https://youtu.be/3mLYNxgw9wE There is a mantra going around about how "AI is the worst it will ever be" and it becomes dangerously misleading. Or should I say: This is the least misleading it will ever be. More and more people outsource not just the work, but thinking to the AI models and more increasingly AI agents. In fact, I do as well. And it becomes increasingly more uncanny, it becomes very difficult to distinguish reality from fiction and—worst of all—you are unable to verify it if you are not an expert in the subject. It can be seen in many subjects, but one of the most prominent ones is software engineering. Now that people who don't have any experience blindly rely on AI to make sure things are right and that there is no mistakes. The issue is still the same and as old as it was 3 years ago: it can confidently lie and with agentic now it can confidently confuse itself before lying. And even if you can its mistake, the model will simply tell you "you are absolutely right" and…
No comments yet. Log in to reply on the Fediverse. Comments will appear here.