However, experts are concerned about the potential misuse of AI technology. During a recent U.S. Senate Judiciary Committee hearing, Dario Amodei, CEO of AI company Anthropic, warned that if AI development continues at its current pace, a wide range of people could gain access to scientific knowledge that even experts do not currently possess within the next two to three years. This could potentially lead to misuse of AI systems in areas such as cybersecurity, nuclear technology, chemistry, and biology.
Key takeaways:
- AI systems have developed rapidly in the last decade due to progress in three inputs—compute, data, and algorithms. The pace of progress is so fast that it has been described as 'scary' by some.
- Compute has been increasing due to Moore's law and the willingness of developers to spend more on training larger models. This has allowed AI systems to process larger amounts of data and draw more accurate conclusions.
- Data is crucial for AI systems to build accurate models. The amount of data used to train AI systems has increased exponentially, with some predicting that developers will run out of high-quality language data by 2026.
- Algorithmic progress has been a key factor in making efficient use of compute and data. Experts predict that AI progress will continue at a rapid pace for the next few years, raising concerns about potential misuse in areas like cybersecurity, nuclear technology, chemistry, and biology.