Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

I've been using ChatGPT 4.0 to summarize articles I don't have domain knowledge in, in case anyone is interested: The research looked into whether changing when and for how long mice with lung cancer could eat would affect the growth of their cancer. The idea was that sticking to a strict 6-hour window for eating each day could mess with the cancer's "body clock" and slow its growth. They tested this by allowing some mice to eat whenever they wanted and others only during a 6-hour period.

They used various tests and experiments with cancer cells and mice to see how the restricted eating schedule affected the cancer. The findings were promising: the cancer didn't grow as much in the mice with the restricted eating times. They noticed that this eating pattern seemed to disrupt the activity of a specific gene related to the cancer's internal clock, leading to an increase in a cleanup process in the cells (called autophagy), which helps to break down and recycle parts of the cell.

In simple terms, by controlling when mice ate, researchers were able to slow down the growth of lung cancer. They also think that if they combine this eating strategy with other treatments that turn off the cancer's clock gene, they might get even better results in fighting cancer.



This is a decent summary. I'd have changed the prompt to generate four bullets, personally, but this is a good use of chatgpt, as long as it aids learning rather than replaces it maybe?

The downvotes are likely because it's autogenerated content and doesn't add much to the discussion. Summary bots are a reddit thing.


Fair point on not adding much to the discussion. My reason for posting was that I found the article too jargon heavy for my taste and I thought the output was good and might be of use to someon else.


Bad science journalism, now even faster.

It's ok for your own perusal, but that's it.


You’re free to not like autogenerated content on that basis alone.

But is there something in particular you find “bad” about this summary?


1. It's almost as long as the abstract.

2. Yet it manages to leave out all the details mentioned in the abstract.

3. Some of the text is wrong, e.g., "The idea was that sticking to a strict 6-hour window". Instead, the idea was "circadian rhythm by dietary strategy," which follows as a logical step from the very first sentence. 6 hour TRF was an arbitrary step in this process. It also doesn't mess with the cancer's body clock, as the summary says, but with the host's.

4. The concluding remark, "by controlling when mice ate, researchers were able to slow down the growth of lung cancer," suggests hope (as does "the findings were promising", which does not have relation to the article) when in reality they found a small effect in lab cell lines and mice, and a very artificial diet. It also omits that chemically induced cancer had less chance on the diet, as well as the cautions at the end of the discussion.

5. Given the statistics, it is uncertain this can be reproduced fully.

6. And I think the last sentence "combine ... better chance" is an unnecessarily emotionally charged misinterpretation of "Combining TRF with TIM inhibition further enhanced the anti-tumor effect." The effect of this treatment is worse than chemo-therapy, and the combined effect is unknown.

It simply does not do a good job. It's as bad as superficial science journalism, i.e. click-bait, but nobody even looked at it, let alone thought about it, before pasting it on a publishing medium. Nobody is responsible for this text. It could have said the opposite, and no-one would have batted an eyelid.


They said the science was bad, not the summary.

If you’re curious, the reason the science is bad because it’s explanationless


I like this response. Can I ask what was your prompt for this type of summary?


I asked it to summarize the article but it was still too domain specific so I asked it to summarize in layman's terms




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: