI today is the least capable that it will ever be, it’s only going to get better.

All of us need to become much more familiar with both the risks and the benefits of AI, because very soon we are going to have to take critical decisions on how to move forward with it. This feels like a great opportunity to catch up with AI's potential to drive innovation. But even if you're more on the concerned side (as I am, honestly), there is a lot of insight into the capabilities of AI, and just a bit of imagination will let you start to think of ways that a misaligned or misused version of this technology could prove dangerous or harmful.

I'll try to keep a good balance by making my next AI post about AI Safety and Governance (like this one from a few months ago).