
To use the currently trendy terminology, what weâre teaching students is
meta-learning
âa strong foundation of approaches, ideas, understanding, and tools so that they will be able to quickly learn and evolve over the following decades, as science and engineering changes 2/7


Yes, we at @Stanford do try to ensure we also teach some up-to-date approachesâand we have a great up-to-date course on neural meta-learning too https://cs330.stanford.edu
âbut when AI faculty talk, they mainly wring their hands about how little students learn from before 2010 3/7

I think your sense of a âparadigmâ is offâscientists immediately think of Kuhnâs paradigms: https://en.wikipedia.org/wiki/Paradigm_shift. At most theyâd count modern deep learning as a paradigm shift. Itâs now 15 years from the initial deep learning breakthrough. So seeing some old ideas helps. 4/7
And many might argue that itâs just a development of the machine learning approach to AI that replaced knowledge-based approaches beginning in the late 1980s. We want students to be able to absorb and be productive beyond paradigm shifts, but they donât happen that often. 5/7
Finally, universities do not only educate you, they also provide you with an extensive network of people, who will be of enormous value to you in your future life in many ways, one of which is that they make it easier for you to hear about and learn new things as time passes 6/7
What parts are right? Yes, all technical professionals today need to engage in continuous education to keep their skills up-to-date. The meta-learning will come in useful.
And, yes, experience on @kaggle is a really useful way to build and refresh your skillsârecommended! 7/7
