You can add location information to your Tweets, such as your city or precise location, from the web and via third-party applications. You always have the option to delete your Tweet location history. Learn more
A more interesting theoretical distinction between "supervised" and "unsupervised" learning could be: 1. Forms of learning driven by an explicit, known objective vs. 2. Forms of learning driven by open-ended discovery and ever-shifting objectives elected by the system itself
All current "unsupervised learning" fits in the first category (for instance, PCA uses data reconstruction as its objective). The second category is basically exclusive to biological systems at this time. It will probably be a while until AI meaningfully encroaches on it.
The second one fits with @jeffclune and @kenneth0stanley open-endedness
I was thinking exactly same thing
(1) could always have sub-objectives elected by system itself, and (2) will arguably always have a super-objective that may be implicit (as in biological systems), so I think it gets confusing and hinges on a vague notion of an "explicit" objective.
This super-objetive could be an intrinsic one, like curiosity is for actions, so you choose the reward function that max this intrinsic in your state, then use this reward to choose an action, change state, and loop. Now it is 100% unsupervised.
It seems like Reinforcement Learning... Isn't it?
In RL an external reward signal is needed, and the question is, in the absence of any external reward signal, how could an agent build his own one? What makes a reward signal good or bad for an agent?
Biological system also has a known objective: to survive. They may change their mini-goal along the way but they clearly have a defined objective. And this clearly falls under supervised learning. In absolute terms, there is no system capable of learning in an unsupervised way.
Since the notion of "supervision" is no longer the explicit labeling of the data but the defined goal of the algorithm, the term "unsupervised" is no longer relevant.
WRT 2 (~∞ game): curiosity gave good results but AI lacks the human bias to find causality, even meaning & agency Problem IMHO: AI is too happy with epicycles & probabilistic models. Science progressed by fighting that, pursuing causality & elegance, a "minimal" set of laws
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.
