A BBC story today announces Cambridge dictionary's Word of the Year as
'Hallucinate'.

<https://www.bbc.com/news/uk-england-cambridgeshire-67424335>

Article title: "Hallucinate is Cambridge Dictionary AI-inspired word of
2023"

The 'AI Inspired' bit is rather puzzling. The word hallucinate is
neither new, nor 'AI-inspired'. The word existed long before AI.

There have been - and still are - a lot of hallucinating human beings,
not least the compilers of the Cambridge Dictionary.

On 2023-11-15 05:45, occam wrote:
> A BBC story today announces Cambridge dictionary's Word of the Year as
> 'Hallucinate'.
>
> <https://www.bbc.com/news/uk-england-cambridgeshire-67424335>
>
> Article title: "Hallucinate is Cambridge Dictionary AI-inspired word of
> 2023"
>
> The 'AI Inspired' bit is rather puzzling. The word hallucinate is
> neither new, nor 'AI-inspired'. The word existed long before AI.
>
> There have been - and still are - a lot of hallucinating human beings,
> not least the compilers of the Cambridge Dictionary.

Did you actually read the article?
The 'AI inspired' reason is that a new definition has been added.
There were no claims of the word being new. Only that the new definition
was inspired by what AI appears to do; reporting inaccurate or
misleading information.

--
"What is H3H3?"
"Laughing gas?"
"No, but it should be."

On Wed, 15 Nov 2023 12:45:37 +0100, occam <occam@nowhere.nix> wrote:

>A BBC story today announces Cambridge dictionary's Word of the Year as
>'Hallucinate'.
>
><https://www.bbc.com/news/uk-england-cambridgeshire-67424335>
>
>Article title: "Hallucinate is Cambridge Dictionary AI-inspired word of
>2023"
>
>The 'AI Inspired' bit is rather puzzling. The word hallucinate is
>neither new, nor 'AI-inspired'. The word existed long before AI.
>
>There have been - and still are - a lot of hallucinating human beings,
>not least the compilers of the Cambridge Dictionary.
>
Well, 'AI-inspired' clued me in as to why they picked
Hallucinate. That problem of the Chat-bots was intriguing
to me six months ago -- especially since it is a problem
that they all exhibited. Does that imply something about
human thinking, which also is highly associational? But
I haven't seen recent commentary.

I have hoped to hear that developers have figured how
to apply fact-checking. I'm still waiting.

--
Rich Ulrich

On 15/11/23 22:45, occam wrote:
> A BBC story today announces Cambridge dictionary's Word of the Year
> as 'Hallucinate'.
>
> <https://www.bbc.com/news/uk-england-cambridgeshire-67424335>
>
> Article title: "Hallucinate is Cambridge Dictionary AI-inspired
> word of 2023"
>
> The 'AI Inspired' bit is rather puzzling. The word hallucinate is
> neither new, nor 'AI-inspired'. The word existed long before AI.
>
> There have been - and still are - a lot of hallucinating human
> beings, not least the compilers of the Cambridge Dictionary.

This is part of what I see as a very new problem. The article says
"Inaccurate or misleading information has long been with us, of course,
whether in the form of rumours, propaganda, or fake news." That's
technically correct, I suppose, but it misses an important point: the
/volume/ of misleading information has risen dramatically in recent
years. We started noticing this with the rise in Trumpism, but AI is
also part of the problem. One feature of new AIs is that they can
plausibly lie.

Another part of the problem is social media platforms that are unwilling
to root out lies.

A relatively new organisation in Australia is a right-wing "grass-roots"
movement called Advance, backed by a lot of funding from the very rich.
It has unashamedly adopted the campaigning tactic of muddying the waters
by spreading as many lies as possible. A generation ago this would have
been seen as shocking. Now lying is, apparently, recognised as a
legitimate political strategy.

Recently I mentioned a new referendum in Australia. It was to recognise
indigenous people in the constitution, and initially it was obviously
going to pass. The flood of misinformation completely killed it. What I
see as significant here is that this is the first referendum we have had
in the post-truth era. It looks to me as if no referendum can ever again
succeed again in this country, because the nature of political
campaigning has changed.

OK, most of that is not about AI. But it's relevant now that AI has
become one of the tools used to spread misinformation.

--
Peter Moylan http://www.pmoylan.org
Newcastle, NSW

On 16/11/2023 00:28, Peter Moylan wrote:
> On 15/11/23 22:45, occam wrote:
>> A BBC story today announces Cambridge dictionary's Word of the Year
>> as 'Hallucinate'.
>>
>> <https://www.bbc.com/news/uk-england-cambridgeshire-67424335>
>>
>> Article title:  "Hallucinate is Cambridge Dictionary AI-inspired
>> word of 2023"
>>
>> The 'AI Inspired' bit is rather puzzling. The word hallucinate is
>> neither new, nor 'AI-inspired'. The word existed long before AI.
>>
>> There have been - and still are - a lot of hallucinating human
>> beings, not least the compilers of the Cambridge Dictionary.
>
> This is part of what I see as a very new problem. The article says
> "Inaccurate or misleading information has long been with us, of course,
> whether in the form of rumours, propaganda, or fake news." That's
> technically correct, I suppose, but it misses an important point: the
> /volume/ of misleading information has risen dramatically in recent
> years. We started noticing this with the rise in Trumpism, but AI is
> also part of the problem. One feature of new AIs is that they can
> plausibly lie.

Yes, but 'volume' has never been part of the definition of fake news.
The *real* problem of the definition of AI hallucination is that the AI
*does not know* it is generating false information or fake news. It is
*not sentient* for it to do so.

>From the article:
""Whereas these are normally thought of as human products, hallucinate
is an evocative verb implying an agent experiencing a disconnect from
reality. "

If we had to generate a new entry in a dictionary for every new 'agent'
experiencing something, we should have a new definition of 'dreaming'
for humans, dogs, frogs, etc.

Similarly, we should have a different definition of 'translating' for
the translations an AI does. Or painting, or poetry. In my opinion,
these AI activities do not deserve new definitions. The prefix "AI-"
should suffice.

>
> Another part of the problem is social media platforms that are unwilling
> to root out lies.
>

Unwilling or unable? (see the example of AI generated faces in Larry's
link.) The moment AI can detect 'fakes', you can bet your bottom dollar
that it will be deployed - same as plagiarism detection programs are

On 16/11/23 19:22, occam wrote:
> On 16/11/2023 00:28, Peter Moylan wrote:

>> This is part of what I see as a very new problem. The article says
>> "Inaccurate or misleading information has long been with us, of course,
>> whether in the form of rumours, propaganda, or fake news." That's
>> technically correct, I suppose, but it misses an important point: the
>> /volume/ of misleading information has risen dramatically in recent
>> years. We started noticing this with the rise in Trumpism, but AI is
>> also part of the problem. One feature of new AIs is that they can
>> plausibly lie.
>
> Yes, but 'volume' has never been part of the definition of fake news.
> The *real* problem of the definition of AI hallucination is that the AI
> *does not know* it is generating false information or fake news. It is
> *not sentient* for it to do so.

That is also true for some humans.

--
Peter Moylan http://www.pmoylan.org
Newcastle, NSW