Send Link mailing list submissions to
[email protected]
To subscribe or unsubscribe via the World Wide Web, visit
https://mailman.anu.edu.au/mailman/listinfo/link
or, via email, send a message with subject or body 'help' to
[email protected]
You can reach the person managing the list at
[email protected]
When replying, please edit your Subject line so it is more specific
than "Re: Contents of Link digest..."
Today's Topics:
1. ChatGPT's hallucination problem is getting worse according to
OpenAI's own tests and nobody understands why (Antony Barry)
2. Re: ChatGPT's hallucination problem is getting worse
according to OpenAI's own tests and nobody understands why
(Tom Worthington)
----------------------------------------------------------------------
Message: 1
Date: Fri, 9 May 2025 15:39:15 +1000
From: Antony Barry <[email protected]>
To: Link list <[email protected]>
Subject: [LINK] ChatGPT's hallucination problem is getting worse
according to OpenAI's own tests and nobody understands why
Message-ID:
<CAECOtWxiG2YJn2wrjM2AC3JAWkrPz4Qe5BazZk9keFJ=bgf...@mail.gmail.com>
Content-Type: text/plain; charset="UTF-8"
Recent testing by OpenAI has found that its latest AI models, GPT-4-mini
and GPT-4 o3, are more prone to hallucinations?producing false or
misleading information?than earlier models like GPT-4 o1. In benchmark
tests, GPT-4 o3 hallucinated 33?51% of the time depending on the task,
while GPT-4-mini hallucinated as much as 79%, compared to o1?s 44%. These
models are part of a new generation of ?reasoning? LLMs designed to mimic
human-like step-by-step thinking, but they are currently producing more
errors, not fewer. Although OpenAI disputes the idea that reasoning models
inherently hallucinate more, the issue raises concerns about the
reliability of AI outputs, especially in applications meant to save time or
effort.
Summary by Chat-GPT
https://www.pcgamer.com/software/ai/chatgpts-hallucination-problem-is-getting-worse-according-to-openais-own-tests-and-nobody-understands-why/?lctg=1980929&utm_source=digitaltrends&utm_medium=email&utm_content=subscriber_id:1980929&utm_campaign=DTDaily20250507
--
Mob:04 3365 2400 Email: [email protected], [email protected]
------------------------------
Message: 2
Date: Sat, 10 May 2025 08:22:08 +1000
From: Tom Worthington <[email protected]>
To: [email protected]
Subject: Re: [LINK] ChatGPT's hallucination problem is getting worse
according to OpenAI's own tests and nobody understands why
Message-ID: <[email protected]>
Content-Type: text/plain; charset="utf-8"; Format="flowed"
On 5/9/25 15:39, Antony Barry wrote:
> Recent testing by OpenAI has found that its latest AI models, GPT-4-mini
> and GPT-4 o3, are more prone to hallucinations ...
Perhaps AI is mimicking human thinking too well. The problem being there
needs to be a way to suppress the daydreams we are all subject to, but
never say out loud. ;-)
--
Tom Worthington http://www.tomw.net.au
-------------- next part --------------
A non-text attachment was scrubbed...
Name: OpenPGP_signature.asc
Type: application/pgp-signature
Size: 665 bytes
Desc: OpenPGP digital signature
URL:
<https://mailman.anu.edu.au/pipermail/link/attachments/20250510/bb797fa7/attachment-0001.sig>
------------------------------
Subject: Digest Footer
_______________________________________________
Link mailing list
[email protected]
https://mailman.anu.edu.au/mailman/listinfo/link
------------------------------
End of Link Digest, Vol 390, Issue 8
************************************
