AI chatbots are better at diagnosing illness than doctors?

According to a report, OpenAI’s chatbot scored an average of 90% diagnosing medical conditions from case reports, while doctors scored 76% when using the chatbot and 74% without it.

Here’s the article: https://www.nytimes.com/2024/11/17/health/chatgpt-ai-doctors-diagnosis.html

This result is both surprising and not so surprising. AI has a vast database to draw from, while doctors might excel in some areas and struggle in others. ChatGPT can process immense amounts of medical knowledge instantly, which explains these results. Thoughts?

Welcome to this forum

Posting Guidelines


Follow these tips when sharing news:

  • Posts should be over 100 characters—give details.
  • Share direct links to news articles or blogs.
  • Mention your connection to the source, if any.
  • Add a summary of the news to help others understand the context.

Let the mods know if you need any help or have questions. Thanks for contributing!

I don’t like the title. The AI was only working with reports, not doing in-person checkups. This makes it better at analyzing data, not treating people.

Mika said:
I don’t like the title. The AI was only working with reports, not doing in-person checkups. This makes it better at analyzing data, not treating people.

So, are you saying AI is better with things like digital data?

@Aza
Exactly, but I think the main issue is the title. It’s clickbait and misleading.

Mika said:
@Aza
Exactly, but I think the main issue is the title. It’s clickbait and misleading.

It’s actually the original title from the New York Times article, but yeah, I get what you mean.

Mika said:
I don’t like the title. The AI was only working with reports, not doing in-person checkups. This makes it better at analyzing data, not treating people.

Here’s another article on this topic: Google AI has better bedside manner than human doctors — and makes better diagnoses

Mika said:
I don’t like the title. The AI was only working with reports, not doing in-person checkups. This makes it better at analyzing data, not treating people.

Doctors have the advantage of seeing patients in person, using intuition, and picking up on subtle signs. But even intuition can sometimes lead to errors. AI might eventually interpret visual info too, which could close that gap.

@Ellis
AI can already interpret visual data in some cases. Many diagnoses don’t even need in-person exams, and for those that do, AI is catching up fast. The real value of doctors today is building trust with patients, which AI can’t replicate. But when access to healthcare is limited, AI might provide an affordable option for basic medical needs.

As a doctor, I think these studies highlight a lot of interesting things. AI does well with structured data, like exams or tests, but not so much with unstructured, real-world situations. Diagnosing often involves processing messy, incomplete information. We’re far from AI replacing all of that.

@Archer
That’s true for now, but it’s just a matter of time. AI is already getting better at handling more complex inputs. Imagine if it had access to all patient interactions through a device like smart glasses—it could analyze everything in real time.

@Kim
If AI can interpret all types of data and perform procedures independently, that sounds more like general AI, not just a medical tool. At that point, it would be replacing every kind of job, not just doctors.

@Kim
AI might become capable of a lot, but holding it accountable for ethical medical decisions is a big challenge. Medicine is about more than data; it’s about understanding context and taking responsibility for tough calls.

@Archer
AI is just a tool, like calculators or assembly lines. It feels scary because it’s new, but it’ll just become part of how we work.

I’d trust an AI doctor more than a human. Can’t wait for that day.

Luca said:
I’d trust an AI doctor more than a human. Can’t wait for that day.

Bad idea. AI makes mistakes too. It’s better to use it as a second opinion alongside real doctors.

Nico said:

Luca said:
I’d trust an AI doctor more than a human. Can’t wait for that day.

Bad idea. AI makes mistakes too. It’s better to use it as a second opinion alongside real doctors.

Human doctors make mistakes all the time as well. Nobody’s perfect.

The big hurdle for AI in healthcare is all the regulations. It’ll take a while before they can officially give medical advice.

This doesn’t make me impressed with AI—it makes me worried about our doctors.

Thanks for posting this.