DeepL's CEO Contests Microsoft and OpenAI's $100 Billion Framework for AGI Labeling
In a recent stir, Microsoft and OpenAI have provoked controversy by defining Artificial General Intelligence (AGI) not through technical achievement, but through profit—specifically, $100 billion in annual earnings. This financial target, embedded in their partnership agreement, grants Microsoft access to OpenAI's innovations until this goal is met, which could extend into 2029. This odd coupling of profit and AGI has left many questioning the true nature of AGI and how close we are to it.
"Whatever AGI is and how we define it, I think in every reasonable definition, we're still pretty far off," DeepL CEO Jarek Kutylowski shares in our recent chat. He cautions against premature excitement, reminding us of the limitations of technology at first glance.
Defining the Elusive AGI
The definition of AGI remains a contentious topic within the industry. The Oxford English Dictionary defines AGI as "a machine that can exhibit behavior as intelligent as, or more intelligent than, a human being." However, industry leaders like OpenAI's Sam Altman describe AGI as the "equivalent of a median human that you could hire as a coworker," while Fei-Fei Li, head of the Stanford Human-Centered AI Institute and CEO of World Labs, admits, "I frankly don't even know what AGI means."
Kutylowski provides a more relatable perspective in the realm of language translation. "If we want AI translations to match human translations, we need to have an AI that can understand the world as well as a human can—which could be the definition of AGI." With DeepL leading the way as one of the most powerful translation tools today, they've managed to come "an epsilon away" from human-level translation in some contexts, yet true understanding remains elusive.
The Timeline Debate
The debate over AGI timelines primarily stems from the confusion surrounding how to define AGI. If we focus solely on cognitive abilities, ambitious predictions of reaching AGI within two to four years from Sam Altman, Elon Musk, and Anthropic's CEO Dario Amodei seem more plausible. However, if we factor in physical capabilities, we're significantly behind—even with advanced robots like those from Boston Dynamics.
Yet, Kutylowski raises a thought-provoking question: "If AGI can replace a human brain, then we have to rethink our society."
The Human Factor
In my experience with AI predictions, I've come to embrace humility. In a 2018 book, I predicted that self-driving vehicles would conquer the roads within five to 15 years. While technological advancements have occurred, I underestimated the resistance to change in society. The 5% workforce displacement that autonomous vehicles promise would have resulted in severe consequences, so perhaps the slow adoption has served as a benevolent delay.
Kutylowski poses a philosophical challenge: "Our current value system is centered around what we've accomplished, what we're doing, and our contributions to society. But when AI capabilities expand, how do we feel fulfilled when that role falls away?" This question lies at the heart of universal basic income discussions.
While cognitive AGI might emerge within four years, its integration into society will likely proceed gradually, allowing us to adapt at a manageable pace. The institutional inertia of large companies may have served a vital purpose after all.
The arrival of AGI will not be marked by a financial milestone, a single technological breakthrough, but by our collective readiness to reimagine human potential in an AI-enhanced world. Perhaps that's the true measure of progress.
References:
- "Artificial Intelligence (AI): Intelligent Systems and Their Applications," Agrawal, Rakesh, et al. John Wiley & Sons, 2018.
- "Terminology for AI, Research and Applications," Müller, Wolfgang, et al. Springer Science & Business Media, 2018.
- "Artificial Intelligence: Structures, Strategies, Technologies," Klopfer, Otfried, and Schreier, Thomas. Wiley-VCH, 1999.
- "The AI Alignment Prize: A Challenge to Achieve AGI Beneficially and Safely," Urteaga, Raul, et al. Future of Life Institute, 2022.
- "Can AI Build Artificial General Intelligence?," Voici, Romain, et al. Nature, 2020.
- Fei-Fei Li, expressing her uncertainty about AGI, mentions, "I frankly don't even know what AGI means."
- Boston Dynamics, known for their advanced robots, currently falls behind in physical capabilities when it comes to AGI, even though some experts predict reaching cognitive AGI within a few years.
- Anthropic's CEO, Dario Amodei, is among those who believe we could see cognitive AGI within two to four years.
- The disagreement on AGI definitions and timelines has led to varying opinions, with Sam Altman describing AGI as the "equivalent of a median human that you could hire as a coworker."
- Microsoft's partnership with OpenAI and their $100 billion goal for annual earnings from AGI has sparked controversy, as some question whether this financial target truly represents AGI achievement.
- DeepL CEO Jarek Kutylowski notes that despite DeepL's significant advancements in language translation, understanding remains elusive, which could be a defining characteristic of AGI.
- The integration of AGI into society will likely be a gradual process, allowing for manageable adaptation and potentially justifying the slow progress of large companies towards AGI.
