AI researchers have typically tried to get around the issues posed by by Montezuma’s Revenge and Pitfall! by instructing reinforcement-learning algorithms to explore randomly at times, while adding rewards for exploration—what’s known as “intrinsic motivation.”
But the Uber researchers believe this fails to capture an important aspect of human curiosity. “We hypothesize that a major weakness of current intrinsic motivation algorithms is detachment,” they write. “Wherein the algorithms forget about promising areas they have visited, meaning they do not return to them to see if they lead to new states.”
Group items tagged
When Splitters become Lumpers: Pitfalls of a Long History of Human Rights « L... - 0 views
-
"For a close reader of Moyn's work on human rights the differences between his two works are head-spinning. Where Last Utopia attacked the very idea of historic continuity in explaining the human rights movement that emerged in the 1970s, Not Enough builds an entire narrative on continuities. The result is an aspirational history for a reformed human rights movement, a history of roads not taken - with respect to equality, in particular, which Moyn elevates to the 'original' position - that can still be reclaimed. Not Enough lacks the skepticism that Moyn employed so effectively in The Last Utopia to explain how disconnected contemporary human rights was from its claimed antecedents and undermines arguments in both books. In addition, by not heeding his own lessons from Last Utopia, Moyn understates the emergent human rights movement's inability to contest what became neoliberalism. As someone who confronted those issues at the time, it is harder to dismiss the claims of complicity."
Uber has cracked two classic '80s video games by giving an AI algorithm a new type of m... - 0 views
-
-
The team’s new family of reinforcement-learning algorithms, dubbed Go-Explore, remember where they have been before, and will return to a particular area or task later on to see if it might help provide better overall results. The researchers also found that adding a little bit of domain knowledge, by having human players highlight interesting or important areas, sped up the algorithms’ learning and progress by a remarkable amount. This is significant because there may be many real-world situations where you would want an algorithm and a person to work together to solve a hard task.
1 - 5 of 5
Showing 20▼ items per page