AI and Machine Learning
Introduction to AI in DAMs
Every year the Digital Asset Management community holds a DAM conference in Los Angeles, New York and London. The frequency of the conference keeps attendees up-to-date on the latest trends and innovations. In March of 2023, the Los Angeles DAM included many hot topics, such as building a DAM and working with AI. Many of the presentations were given by digital asset managers with decades of experience at their institutions, learning to navigate fairly new digital asset management systems. Shared concerns and challenges included choosing platforms, working with vendors, managing massive amounts of assets, migrating assets from one system to another, and learning to use AI to manage digital assets effectively.
Among some DAM experiences, one digital asset manager shared his experience working with AI. He talked about a video of the Amazon covered in trees, that was consistently misidentified by an automated system. The auto-generated tags identified the top of the trees as “broccoli.” So every three seconds the word “broccoli” appears in the descriptive language. The AI recognized the trees as “broccoli” because of how it was programmed. The incorrect AI required a team member to go in and manually correct each “broccoli” reference, according to the speaker. Although this was a very funny story and a good example of how AI can go terribly wrong, it sounds a bit like an urban myth. Logically, if you have ever used the Word function to replace one word with another in a document, wouldn’t it make sense that you could do the same within the context of a specific video, pertaining to it’s auto-generated tags? Learning to work with AI will soon be as common as learning how to use Word or Excel at work, notes Tenovos.
Development of AI
Michael Waldron, CMO of Tenovos ask how digital asset managers are currently using AI at work. He used a poll to survey the audience of DAM professionals from across the world on how they are using AI: Some said their DAM does not have AI capabilities, and some have the capabilities but they are not using them, while others use AI in limited capacities, such as to label content and deliver insights. Still, a few express concerns that AI presents some serious risks and potential dangers. According to this brief survey of the audience at this presentation, over 50 percent of the represented Digital Asset Management Systems do not yet have AI capabilities. Less than 10 percent expressed concerns over using AI, and potential negative ramifications of implementation. The presentation attempts to explain both what AI is and what AI is not. Artificial Intelligence can be misunderstood. Over the past decade, AI has meant that humans have programmed machines and systems to provide automated responses, when prompted to give an answer to a question. This lighter version of Artificial Intelligence depends on hard coding by programmers. In other words, the machines merely answer what they are programmed to respond. They are not making any great leaps and bounds in processing communications and providing human-like responses, other than the programmers who programmed them are humans and write code to simulate human interactions. These easier versions of artificial intelligence were dependent on the information provided by the programmer. The AI was not based on any reasoning or logical deduction made by the machine or system to come up with the correct response. The answers were not provided based on deductive reasoning, in a manner similar to how a human responds.
Evolution of AI
In this recent presentation on the evolution of artificial intelligence, Tenovos defines true AI as “The simulation of human intelligence in machines that are programmed to perform tasks that typically require human intelligence, such as visual perception, speech recognition, decision making, and language translation,” (Tenovos, 2023, April 25.). In the last couple of years, true AI has evolved to include these four capabilities: decision making, language translation, speech recognition and visual perception. With these capabilities, AI helps reduce repetitive tasks that require a lot of time and manpower. AI can be used to automatically tag images and videos, to recognize backgrounds, faces, objects and voices, and to analyze user preferences and enhance content search ability and user adoption. These are features and functions that true AI should be capable of in order to help digital asset managers streamline workloads in their DAMS.
Advances in AI and Machine Learning
Part of artificial intelligence is machine learning, which eventually learns human-like behaviors beyond how its programmed. As MIT Sloan professor Thomas Malone explains, “In just the last five or 10 years, machine learning has become a critical way, arguably the most important way, most parts of AI are done,” (Brown, 2021). Computer programmers agree that the ultimate goal of true AI is to be able to learn human-like decision making without needing a human to program those responses. These learned skills allow machines to respond to their physical environment as a human might. The three main functions of true AI include: Descriptive, predictive, and prescriptive behaviors. (Brown, 2021).
Analysis of AI
According to Tenovos, key skills required to leverage AI include critical thinking and data analysis. As a digital asset manager, you must “consider the ethical implications of using AI in your DAM; evaluate and analyze AI-generated recommendations and insights; identify potential biases in data; and, make informed decisions on how to optimize workflow and performance based on AI insights,” (Tenovos, 2023, April 25.) Some of those biases include who is inputting the data and what natural biases impact the type of data being programmed into the machine or system. Biases can be built on the programmer’s background, including race, gender, sexual identity, culture and community. This is one reason why advancements in AI benefit from advancements in diversity, accessibility, inclusivity and equity in the Library and Information Sciences profession.
Tenovos reminds users that AI is a human construct, so biases in the data input can impact the data output. The digital asset manager is responsible for analyzing that output from the AI, and verify that it does not violate any company policies, such as presenting racial biases, harmful language or inaccurate representations. Once programmed properly, AI can work to implement legal perimeters, check for use rights and red-flag any inappropriate or unacceptable content, in text, images and videos, throughout the DAM. AI can approve content based on brand guidelines or identify rights violations, notes Tenovos. AI can be used through out the workflow of an asset, assigning it to teams, authorizing access, auto-tagging metadata, updating metadata language, identifying faces, objects and voices, helping with accessibility issues, verifying rights and distributing assets to partners, for example.
Training Your AI
Waldron encourages you to talk to your AI on a regular basis. Sign up for a CHAT GPT account. Manage your AI like you would manage someone on your team. How you extract the greatest value from this team member, depends on how well you monitor the results that you are getting from your AI. Tenovos recommends playing with your AI on a daily basis, outside of your DAM. Vendors in DAMS will not always be caught up on the latests technologies in AI. So it will be up to you to talk with your AI and learn how to communicate with it, because you will need to know how it responds and what it is capable of in its responses. Recommended reading, AI, Machine Learning, DAMS, and the Future of Data.
Brown, Sara (2021, April 21). Machine learning explained. MIT.
https://mitsloan.mit.edu/ideas-made-to-matter/machine-learning-explained
Tenovos (2023, April 25). The impact of AI on digital asset management: Key skills for dam managers to stay competitive
No comments:
Post a Comment