What is Machine Learning?


Machine learning refers to computers that are able to act and react without being explicitly programmed to do so. Practical speech recognition, semantic applications, and even self-driving cars all leverage machine learning via data systems that not only intake, retrieve, and interpret data, but also learn from it. To do this, the machine must make a generalization, using algorithms to respond to new inputs after being “trained” on a different learning data set — much like a human learns from experiences and uses that knowledge to respond appropriately in a different encounter. In this sense, machine learning is widely considered by many researchers and thought leaders as a step towards human-like artificial intelligence. Recent incarnations of machine learning include a university-developed telescope that can automatically detect significant changes pointing to supernova occurrences. The software Xapagy improvises dialogue and plot moves in stories fed to it by users. The potential of machine learning for education is still some years away, but the potential of learning systems that can adapt and learn on their own is driving research around the world.

INSTRUCTIONS: Enter your responses to the questions below. This is most easily done by moving your cursor to the end of the last item and pressing RETURN to create a new bullet point. Please include URLs whenever you can (full URLs will automatically be turned into hyperlinks; please type them out rather than using the linking tools in the toolbar).

Please "sign" your contributions by marking with the code of 4 tildes (~) in a row so that we can follow up with you if we need additional information or leads to examples- this produces a signature when the page is updated, like this: - Larry Larry Feb 7, 2012

(1) How might this technology be relevant to the educational sector you know best?

  • IBM's Watson has shown that a controlled dictionary/vocabulary in a given content niche (medicine and Jeopardy) can provide intelligent answers to common (and not so common) questions. Many visitors to museums seek out volunteers, guards and other visitors in their quest to get a quick answer to a question about an object, technique, artist, or any other curious thought that pops in their head while in the museum. Machine Learning (and intelligent Virtual Assistants) can provide a great deal of this front-line Q&A. - dallen dallen Feb 28, 2016
  • Because so much of our industry is visually based, this topic should include research on object recognition technologies. It won't be long before software will be able to "guess" and object's categorization and even the time period and artist. Better still for our woefully limited object metadata, it will be wonderful when keywords can be generated (trees, cocker spaniel, piano, sheet music, oranges, etc.) from recognizing items in the image itself. - dallen dallen Feb 29, 2016
  • Museum data is a great repository for utilization in machine learning. Taking metadata from a single institution might yield interesting results, however bringing together many datasets (possibly though LOD) could help make connections and identify new ways of looking at the objects we hold dear. I am not only referring to text data but visual data as well, the images of the works of art can be analyzed and processed to find new relationships and themes that might not be obvious to the human eye. - kjaebker kjaebker Feb 29, 2016
  • IBM's Watson has demonstrate how AI/Machine Learning can be massively disruptive in professions that traditionally depended on mastery (or the appearance of mastery) of vast amounts of information: law, medicine, investment. AI/ML has similar potential to disrupt academic disciplines such as art history, history and science, supplanting some human research tasks, and supplementing higher level analysis. For museums, most significantly, an AI/ML intermediator for museum collections & related data could put the user in charge of their own learning, rather than being dependent on what the curator/educator/museum wants to teach. Museums are experiencing their own mini-explosion of open data, but AI/ML tools have the potential to make this data accessible and meaningful to the non-specialist. - elizabeth.merritt elizabeth.merritt Mar 1, 2016

(2) What themes are missing from the above description that you think are important?

  • The research and solutions in this area are loosely bundled under the broader concept of Artificial Intelligence (AI). The real benefit to the user will be when virtual assistants "front-end" an intelligent program that can parse the natural language interface and provide cogent answers. Perhaps you should consider renaming this area AI and include the idea of Virtual Assistants. - dallen dallen Feb 29, 2016
  • How does this effect current staffing structures of museums. If we can train computers to help in research and content generation, how does that shift the staffing profile in a museum. Does it replace or augment current staffing levels or require new positions to manage the quantity of data that is necessary and could be produced using these systems. - kjaebker kjaebker Feb 29, 2016
  • ML can be applied to human behavior as well--algorithms can respond to individual responses and learning preferences. For this reason, ML can be be a powerful tool driving personalized learning, a la the "Young Lady's Primer" in Neal Stephenson's "Diamond Age."- elizabeth.merritt elizabeth.merritt Mar 1, 2016

(3) What do you see as the potential impact of this technology on education and interpretation in museums?

  • - jasper jasper Feb 19, 2016 I don't know of any real-world examples of machine learning in museums, but I do know is that my students (both in Amsterdam as in Moscow) are bringing machine learning into class as a trend they want to work with in museums ever more often. They are especially interested in the potential to connect collections, and find hidden narratives that are invisible even the the best curator's eye. I'm encouraging them to make this reality! :-)
  • An intelligent art database could incorporate all known scholarly research and metadata surrounding an object. With a natural language interface, it would be easy to ask the program questions such as "Who was her teacher or mentor?"; "Was this work influenced by another artist?"; "Who owned this painting before it was acquired?", etc. By combining general knowledge from history, you could even ask questions such as "Who was the ruler of the country at the time this was created?" or "What composers were alive at this time and what works were they composing?"- dallen dallen Feb 29, 2016
  • This can open up new directions for researchers to take when working with collections. Having computers crunch though mounds of data that would take humans years to process will allow for new thinking processes to develop. We are at the early stages in this technology and the future will only see more enhancements that can lead to new and quicker discoveries. - kjaebker kjaebker Feb 29, 2016
  • See my comments above about the potentially disruptive effects of AI/ML. Museums have traditionally all about authority, acting as gatekeepers to knowledge and learning. Tools like AI/ML can bypass the gate, giving lay users meaningful access to data about and related to museum collections. It may make museum open data sets broadly useful to self-directed learners as well as businesses and researchers - elizabeth.merritt elizabeth.merritt Mar 1, 2016

(4) Do you have or know of a project working in this area?


Please share information about related projects in our Horizon Project Sharing Form.