By thetransmitter.org
Publication Date: 2026-01-19 05:00:00
Artificial intelligence agents have amazing strengths, but also weaknesses. They are great programmers: They can turn ideas into working computer programs much faster and with fewer errors than almost all people. The newest agents are so good at math that the International Mathematical Olympiad is now asking AI companies to give human participants a day to enjoy their success before the AI results are announced. And these agents are also becoming better at visual thinking, such as interpreting graphical data charts. On the other hand, “hallucinations” still occur and can have real consequences. Lawyers, for example, had to expect fines citing non-existent cases who were hallucinated by large language models.
Scientists are enthusiastic about the possibilities – now even more than 60 percent of researchers use AI to support their work–but they are understandably cautious about putting too much faith in LLMs until we find ways to ensure…
