AI and ML

The document provides an update from the Standards Vice President on the work of the Joint ETC Task Force on artificial intelligence (AI) and media, which aims to identify key issues around using AI and machine learning in media such as embedded biases in training data and how to address challenges. It recommends listening to an expert on the dangers of AI bias and provides links to webcasts on the task force's progress and upcoming standards meetings.

Uploaded by

Saqib Kamal

We take content rights seriously. If you suspect this is your content, claim it here.

Available Formats

Download as PDF, TXT or read online on Scribd

0% found this document useful (0 votes)

69 views1 page

AI and ML

Uploaded by

Saqib Kamal

We take content rights seriously. If you suspect this is your content, claim it here.

Available Formats

Download as PDF, TXT or read online on Scribd

You are on page 1/ 1

UPDATE FROM THE STANDARDS VICE PRESIDENT

Bruce Devlin

AI and ML

M PTE and the Enter- important that the training set con- assistants, then I recommend listening

S tainment Technology Center

[(ETC) etcenter.org] have a
long histor y of working
tains images of a variety of animals
and shapes and random objects that
you might take pictures of. This
to the TED Talk (ted.com) from Kriti
Sharma who is an expert in this field.
The aim of the ETC and SMPTE
together. One major success is the seems obvious. Joint Task Force is to identify key
Interoperable Mastering Format Now let us take a different use questions that need to be explored in
(IMF) and another project that we case. Imagine that you are trying to detail to make AI and ML productive
hope will bear similar success is the recruit a new director of marketing. in media. From ethics to Application
Joint ETC Task Force on artificial The head-hunting firm has trained its Planning Interfaces (APIs), all ques-
intelligence (AI) and media. We AI/ML resume filter algorithm based tions will be considered and priori-
know that the use of machine learn- on its 30-year history of resumés tized. For a status report directly from
ing (ML) techniques in the media received and on the people that were
the Task Force, visit the SMPTE web-
space will have an impact. Some placed as directors of marketing. You
site at https://www.smpte.org/webcast/
of these will be beneficial, such as do not have to be a rocket scientist (or
standards-april-task-force-updates
improved metadata collection and a media engineer) to realize that this
where a progress report was given by
inference, and some will have a training data is almost certainly skewed
Yves Bergquist and Fred Walls who
negative effect, such as the ability toward white middle-class males who
leads the group.
to create the so-called deep fakes that have traditionally played that kind of
cannot be easily distinguished from role in the past. This is despite recent Links to our quarterly standards
genuine material. Other issues will studies that have shown that a more meetings and update webcasts, which
be more difficult to address, partic- inclusive and diverse workforce leads feature guest appearances from
ularly the problem of embedded bias to more successful companies. These experts who help influence SMPTE
in training sets. To understand this, invisible biases are a threat to the suc- standards, are also on the website.
it is worth knowing a little about cess of using AI and ML in the media You can attend your first SMPTE
how ML algorithms work. space and this is one of the issues that standards meeting free of cost by fill-
The majority of algorithms try the task force is considering in its work. ing in a guest participation form that
to establish a degree of confidence If you are interested in the dangers of you will find in the governance sec-
between some incoming data (e.g., AI bias, especially on gender issues tion of the website. We look forward
an image) and a training set (e.g., and the voice bias inherent in digital to seeing you (virtually) in 2021.
thousands of existing preclassified
images). When the goal of the algo-
rithm is to distinguish a dog from a 2021 TECHNOLOGY COMMITTEE MEETINGS
horse, it is important that the train-
ing set contains images of both dogs 7–10 June 2021 Virtual
and horses. In fact, it is probably 17–20 August or 14–17 September 2021 Virtual (TBD)