Tell HN: It looks like ChatGPT can read blog posts now | Hacker News Reader

Tell HN: It looks like ChatGPT can read blog posts now

3 points_boffin_6 comments

Me > Summarize this blog post: https://www.federalreserve.gov/newsevents/pressreleases/monetary20230201a.htm

It > The Federal Reserve's press release on February 1st, 2023 announces the results of its latest monetary policy meeting. The Federal Open Market Committee (FOMC) decided to maintain the target range for the federal funds rate at 0.00-0.25 percent. The FOMC also emphasized that they will continue to support the US economy through the ongoing public health crisis and resultant economic challenges. The Committee noted that the US economy is recovering but at a slower pace than they had previously anticipated. To support the recovery, the FOMC plans to continue purchasing Treasury and agency mortgage-backed securities at the current pace. Additionally, the Committee stated that it will continue to provide support to the US financial system and economy through its lending facilities.

Are you sure it's not just making a good guess based on previous similar URLs? I tried asking it for the contents of this post:

"As a language model AI, I do not have access to the current contents of websites. However, the URL https://news.ycombinator.com/item?id=34632249 is a link to a specific item or article on the Y Combinator News website. The contents of the article can vary and it is impossible to determine what it is without accessing the website."

A quick way to confirm that this isn't summarizing the given article is to note that the summary is not what the article you linked actually says: the federal funds rate is over 4% today, and inflation pressures are the topic of this article (not economic recovery and purchasing of securities).

It seems super likely that this saw the URL you requested and regurgitated some representative text from its training data a few years back.

Does it read blog posts/arbitrary URLs or rather is it being regularly updated with new text? Can you ask it to summarize a doc just put online and watched to ensure it's not acquired in any other fashion?

I'm not able to do that at the moment, but that article from the Fed was posted yesterday.

Random thought: How would it handle a word bomb or a zip bomb sort of attack while reading a blog post? To me, this opens up some avenues of attack.

I don't have the experience to test though.

The same way Google handles it when they crawl these things

Reply on news.ycombinator.com