1. For coding, and the reason coders are so excited about GenAI is it can often be 90% right, but it's doing all of the writing and researching for me. If I can reduce how much I need to actually type/write to more reviewing/editing, that's a huge improvement day to day. And the other 10% can be covered by tests or adding human code to verify correctness.
2. There are cases where 90% right is better than the current state. Go look at Amazon product descriptions, especially things sold from Asia in the United States. They're probably closer to 50% or 70% right. An LLM being "less wrong" is actually an improvement, and while you might argue a product description should simply be correct, the market already disagrees with you.
3. For something like a medical question, the magic is really just taking plain language questions and giving concise results. As you said, you can find this in Google / other search engines, but they dropped the ball so badly on summaries and aggregating content in favor of serving ads that people immediately saw the value of AI chat interfaces. Should you trust what it tells you? Absolutely not! But in terms of "give me a concise answer to the question as I asked it" it is a step above traditional searches. Is the information wrong? Maybe! But I'd argue that if you wanted to ask your doctor about something that quick LLM response might be better than what you'd find on Internet forums.
1. For coding, and the reason coders are so excited about GenAI is it can often be 90% right, but it's doing all of the writing and researching for me. If I can reduce how much I need to actually type/write to more reviewing/editing, that's a huge improvement day to day. And the other 10% can be covered by tests or adding human code to verify correctness.
2. There are cases where 90% right is better than the current state. Go look at Amazon product descriptions, especially things sold from Asia in the United States. They're probably closer to 50% or 70% right. An LLM being "less wrong" is actually an improvement, and while you might argue a product description should simply be correct, the market already disagrees with you.
3. For something like a medical question, the magic is really just taking plain language questions and giving concise results. As you said, you can find this in Google / other search engines, but they dropped the ball so badly on summaries and aggregating content in favor of serving ads that people immediately saw the value of AI chat interfaces. Should you trust what it tells you? Absolutely not! But in terms of "give me a concise answer to the question as I asked it" it is a step above traditional searches. Is the information wrong? Maybe! But I'd argue that if you wanted to ask your doctor about something that quick LLM response might be better than what you'd find on Internet forums.