At Rise of AI 2019 in Berlin, Charles-Edouard Bouee, CEO of Roland Berger, revisits his 2018 thesis and reports on where AI actually stands a year later. He recaps "portable AI," a personal cloud he compares to Jiminy Cricket and to Joy from Blade Runner 2049, arguing that GAFAM and the Chinese BAT platforms are on the last stretch of their exponential growth because portable AI will eventually replace their revenue sources, while traditional companies fighting digital leaders, startups, platforms and their own IT systems have more future than they think. He says progress has been slower than he expected: every client mentions AI on earnings calls, but few actually use it, and GAFA companies mostly apply basic machine learning to harvest existing data, such as Amazon optimizing warehouse robots and server uptime, rather than building genuinely advanced AI. He argues GDPR benefits Europe by protecting the personal-cloud data that portable AI will need, and gives companies practical advice: start small, focus on real business problems, and treat cloud computing skeptically, since he estimates 95 percent of it is storage rather than processing. He closes with predictions for 2029, picturing AI as everything, everywhere, for everyone.
Key takeaways
- Bouee says traditional companies have more future than expected once portable AI helps them fight digital leaders and platforms.
- He says GAFA companies mostly harvest existing data with basic machine learning rather than build advanced AI.
- He argues GDPR benefits Europe by protecting the personal-cloud data future portable AI systems will need.
- He estimates 95 percent of cloud computing is data storage rather than actual data processing.
- He predicts AI in 2029 will be everything, everywhere, for everyone, reshaping every industry like the internet did.
Detailed notes
Recap: portable AI and human augmented intelligence
Bouee recaps three points from his 2018 talk: portable AI, a personal cloud he illustrates with Jiminy Cricket and Joy from Blade Runner 2049, will be the next big thing and will replace today's platform winners the way earlier ten-year cycles replaced IBM and mobile phones; GAFAM and China's BAT platforms are on the last part of their exponential curve because portable AI will wipe out their current revenue sources; and traditional companies, besieged by digital leaders, startups, platforms and their own legacy IT systems, have more future than they think once positive AI lets them fight back. He also recaps his idea of human augmented intelligence (HAI), arguing humans are coming back into the picture and citing consulting as a profession that will persist.
Why AI adoption is slower than expected
Bouee reports that a year on, the field is moving more slowly than he expected: every client now mentions AI on earnings calls because analysts punish companies that do not, yet very few companies actually use it. He argues that what GAFA companies present as advanced AI is mostly basic machine learning and deep learning applied to harvesting data already collected from users, for example Amazon optimizing a warehouse of 5,000 robots, its logistics chain and server uptime rather than predicting the future. On data protection, he takes the opposite view from Kai-Fu Lee, arguing GDPR is good for Europe because it protects the personal-cloud data that portable AI will depend on in ten years.
Practical advice for traditional companies and startups
Bouee tells traditional companies to "learn to walk, drive, fly" before expecting to "go to the moon," and tells startups to sell clients concrete machine learning and deep learning applications on existing data rather than promises about huge data lakes, noting that he estimates 95 percent of cloud computing is data storage, not processing. He names healthcare, fintech and education as the industries with the most near-term potential because of high transaction costs and volume, citing France as home to two of the world's five best healthcare databases. His golden rules: proofs of concept are easy to sell but production is hard to maintain, so ask what business problem needs solving before proposing AI, favor business-driven innovation, take baby steps, and think about the data first.
Predictions for 2029
Bouee predicts AI in 2029 will be defined by three E's: everything, everywhere, for everyone, touching every industry the way the internet did in the 1990s. He illustrates this with existing products: Predictive, which estimates the odds of winning a legal case; Detective, software that compiles a full briefing on people before a meeting; Neo, which recommends who to team up with on a project; and Netflix, where he says algorithms acting as artificial clients now shape what gets produced. He also cites Google Duplex, a movie script he says was built with Watson from a corpus of science-fiction films, which qualified for Writers Guild of America royalties, and the AI-powered website builder The Grid. He demonstrates the WT2 earphone translator, a device from a Chinese company inspired by the film Wandering Earth by Liu Cixin, that translates spoken conversation in real time across more than 20 languages.
Quotes
First, learn to walk, drive, fly, and then maybe I can bring you to the moon. There's no magic.
Charles-Edouard Bouee
If it's written in Python, it's machine learning. If it's written in PowerPoint, it's probably AI.
Charles-Edouard Bouee
Questions answered in this talk
What is "portable AI"?
According to Charles-Edouard Bouee at Rise of AI 2019, portable AI is a personal cloud, comparable to Jiminy Cricket or Joy from Blade Runner 2049, that he expects to eventually replace the revenue sources of today's platform companies.
Is GDPR good or bad for Europe's AI future?
According to Charles-Edouard Bouee at Rise of AI 2019, GDPR is good for Europe, unlike the view held by Kai-Fu Lee, because in the coming personal-cloud era it will protect the data that portable AI systems need.
How much of cloud computing is actual data processing rather than storage?
According to Charles-Edouard Bouee at Rise of AI 2019, about 95 percent of the cloud is used for storing data, with only the remaining share used for deep learning and machine learning processing on demand.