Paleonet: Mucking about with ChatGPT

Fri Oct 13 10:52:18 UTC 2023

On Fri, Oct 13, 2023 at 11:39 AM Luis Azevedo Rodrigues <
laz.rodrigues at gmail.com> wrote:

> Dear colleagues,
>
> If you allow me to share some previous practical experience in the last
> year:
>
>
> At our Science Centre, we've been using some AI tools, particularly in
> content production (video, text and images)
>
> Briefly:
>
> 1 we use the paid version of ChatGpt, version 4.
> 2 most of the results are good, despite the fact the ALL of them need fact
> checking.
> 3 the results vary significantly on how to write your prompt - e.g. if you
> start your prompt with "act as a paleontologist..." the results will be
> different that a more general prompt.
> 4 If you "feed" ChatGpt, both with upload PDF, or with text, the results
> improve a lot.
> 5 We used several text to image tools and we finally decided to buy
> Midjourney and, like many others, the results of drawing dinosaurs were
> awful opposite to what happened with landscapes - please see attached
> examples.
> 6 Dall-e 3 (free), accessed through Bing, although the results are (still)
> not as good or configurable as Midjourney, improved a lot in the last
> months.
> 6 and advise my team to use perplexity,ai instead of Google - besides
> giving references from the information sources, Perplexity write a brief
> resume of the subject or theme instead of a list of links.
>
> And don't want to be exhaustive, feel free to ask.
>
> [image:
> laz.rodrigues_a_landscape_of_cretaceous_vegetation_the_fossil_r_87ecf028-6648-4c4b-b582-687b2c2780ef.png]
>  [image:
> laz.rodrigues_a_lush_forest_where_giant_tree_ferns_grow_with_hu_6c7d191f-d3b2-4c37-b168-a9e04f66406c.png]
> [image:
> laz.rodrigues_Capture_the_majesty_of_the_sauropod_dinosaur_in_m_6031a547-d2d7-45e6-8613-6e6097aeec4c.png]
>
> Best,
>
> Os melhores cumprimentos,
>
> Luis Azevedo Rodrigues
>
> Executive Director
> Lagos Ciência Viva Science Centre
> http://lagos.cienciaviva.pt
> <http://lagos.cienciaviva.pt/>____________________________
> Paleontologist (PhD) | Science communicator
> Editor of Geoscience Communication journal
> <https://www.geoscience-communication.net/>
>
> Rua Dr. Faria e Silva 34
> 8600-734 Lagos
> Portugal
> +351 282 770 004 | GPS 37.104054, -8.673368
> Personal mail | laz.rodrigues at gmail.com
> Skype: Luis_Azevedo_Rodrigues
> ___________________________
> Science Blogs Brazil <http://bit.ly/1P4qqYd>  https://bit.ly/1P4qqYd
> Youtube  <http://soundcloud.com/ciencia_viva_conversa/>
> https://www.youtube.com/@Luis_Azevedo_Rodrigues
> Academia <http://bit.ly/1V8OKzn>  https://bit.ly/1V8OKzn
> Twitter <https://twitter.com/cienaonatural>  https://bit.ly/2T6oGvh
> LinkedIn <http://bit.ly/2wtAsSU>  https://www.linkedin.com/in/
> luisazevedorodrigues/
>
>
> On Fri, Oct 13, 2023 at 10:57 AM Jere H. Lipps <jlipps at berkeley.edu>
> wrote:
>
>> From Nature:
>> "Large language models (LLMs), such as ChatGPT, could become regular
>> assistants for writing manuscripts, peer-review reports and grant
>> applications. These artificial-intelligence (AI) tools could change how
>> scientists interrogate and summarize results
>> <https://nature.us17.list-manage.com/track/click?u=2c6057c528fdc6f73fa196d9d&id=cfe46567db&e=69bb79007c>,
>> producing ‘papers on demand’ from experimental data and vastly expanding
>> the scope of meta-analyses and reviews. But publishers worry that LLMs’
>> propensity to make up information might lead to a flood of error-strewn
>> manuscripts — and possibly AI-assisted fakes. And because LLMs trawl
>> Internet content without concern for bias, consent or copyright, their use
>> is “automated plagiarism by design”, suggests cognitive scientist Iris van
>> Rooij."
>>
>> I'm not certain that it's plagiarism.  AI does not always copy material
>> but mixes it up with paraphrased stuff.  Stealing the material to begin
>> with is another issue.
>>
>> Jere
>> Jere H. Lipps
>> jlipps at berkeley.edu
>>
>>
>>
>> On Wed, Oct 11, 2023 at 6:23 AM John Pandolfi <j.pandolfi at uq.edu.au>
>> wrote:
>>
>>> I’d still take Richard Cowen’s ‘History of Life’ limerick collection
>>> over ChatGPT any day of the week!
>>>
>>>
>>>
>>> *From: *Paleonet <paleonet-bounces+j.pandolfi=uq.edu.au at paleonet.org>
>>> on behalf of David Kopaska-Merkel <davidkm at gsa.state.al.us>
>>> *Date: *Wednesday, 11 October 2023 at 10:12 am
>>> *To: *PaleoNet <paleonet at paleonet.org>
>>> *Subject: *Re: Paleonet: Mucking about with ChatGPT
>>>
>>> This may be an appropriately unserious use of AI, but this limerick
>>> isn’t ready for prime time. It doesn’t follow the required meter. Most
>>> AI-generated text has serious problems like this, thank Ghu!
>>>
>>>
>>>
>>> David
>>>
>>>
>>>
>>> David C. Kopaska-Merkel
>>>
>>> Geological Survey of Alabama
>>>
>>> Box 869999
>>>
>>> Tuscaloosa AL 35486-6999
>>>
>>> 205-247-3695
>>>
>>> dkm at gsa.state.al.us
>>>
>>> he/him/his
>>>
>>>
>>>
>>> *From:* Paleonet <paleonet-bounces+davidkm=gsa.state.al.us at paleonet.org>
>>> *On Behalf Of *Daniel Phelps
>>> *Sent:* Sunday, October 8, 2023 5:41 PM
>>> *To:* PaleoNet <paleonet at paleonet.org>
>>> *Subject:* Re: Paleonet: Mucking about with ChatGPT
>>>
>>>
>>>
>>> I never use ChatGPT for anything serious:
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>>
>>> Sent from my iPad
>>>
>>> > On Oct 8, 2023, at 3:54 PM, David Campbell <pleuronaia at gmail.com>
>>> wrote:
>>> >
>>> > 
>>> > More generally, many online tools make it easy to get
>>> authoritative-looking output from doubtful data.  Checking the actual
>>> information behind a result is essential.
>>> >
>>> >
>>> > --
>>> > Dr. David Campbell
>>> > Associate Professor, Geology
>>> > Department of Natural Sciences
>>> > 110 S Main St, #7270
>>> > Gardner-Webb University
>>> > Boiling Springs NC 28017
>>> > _______________________________________________
>>> > Paleonet mailing list
>>> > Paleonet at paleonet.org
>>> > http://lists.paleonet.org/mailman/listinfo/paleonet
>>> _______________________________________________
>>> Paleonet mailing list
>>> Paleonet at paleonet.org
>>> http://lists.paleonet.org/mailman/listinfo/paleonet
>>>
>> _______________________________________________
>> Paleonet mailing list
>> Paleonet at paleonet.org
>> http://lists.paleonet.org/mailman/listinfo/paleonet
>
>
>
> --
> ___________________________________________
> Luis Azevedo Rodrigues
> Paleontologist (PhD) | Science communicator
> Editor of Geoscience Communication journal
> <https://www.geoscience-communication.net/editorial_board.html>
> Público Newspaper invited blog
> Science Blogs Brazil | Ciencia Ao Natural
> <http://scienceblogs.com.br/cienciaaonatural/>
> Podcast <http://bit.ly/1Jdp0cC>
> independent.academia.edu/LuisARodrigues
> *Skype*: Luis_Azevedo_Rodrigues
> *Twitter*: @CienAoNatural
>
>

