Seton Corridor College welcomed alumnus Joshua Meyer ’12 for a workshop titled “AI
101,” introduced by the Instructing, Studying and Know-how Roundtable (TLTR) Synthetic
Intelligence Committee on November 12 within the Bishop Dougherty College Heart. The
session drew college students, school and directors from throughout campus who have been
in understanding how fashionable synthetic intelligence instruments work and the way they are often
used responsibly in educational {and professional} settings.
TLTR AI Committee co-chair Jessica Rauchberg opened this system with an introduction
to the committee’s objectives and the broader mission of TLTR, which is to help the
considerate, moral and efficient use of know-how. Rauchberg then launched Meyer,
an AI researcher and business chief whose work spans language applied sciences, machine
studying and product growth. Meyer holds a Ph.D. in automated speech recognition
from the College of Arizona and a B.A. in liberal research from Seton Corridor.
Meyer started by explaining the foundations of huge language fashions, or LLMs, describing
them as techniques skilled to foretell the following phrase in a sequence moderately than conventional
databases that retailer and retrieve information. He demonstrated how this predictive design
allows LLMs to generate fluent textual content, but additionally makes them liable to errors when trying
duties akin to math or detailed reality recall. Meyer demonstrated stay examples of those
limitations, together with fixing easy arithmetic issues.
Meyer then launched the idea of agentic AI. Agentic techniques can name exterior
instruments or take actions to finish a activity. Utilizing an instance of a mannequin triggering a
calculator program on a pc to compute 646 plus 101, Meyer illustrated how these
brokers lengthen the capabilities of LLMs by connecting them with instruments that may carry out
duties they can not reliably deal with internally. He emphasised that this shift represents
a brand new frontier in AI growth as a result of it hyperlinks reasoning fashions with real-world
features and decision-making.
He additionally addressed widespread points, akin to hallucinations, during which a mannequin produces
assured however incorrect statements. Meyer shared examples that included fabricated
biographical particulars about himself, akin to taking part in soccer at Seton Corridor and in
the NFL, and starring on NBC’s “The Workplace”. He demonstrates how these techniques can
create believable however false info when introduced with a query missing ample
context. He reminded attendees that LLMs don’t perceive reality in a human sense
and rely solely on statistical patterns in language.
Meyer additionally mentioned the rising drawback of what he known as “AI slop,” referring to
content material that seems polished at a look however reveals inaccuracies, imprecise claims,
or clear indicators of machine technology upon nearer inspection. He urged college students not
to rely solely on AI to supply educational work and suggested school to stay vigilant
in regards to the high quality of writing generated by automated instruments. He famous that AI ought to
not change human judgment or creativity and that sustaining “style” and demanding
analysis is important whilst these instruments turn out to be extra superior.
All through the presentation, Meyer emphasised the significance of human management and
intentionality in the usage of AI. Advising members to “be the pilot, not the passenger,”
he inspired customers to take a managerial method by offering clear route, context
and constraints. He defined that LLMs carry out greatest when given particular directions
and detailed context. Their efficiency improves with bigger context home windows, which
allow them to course of extra textual content concurrently.
Meyer concluded by outlining sensible methods during which school, college students, and directors
can make the most of AI to speed up analysis, arrange info, and help artistic
problem-solving. He emphasised that though AI can help with processing massive volumes
of textual content or figuring out patterns, customers should nonetheless confirm outcomes and guarantee accuracy.
The TLTR AI Committee, co-chaired by Rauchberg and Ruchin Kansal, researches the results
of AI on increased training, helps coverage growth, curates greatest practices for
integration and examines implications for the long run office.
For extra info, click here.
Classes:
Science and Technology
