Re: Can AI distinguish "libre" from "gratis"?

[Jean Louis]

* Akira Urushibata <afu@wta.att.ne.jp> [2025-04-28 20:08]:
> Other items of the list discussed ticket prices, access to the
> grounds.  Obviously the above text was generated by AI, for it is
> unlikely for a human being to err in this manner, but Google does not
> clarify.  (Confirmed April 28 Japan time)  So, one part of Google
> (probably AI) treats the exposition as canceled while another part
> understands that it is in progress.  There is a failure to notice
> an obvious contradiction.

It is their organization. You can complain to them. Take it
realistically, you can't do nothing much about organizations that are
not under your control.

Don't expect "corporate responsibility" as they are money driven, they
have responsibility to making income, not to random people.

> Can AI distinguish between "free" as in "free drinks" and "free" as in
> "freedom"?

Run the so-called "AI" or the Large Language Model (LLM) on your own
computer and find it out!

See https://www.gnu.org/philosophy/words-to-avoid.html#ArtificialIntelligence

The word AI as such is not wrong, but when used in wrong context for
sensationalism or anthropomorphism of computer programs it leads
people to wrong conclusions. Point is that it doesn't think.

Your message proves that it doesn't think.

Back to question if Large Language Model (LLM) can understand "free" as in 
freedom related to free software:

- ask your favorite LLM what is free software. Sample answer by
  GLM-4-9B-0414-Q8_0.gguf Large Language Model (LLM) says that it can
  understand it: 
https://gnu.support/files/tmp/clipboard-2025-05-01-08-35-21.html

If you start talking to the LLM in the context of price, it may probably answer 
in that way. Let's try:

> I can't afford Microsoft proprietary software, should I instead use free 
> software?

Answer:
https://gnu.support/files/tmp/clipboard-2025-05-01-08-37-51.html

Answer is quite aligned to meanings of free software.

I have other 100 models which I could test, but there is no need to
exaggerate. I can tell you that my experience is that it does
understand the meaning of free software as majority of models have
been trained on information which provides those meanings.

> We should watch out for this. 

You can watch, but is waste of time.

If you wish to make things right, take datasets and fine-tune some
models. But I think that is also waste of time.

The effort to correct society on the importance of free software,
especially when common knowledge lacks sufficient references, would be
like trying to teach a new language to a community that still speaks
an old one.

Reality is that there is sufficient information on free software and
that information is part of each Large Language Model (LLM) so far
that I have tried out, it is more than 100+ different Large Language
Models (LLM).

> The above makes me think that we should not expect much.

Expecting is passive attitude of an observer. You can be causal and
influence society. 

First is to understand that you can run the Large Language Model (LLM)
locally.

THUDM/GLM-4: GLM-4 series: Open Multilingual Multimodal Chat LMs | 开源多语言多模态对话模型
https://github.com/THUDM/GLM-4

Get your hardware and run it locally. 

There is no point in staying behind the computer technology
developments. Run it locally, do the research, understand that there
there many many different Large Language Model (LLM) and then start
benchmarking and testing.

> Even if we observe AI getting the distinction right in a number of
> cases we should not expect it to be always correct.

Good conclusion!

I consider all information generated by the Large Language Model (LLM)
as a draft of a possibly correct answer.

> For one thing many people are confused and much has been written
> based on the mistake.  it is unreasonable to expect that none of the
> erroneous material has been fed to the large language model (LLM)
> neural networks for training.  In addition the above example
> indicates that AI is not good at coping with lack of integrity.
> Perhaps a contradiction which is obvious to a human being is not
> so for AI.

What is erroneous material or not, is not subject for the Large
Language Model (LLM). That is subject of human discussions and
opinions. 

There is no right or wrong for the Large Language Model (LLM).

Datasets used in training are by the rule biased by people who created
There is no right or wrong for the llmthe information.

There is no right or wrong for the Large Language Model (LLM).

It is like watching TV stations over the world. You cannot get the
single objective truth, but you get variety of human information where
you need to keep the principle of deciding what is right or wrong for
yourself.

Anybody who thinks “let me teach Large Language Model [LLM] what is
right or wrong” is automatically creating a new biased and censored
version that won’t represent information to the user so they can
decide about it, as what was considered wrong by the author may not be
wrong for the user.

Your attempt of wronging any Large Language Model (LLM) is futile. It
is text generation tool. To blame anyone condition is to know who was
responsible for "training" of the Large Language Model (LLM). And why
do that in first place? Anybody can make their own Large Language
Model (LLM), similarly like people are creating websites individually.

> Related article: Researchers have found methods to investigate what is
> going on inside the neural networks that power AI.  They have
> discovered that the process differs greatly from human reasoning.

There are sciences of mind though so far there is no consensus how
human mind works exactly. I don't think we have a stable information
about human reasoning that could be compared to computer base neural
networks.

But -- when I hear that researches have found methods to investigate
what is going on inside the neural networks, it only means that those
were beginners, and there were previous researchers who created
neuaral networks knowing exactly what is going on.

So article is in the sense contradictory and sensationalist.

> We Now Know How AI 'Thinks'--and It's Barely Thinking at All
> https://www.yahoo.com/finance/news/now-know-ai-thinks-barely-010000603.html

Is not worth reading even. It is made for Yahoo readers, part of entertainment. 

It is well established that Large Language Model (LLM) doesn't
think. There is no "barely thinking", there is ZERO thinking. It is
data and computer program giving you output of that data, giving the
impression of mimicking, due to usage of natural language.

I think it is greatest tool so far in 21st century as it empowers
human to be 10 times or 100 times faster in various tasks.

Of course, it doesn't think.

> Though I don't follow developments in this field closely, what is
> written in this article confirms my suspicions.

Good cognition!

> Another way to see this is that the AI developers now have the
> equivalent of a debugger which allows them to probe the internal
> process.  This is likely to affect development.

Of course! All tokens can be seen, analysed, layers of the Large
Language Model (LLM) can be visually inspected, analyzed to every
detail when necessary.

Jean Louis