Hacker Newsnew | past | comments | ask | show | jobs | submitlogin
OpenAI blames Open-source library for Chat History Bug (twitter.com/sama)
29 points by behnamoh on March 22, 2023 | hide | past | favorite | 17 comments


Maybe it's my bias, but the tone of the title seems somewhat misaligned with the actual tweet.

They provided a reason for the issue and said they felt bad. Using the word "blames" makes it seem as if they're shirking responsibility and blame someone else.

The tweet tells us that there was a bug, the bug caused an issue, and the bug has been fixed.


I don't know if it's your bias. I have no particular love for OpenAI but it seems sensational to characterize this tweet as trying to shift blame.


Yeah @dang the title is badly editorialized.


The actual Tweet:

    we had a significant issue in ChatGPT due to a bug in an open source library, for which a fix has now been released and we have just finished validating. a small percentage of users were able to see the titles of other users’ conversation history. we feel awful about this.
Imagine if the bug was caused by closed-source library. Then maybe he'd have posted

    we're having a significant issue in ChatGPT due to a bug in a proprietary, closed-source library. Unfortunately the developers have not responded to our report, and as we do not have access to the code, we cannot fix the bug ourselves. we feel awful about this. perhaps this is why OSS is so widely-accepted in the software engineering community, and many of the widely-used libraries of today are open source.


So when it works gloriously it’s OpenAI’s private genius, when it breaks it’s the fault of open source… I see.


More disgusting every day. We need open source competitors.


Blaming open source for your shortcomings isn't professional at all.


Why is it not just stating a fact? Bugs happen - they have notified users and are fixing it - I don’t see how you read “blame” into one tweet.

I’m no fan, I just don’t think your response is fair or appropriate for HN.

The tweet text is:

  @sama 3h
  we had a significant issue in ChatGPT due to a bug in an open source library, for which a fix has now been released and we have just finished validating.
  a small percentage of users were able to see the titles of other users’ conversation history.
  we feel awful about this.


You're correct, it is just a fact, but he's presenting in a misleading way. The statement presents the issue as if a third party "open source" group is responsible, not OpenAI. In reality, anyone who uses open source code is 100% liable for all its faults -- their bugs are now your bugs. It's not a bug in an open source library, it's a bug in ChatGPT.


The HN title primes us using the clickbaity word “blames”. Without that title, only a cynic would read that much into what is written by a technical person with a majority technical audience. I think your presumption is that Sam is an idiot (trying to pass the buck to a technical audience would be idiotic), and my presumption is that Sam is smart but in a hurry.

Go ahead and blast SamA for writing something transparently as an engineer, presumably quickly without over-thinking it. But don’t be surprised if he starts to write using highly conditional PR-talk with zero substance.

Meanwhile, wait and see their incidence report. That is writing for a discerning audience, and any whiff of passing-the-buck would be responded to harshly by the highly technical audience. This is an expected class of error, and it happens when a company develops and deploys fast, but I expect they will still try and change some processes to try and avoid similar issues (hard though that is)...

Apparently ‘we identified an issue, we are fixing it as fast as we can’ is not enough for some people. Some people also want Sam to use weasel words to avoid mis-reading of a single paragraph of text. A level of perfection few of us could achieve.


"It was just hallucinating because of no input!" denials before now seem...not great.


Fake it until you make it. You have to respect the hustle. Maybe if we learn from him instead of dwelling on the past we can become leaders of billion dollar corporations too, king!


I know you’re being sarcastic, but my god, I’d rather die dirt poor than live this way. There’s a dignity and peace in acting honourably and lovingly


Maybe they shouldn't release the bugfix in light of competitive landscape. I also thought the tone of the tweet was weird. After all, why would NotOpenAI use an open source library


They probably added "open source" to indicate that they are not at fault, which is understandable. However, they could have stated "bug in a third-party library" instead. By emphasizing that it was Open Source at fault, they perpetuate the "You get what you pay for" anti-open source rhetoric that was prevalent a some years ago.


I bet it was probably some sort of cache. Maybe it has a memory leak when hitting openai scale


This entire Twitter thread is an utter dumpster fire. No surprise that Twitter has been left with the most angry people.




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

Search: