On Dec 18, 2024, at 3:40 PM, Gary Richmond <[email protected]>
wrote:
Frederik, Mike, Tuezuen, Daniel, List,
For the last couple of years I have been dabbling in various AI
programs including those which generate visual images including
diagrams. I have mainly used ChatGPT for any number of purposes but
principally for information gathering (it hasn't completely replaced
search engines and Wikipedia, but I use it frequently when I'm
looking for specific information and not, say, the kind of overview
which Wikipedia offers.
Frederik, your observation that LLMs frequently 'hallucinate' since
"abduction is neither necessary (like deduction) nor probable (like
induction)." On the other hand, abduction viewed as retroduction
(reasoning from effect to cause) seems nearly 'tailor made' for AI
which can search those myriad data bases in search of connections
which might prompt plausible hypotheses. Does that seem correct?
In my first post in this thread I noted how AI has proven useful in
generating hypotheses in the medical field (I've probably read more
about this field of AI hypothesis generation than any other because
of some tech friendly, AI enthusiastic physicians I know in the NYU
Langone health system.
I'm still quite interested in List members' thoughts about the 3
questions I concluded my original post with, namely: 1. How
potentially valuable do you think AI will be in various disciplines,
especially those fields in which one has some expertise? 2. What are
the potential dangers of AI? To which I'd add: Can they be
circumvented? How?. 3. Is the role of the scientist (the creative
'hypothesizer') jeopardized? I
The rest of this post is a ChatGPT outline of some of the fields in
which AI has successfully generated hypotheses (including examples)
and, in some cases, even tested these hypotheses.
*****
*****
AI has shown considerable success in generating hypotheses across
a variety of fields, disciplines, and sciences. Here are some of
the most notable areas:
*1. Healthcare and Medicine*
o *Drug Discovery:* AI has been instrumental in identifying new
drug candidates and repurposing existing drugs. Examples
include identifying potential treatments for diseases like
COVID-19 and rare genetic disorders.
o *Diagnostics:* AI models have proposed novel diagnostic
criteria, such as identifying biomarkers for diseases like
cancer or Alzheimer's from genomic or imaging data.
o *Genomics:* AI is used to hypothesize relationships between
genes and diseases, and predict the functional impact of
genetic mutations.
*2. Biology and Biotechnology*
o *Protein Folding:* AI systems like AlphaFold have solved
hypotheses about protein structure, paving the way for
advancements in molecular biology.
o *Ecology:* AI has helped hypothesize the effects of climate
change on ecosystems and species interactions.
o *Synthetic Biology:* AI-generated hypotheses guide the design
of engineered organisms for biofuels, medicine, or agriculture.
*3. Physics and Astronomy*
o *Astrophysics:* AI has helped hypothesize about the
distribution of dark matter, the structure of the universe,
and the identification of exoplanets.
o *Quantum Physics:* AI models have been used to generate
hypotheses about material properties and quantum states.
o *Particle Physics:* AI helps analyze data from particle
accelerators to hypothesize about fundamental particles.
*4. Chemistry and Materials Science*
o *Material Design:* AI hypothesizes the properties of novel
materials for use in batteries, solar cells, or superconductors.
o *Reaction Mechanisms:* AI can propose mechanisms for complex
chemical reactions, speeding up the discovery of catalysts.
*5. Social Sciences*
o *Behavioral Patterns:* AI generates hypotheses about human
behavior by analyzing large datasets, such as social media
interactions or economic activity.
o *Policy Impact:* AI models help hypothesize the effects of
policy changes on societal outcomes like education, public
health, or economic growth.
*6. Environmental Science*
o *Climate Modeling:* AI hypothesizes the potential effects of
greenhouse gases, deforestation, and other factors on climate
change.
o *Sustainability:* AI generates hypotheses about renewable
energy efficiency, waste management, and conservation efforts.
*7. Economics and Finance*
o *Market Predictions:* AI hypothesizes about market trends and
economic conditions by analyzing large-scale financial data.
o *Economic Modeling:* AI helps explore hypotheses about income
inequality, employment trends, and consumer behavior.
*8. Engineering and Technology*
o *Robotics:* AI hypothesizes how to optimize robot design and
functionality in various environments.
o *Optimization Problems:* In fields like logistics, AI
hypothesizes ways to improve efficiency and reduce costs.
*9. Neuroscience and Cognitive Science*
o *Brain Function:* AI helps generate hypotheses about how
neural networks in the brain relate to behavior and cognition.
o *Mental Health:* AI models propose new treatments for mental
illnesses based on patterns in psychological and neurological
data.
*10. Education*
o *Personalized Learning:* AI generates hypotheses about which
teaching methods or materials work best for individual
learning styles.
o *Curriculum Design:* AI analyzes data to hypothesize about
effective curriculum structures.
*11. Agriculture*
o *Crop Yields:* AI hypothesizes how weather patterns, soil
types, and farming techniques affect yields.
o *Pest Control:* AI models propose sustainable methods for
pest management.
*12. Linguistics and Natural Language Processing*
o *Language Evolution:* AI helps hypothesize about how
languages evolve over time.
o *Semantic Analysis:* AI generates hypotheses about the
relationships between linguistic structures and meanings.
By leveraging vast amounts of data, AI not only generates
hypotheses but also tests them through simulations or by guiding
experimental design. Its versatility and data-driven approach
make it an invaluable tool in advancing knowledge across disciplines.
Best,
Gary R
On Wed, Dec 18, 2024 at 5:02 AM Frederik Stjernfelt
<[email protected]> wrote:
Dear Mike, Gary, Tuezuen, list –
This is a great idea. This would also explain why LLMs
“hallucinate” so much as they do, as abduction is neither
necessary (like deduction) nor probable (like induction). Peirce,
of course, stresses that abduction is indeed the source of new
ideas but that it offers no assurance of their truth which has to
be established by ensuing investigation using de- and inductions.
I have only experimented with the free versions of ChatGPT and
they are, indeed, highly error-prone.
I tend to prefer the program Perplexity which is connected to a
search engine which it utilizes to provide references to where it
scraped its information.
Best
Frederik
Frederik Stjernfelt: /Sheets, Diagrams, and Realism in Peirce/ –
De Gruyter 2022
* “Peirce as a Philosopher of AI”, in Olteanu et al.:
/Philosophy of AI/, forthcoming
*Fra: *<[email protected]> på vegne af Tuezuen
Alican <[email protected]>
*Svar til: *Tuezuen Alican <[email protected]>
*Dato: *onsdag den 18. december 2024 kl. 08.43
*Til: *Mike Bergman <[email protected]>, Gary Richmond
<[email protected]>, Peirce-L <[email protected]>
*Emne: *RE: [PEIRCE-L] AI and abduction
Dear Mike and Gary,
If I’m not mistaken, John Sowa already utilizes LLMs this way. He
argues that LLMs are great for abductive conclusions, and later,
with an Ontology, he checks whether that “hypothesis” is true or
not. At least, that’s my interpretation of his work.
@Mike Bergman <mailto:[email protected]>, sorry for the
duplication; I pressed reply instead of replying to everyone.
Best Regards,
*Dipl.-Ing. Alican Tüzün, BSc*
PhD Candidate
*University of Applied Sciences Upper Austria*
*Josef Ressel Centre for Data-Driven Business Model Innovation*
Wehrgrabengasse 1-3
4400 Steyr/Austria
LinkedIn: https://www.linkedin.com/in/t%C3%BCz%C3%BCnalican/
<https://www.linkedin.com/in/t%C3%BCz%C3%BCnalican/>
Phone: +43 5 0804 33813
Mobil: +43 681 20775431
E-Mail: [email protected]
<mailto:[email protected]>
Web:www.fh-ooe.at <http://www.fh-ooe.at/imm>
Web: https://coe-sp.fh-ooe.at/ <https://coe-sp.fh-ooe.at/>
*From:*[email protected]
<[email protected]> *On Behalf Of *Mike Bergman
*Sent:* Wednesday, 18 December 2024 02:52
*To:* Gary Richmond <[email protected]>; Peirce-L
<[email protected]>
*Subject:* Re: [PEIRCE-L] AI and abduction
You don't often get email from [email protected]. Learn why this
is important <https://aka.ms/LearnAboutSenderIdentification>
Hi Gary,
This is a topic near and dear to me, and one I am very actively
investigating (and using) personally (mostly with ChatGPT 4-o1,
but also the latest version of Grok). My first observation,
granted based on my sample of one, is that abductive reasoning in
a Peircean sense is lacking with current LLMs (large language
models), as is true for all general ML or AI approaches. Machine
learning and deep learning have been mostly an inductive process
IMO. A major gap I have seen for quite some time has been the
lack of abductive reasoning in most ML and AI activities of
recent vintage.
This assertion is most evident in the lack of "new" hypothesis
generation by these systems, the critical discriminator that you
correctly point out from Peirce. One can prompt these new chat
AIs with new hypotheses, and in that form, they are very helpful
and useful. It is for these reasons that I tend to treat current
chat AIs as dedicated research assistants: able to provide very
useful background legwork, including some answers that stimulate
further questions and thoughts, often in a rapid fire
give-and-take manner, but ones that are not creative in and of
themselves aside from making some non-evident connections.
I believe that better matching of current chat AIs with Peirce's
thinking (esp abductive reasoning as he defined) is a
particularly rich vein for next generation stuff. Lastly, my own
personal view is that the current state of the art is not
"dangerous", but we are also seeing very rapid increases of what
Ilya Sutskever <https://en.wikipedia.org/wiki/Ilya_Sutskever>
calls "superintelligence", the speed of which is pretty
breathtaking. We may be close to tapping out on this current
phase with most Internet content already captured for training,
but like with LLMs, there are certainly new innovations not yet
foreseen that may continue to maintain this Moore's law
<https://en.wikipedia.org/wiki/Moore%27s_law>-like pace of
improvements.
Best, Mike
On 12/17/2024 6:00 PM, Gary Richmond wrote:
List,
In a brief article, "How Does A.I. Think? Here’s One Theory"
in the New York Times today, Peter Coy, after noting that
"Computer scientists are continually surprised by the
creativity displayed by new generations of A.I.," comments
on one hypothesis that might help explain that 'creativity',
namely, that AI is using abduction in its machine reasoning.
He writes:
One hypothesis for how large language models such as o1
think is that they use what logicians call abduction, or
abductive reasoning. Deduction is reasoning from general
laws to specific conclusions. Induction is the opposite,
reasoning from the specific to the general.
Abduction isn’t as well known, but it’s common in daily
life, not to mention possibly inside A.I. It’s inferring
the most likely explanation for a given observation.
Unlike deduction, which is a straightforward procedure,
and induction, which can be purely statistical, abduction
requires creativity.
The planet Neptune was discovered through abductive
reasoning, when two astronomers independently
hypothesized that its existence was the most likely
explanation for perturbations in the orbit of its inner
neighbor, Uranus. Abduction is also the thought process
jurors often use when they decide if a defendant is
guilty beyond a reasonable doubt.
Yet Peirce argues in the 1903 Lectures on Pragmatism that
only abduction "introduces any new idea" into a scientific
inquiry:
" Abduction is the process of forming an explanatory
hypothesis. It is the only logical operation which
introduces any new idea; for induction does nothing but
determine a value, and deduction merely evolves the
necessary consequences of a pure hypothesis."
I had always thought of abduction as the unique domain of the
individual scientist, the creative genius (say, Newton or
Einstein) who, fully versed in the most important relevant
findings in his field, retroductively connects those pieces
of scientific information to posit a testable hypothesis
concerning an unresolved question in science.
But it makes sense that an AI program employing large data
bases might indeed be able to 'scan' those huge,
multitudinous bases, connect the salient information, and
posit an hypothesis (or some other abductive idea).
Any thoughts on this? For example: Is it potentially a
valuable feature and power of AI and, thus, for us (the use
of AI in medical research would tend to support this view)?
Is it a potential danger to us (some AI programs have been
seen to lie, to 'hide' some findings, etc.; might this
get out of control)? If AI can create testable hypotheses, is
the role of the 'creative' scientist jeopardized?
Best,'
Gary R
_ _ _ _ _ _ _ _ _ _
ARISBE: THE PEIRCE GATEWAY is now at
https://cspeirce.com and, just as well, at
https://www.cspeirce.com . It'll take a while to repair /
update all the links!
► PEIRCE-L subscribers: Click on "Reply List" or "Reply All"
to REPLY ON PEIRCE-L to this message. PEIRCE-L posts should
go to [email protected] .
► To UNSUBSCRIBE, send a message NOT to PEIRCE-L but to
[email protected] with UNSUBSCRIBE PEIRCE-L in the SUBJECT
LINE of the message and nothing in the body. More at
https://list.iupui.edu/sympa/help/user-signoff.html .
► PEIRCE-L is owned by THE PEIRCE GROUP; moderated by Gary
Richmond; and co-managed by him and Ben Udell.
--
__________________________________________
Michael K. Bergman
319.621.5225
http://mkbergman.com
http://www.linkedin.com/in/mkbergman
__________________________________________
_ _ _ _ _ _ _ _ _ _
ARISBE: THE PEIRCE GATEWAY is now at
https://cspeirce.com and, just as well, at
https://www.cspeirce.com . It'll take a while to repair / update all
the links!
► PEIRCE-L subscribers: Click on "Reply List" or "Reply All" to REPLY
ON PEIRCE-L to this message. PEIRCE-L posts should go to
[email protected] .
► To UNSUBSCRIBE, send a message NOT to PEIRCE-L but to
[email protected] with UNSUBSCRIBE PEIRCE-L in the SUBJECT LINE of
the message and nothing in the body. More at
https://list.iupui.edu/sympa/help/user-signoff.html .
► PEIRCE-L is owned by THE PEIRCE GROUP; moderated by Gary Richmond;
and co-managed by him and Ben Udell.
