Will AI Lead to Superintelligence or Just Super-Automation?

Summary: Dimitris Tsementzis, from Goldman Sachs’ Applied Artificial Intelligence team, explores whether the current trajectory of AI development will lead to superintelligence—AI surpassing human cognitive capabilities—or result in advanced "super-automation." He discusses the conceptual leap from Artificial General Intelligence (AGI) to superintelligence and highlights the technical and ethical challenges in creating a safe and useful superintelligent system.

Key points include:

• The distinction between AI, AGI, and superintelligence.
• The potential tasks superintelligence could revolutionize, such as solving complex conjectures or designing cures for genetic diseases.
• The challenges of achieving "self-learning" AI, structural reasoning beyond correlations, and asymmetric ethical alignment.
• Practical concerns about whether superintelligence would be comprehensible or commercially viable.

Tsementzis concludes that while super-automation is a more immediate prospect, the pursuit of safe superintelligence remains a significant challenge for researchers and technologists.

References:

1. Ivgi, M., Carmon, Y., & Berant, J. Scaling Laws Under the Microscope: Predicting Transformer Performance from Small Scale Experiments (2022).
2. Morris, M. R., et al. Position: Levels of AGI for Operationalizing Progress on the Path to AGI (2024).
3. Bostrom, N. How Long Before Superintelligence? (1998).
4. Pavlus, J. The Computer Scientist Training AI to Think with Analogies (Quanta, 2021).
5. H?lldobler, B., & Wilson, E. O. The Leafcutter Ants: Civilization by Instinct (2010).

Hashtags: #AI #Superintelligence #Automation #ArtificialIntelligence #Innovation #FutureOfAI #TechTrends #MachineLearning #Research