-- 
___________________________________________
Luis Azevedo Rodrigues
Paleontologist (PhD) | Science communicator
Editor of Geoscience Communication journal
<https://www.geoscience-communication.net/editorial_board.html>
Público Newspaper invited blog
Science Blogs Brazil | Ciencia Ao Natural
<http://scienceblogs.com.br/cienciaaonatural/>
Podcast <http://bit.ly/1Jdp0cC>
independent.academia.edu/LuisARodrigues
*Skype*: Luis_Azevedo_Rodrigues
*Twitter*: @CienAoNatural
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.paleonet.org/pipermail/paleonet/attachments/20231013/93a23867/attachment.htm>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: laz.rodrigues_a_landscape_of_cretaceous_vegetation_the_fossil_r_87ecf028-6648-4c4b-b582-687b2c2780ef.png
Type: image/png
Size: 2456782 bytes
Desc: not available
URL: <https://lists.paleonet.org/pipermail/paleonet/attachments/20231013/93a23867/attachment.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: laz.rodrigues_a_lush_forest_where_giant_tree_ferns_grow_with_hu_6c7d191f-d3b2-4c37-b168-a9e04f66406c.png
Type: image/png
Size: 2224198 bytes
Desc: not available
URL: <https://lists.paleonet.org/pipermail/paleonet/attachments/20231013/93a23867/attachment-0001.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: laz.rodrigues_Capture_the_majesty_of_the_sauropod_dinosaur_in_m_6031a547-d2d7-45e6-8613-6e6097aeec4c.png
Type: image/png
Size: 2000167 bytes
Desc: not available
URL: <https://lists.paleonet.org/pipermail/paleonet/attachments/20231013/93a23867/attachment-0002.png>