(C) Daily Kos
This story was originally published by Daily Kos and is unaltered.
. . . . . . . . . .



Sometimes, AI is Just Hype [1]

['This Content Is Not Subject To Review Daily Kos Staff Prior To Publication.', 'Backgroundurl Avatar_Large', 'Nickname', 'Joined', 'Created_At', 'Story Count', 'N_Stories', 'Comment Count', 'N_Comments', 'Popular Tags']

Date: 2023-05-02

I am so, so old.

You may have seen a version of this article floating around. It claims that ChatGPT does better than regular doctors in accuracy and compassion. Wouldn't that be amazing if true? Except, well, the comparison they used seems somewhat flawed:

Researchers compared written responses from physicians and ChatGPT to real-world health questions and found that a panel of licensed healthcare professionals preferred ChatGPT’s responses 79% of the time, rating ChatGPT’s responses as higher quality and more empathetic. .... The study obtained a diverse sample of healthcare questions and physician answers from the social media platform Reddit’s AskDocs, where millions of patients publicly post medical questions to which doctors respond. ChatGPT Beats Doctors in Compassion and Quality of Advice to Patients - Neuroscience News

Yeah. A Reddit Forum was their baseline. Now, to be fair, the subreddit does claim that all of the responses are verified as doctors by the mods. But there are no apparent controls for quality and the study doesn't attempt to compare the responses in AskDoc to traditional medicine. What the study has proven, then, is significantly less impressive than ChatGPT is better and nicer than your doctor. ChatGPT might be better and nicer than random doctors in a subreddit. Not the same thing, despite the headlines.

One of the problems we are going to face in this brave new world of AI is that most AI is being developed by commercial companies. That means they have a vested interest in hyping their abilities beyond what they can actually achieve in order to drive investment and sales. Combine that with the tendency of a certain subset of tech people and reports that wants AI to replace human beings, whether to prove their worldview correct or because they desperately want to convince themselves they are working on something earth shattering, that they aren't just wasting their time making tiny improvements in calling a taxi.

And so you get studies that have a ridiculous premise and articles that treat the ridiculous study as if it is ground breaking. Which is a shame, because there are potentially life-altering benefits from some AI systems. Scientists fomr the NHS claim to have developed a systme that is better than the standard methods of detecting cancer early:

The AI tool designed by experts at the Royal Marsden NHS foundation trust, the Institute of Cancer Research, London, and Imperial College London can identify whether abnormal growths found on CT scans are cancerous. The algorithm performs more efficiently and effectively than current methods, according to a study. The findings have been published in the Lancet’s eBioMedicine journal. New artificial intelligence tool can accurately identify cancer | Cancer | The Guardian

Caution is warranted. In the past these kinds of claims have fallen apart because the system was learning some quirk of the training data, not the actual medical condition. And fanciful studies like the Reddit doctor one above inject even more reasons to be skeptical. But in this case, the study authors are clear that this is preliminary data that there is still a lot more testing to be done to verify its usefulness in clinical settings, and that they haven't really proven anything yet. They aren't over-hyping their results, in other words, and they created a reasonable study.

We can and likely will get many benefits from AI systems in the near future. As always, the question will be to what extent the drawbacks outweigh those benefits and how can we mitigate the damage to the lives and livelihoods of ordinary people. Falling for AI hype isn't going to help us answer any of those questions. It will, in fact, make it harder for us to determine what is actually happening and react appropriately. Make sure you are properly skeptical when you see a result that promises the world. You are likely not going to get the world -- just a tiny giftshop plastic globe.

[END]
---
[1] Url: https://www.dailykos.com/stories/2023/5/2/2166951/-Sometimes-AI-is-Just-Hype

Published and (C) by Daily Kos
Content appears here under this condition or license: Site content may be used for any purpose without permission unless otherwise specified.

via Magical.Fish Gopher News Feeds:
gopher://magical.fish/1/feeds/news/dailykos/