Artificial Intelligence (AI) is undeniably transforming the global investment landscape. Despite some skeptics, J.P. Morgan Wealth Advisors recently dismissed fears of an AI investment bubble. Predicting the future is impossible (as Peter Drucker said, “The best way to predict the future is to create it.”).
But as a founding member of Microsoft Azure and former CEO of Microsoft Ventures’ first partner in Latin America, I offer an insider perspective on why AI is here to stay and how its economics can drastically be changed for the better.
Generative AI began with the 2017 “Attention Is All You Need” paper by Google scientists, and OpenAI brought AI to the masses in November 2022 with ChatGPT. However, AI research, founded in 1956, is broader and includes Natural Language Processing (NLP), Machine Learning, Neural Networks, Deep Learning, and Generative AI.
During my tenure leading Microsoft’s Digital Transformation team in the Middle East & Africa, we collaborated with top consulting firms as well as system integrators to launch 21 cloud and AI vertical solutions in our first year, deploying them across the top 10 industries prioritized for the region. The thinking was that AI should be part of business and board-level discussions, not just IT.
The main barrier to AI’s wider adoption and meaningful ROI isn’t technology but the outdated process of developing and maintaining AI solutions. Despite numerous open-source tools and data science platforms, the discovery phase remains a manual, trial-and-error process. Achieving the best optimization requires extensive experimentation, which impacts ROI.
Assembling multidisciplinary teams is essential; it includes data scientists, AI engineers, product managers, project managers, and increasingly AI ethicists, computer vision engineers, and AI trainers. Attracting, onboarding, training, and retaining this diverse talent is challenging, costly, and time-consuming. For each project, data scientists and AI product managers must select the most suitable ML algorithms, perform model selection and hyperparameter optimization, and more. It’s humanely impossible to experiment with all possibilities before you choose how you devise an AI solution. And even if you had Mr. or Mrs. Perfect onboard, the best AI solutions degrade over time due to constant changes in data and business rules. That contributes to the 75-85% failure rate in achieving expected outcomes from AI according to studies by Deloitte, BCG, Gartner, and other great companies I’ve worked with during my 15 years at Microsoft.
But what if AI could autonomously handle 70-80% of this work and empower individuals with basic skills to leverage AI’s computational power and speed to do that experimentation at massive scale with the highest possible accuracy? Does that sound like AGI or science fiction?
An Israeli startup, Evolution, founded in 2019, uses genetic algorithms (GAs) to transform R&D economics for IT services companies and companies like Hyundai, Johnson & Johnson, and unicorns like StoreDot. These organizations use Earth – a generative AI solution that rather than creating text or images, builds and evolves optimized AI solutions autonomously and affordably, meeting ever-changing market demands, regulatory environments, and business rules.
Now imagine a world where every business – small, medium and large – can reduce a 1–2-year development cycle to 1-2 months. Moreover, it has the capacity to become an “AI factory” and experiment with dozens of projects simultaneously, and for each of them come up with millions, billions, trillions of possible solutions – freeing their resources from endless manual trial-and-error and letting AI-for-AI naturally select the one AI solution that best fits their needs in a fast, accurate and cost-effective way. Companies like Uber, Google, Microsoft and others already do that – because they have the resources and use Genetic Algorithms in their innovation labs! But now we are talking about more than the “democratization of AI”; we’re describing introducing a new threshold for ROI in AI through a profound change in its adoption curve – and success rate! Now think as an investor – is that scenario more compelling to you?
Is the world ready for AI-for-AI? Can we trust AI to make AI more efficient, accurate, explainable, fair, faster, and more accessible? If this sounds like science fiction, consider that humans evolved through natural selection. AI systems can evolve too, but perhaps not solely by human hands. Will society accept this shift? Only time will tell.
Let’s evolve the way we think about AI research and development. We can learn from biology: just as humans evolved, AI can evolve naturally and autonomously. The evolution in AI R&D will significantly impact AI investment returns, reflecting this advancement.