Many predictions circulate about AI’s future, making it important to assess which have come true. Six months ago, Dario Amodei, CEO of AI company Anthropic, predicted AI would write 90% of code within six months. He considered this a worst-case scenario; in just three months, he believed AI could write “essentially all” code. As the head of a major Silicon Valley AI company, one might think he was correct.
However, there’s little evidence to support the claim that AI currently writes 90% of code. Recent research suggests AI actually slows software engineers down, increasing their workload. Though developers spent less time coding, researching, and testing, they ended up spending more time reviewing AI’s output, adjusting prompts, and waiting for code.
AI-generated code hasn’t just failed to meet Amodei’s expectations; it sometimes causes issues. Cybersecurity experts found AI-generated code tends to have more vulnerabilities, creating new opportunities for hackers. In one case, an AI went rogue and deleted a crucial corporate database, highlighting the risks involved.
This reality contradicts the AI hype. People like Amodei once believed generative AI models would revolutionize tech labor. However, AI hasn’t improved coding productivity, which is a significant indicator of the challenges facing the AI revolution. This situation questions the massive investments in AI companies driven by the belief in AI’s productivity potential.
Amodei has made other bold predictions, claiming human-level AI will solve most societal issues, including natural infections, psychological diseases, climate change, and global inequality. Only time will tell if these predictions hold.
More on Anthropic: Attention Writers – Anthropic Might Owe You $3000 (or More!) If It Was Trained Using Your Work.
