Lifeboat Foundation

DARPA, the innovation arm of the U.S. military, wants artificial intelligence to make battlefield medical decisions, raising red flags from some experts and ethicists.

On top of the environmental concerns, Japan has an added motivation for this push towards automation —its aging population and concurrent low birth rates mean its workforce is rapidly shrinking, and the implications for the country’s economy aren’t good.

Thus it behooves the Japanese to automate as many job functions as they can (and the rest of the world likely won’t be far behind, though they won’t have quite the same impetus). According to the Nippon Foundation, more than half of Japanese ship crew members are over the age of 50.

In partnership with Misui OSK Lines Ltd., the foundation recently completed two tests of autonomous ships. The first was a 313-foot container ship called the Mikage, which sailed 161 nautical miles from Tsuruga Port, north of Kyoto, to Sakai Port near Osaka. Upon reaching its destination port the ship was even able to steer itself into its designated bay, with drones dropping its mooring line.

What most people define as common sense is actually common learning, and much of that is biased.

The biggest short term problem in AI: as mentioned in the video clip, an over-emphasis on data set size, irrelevant of accuracy, representation or accountability.

The biggest long term problem in AI: Instead of trying to replace us we should be seeking to complement us. Merge is not necessary nor advisable.

Continue reading “The biggest problem in AI? Machines have no common sense” »

Which, to me, sounds both unimaginably complex and sublimely simple.

Sort of like, perhaps, like our brains.

Building chips with analogs of biological neurons and dendrites and neural networks like our brains is also key to the massive efficiency gains Rain Neuromorphics is claiming: 1,000 times more efficient than existing digital chips from companies like Nvidia.

Sanctuary, a startup developing human-like robots, has raised tens of millions in capital. But experts are skeptical it can deliver on its promises.

Introducing a novel visual tool for explaining the results of classification algorithms, with examples in R and Python.

Classification algorithms aim to identify to which groups a set of observations belong. A machine learning practitioner typically builds multiple models and selects a final classifier to be one that optimizes a set of accuracy metrics on a held-out test set. Sometimes, practitioners and stakeholders want more from the classification model than just predictions. They may wish to know the reasons behind a classifier’s decisions, especially when it is built for high-stakes applications. For instance, consider a medical setting, where a classifier determines a patient to be at high risk for developing an illness. If medical experts can learn the contributing factors to this prediction, they could use this information to help determine suitable treatments.

Some models, such as single decision trees, are transparent, meaning that they show the mechanism for how they make decisions. More complex models, however, tend to be the opposite — they are often referred to as “black boxes”, as they provide no explanation for how they arrive at their decisions. Unfortunately, opting for transparent models over black boxes does not always solve the explainability problem. The relationship between a set of observations and its labels is often too complex for a simple model to suffice; transparency can come at the cost of accuracy [1].

Continue reading “Explainable AI (XAI) with Class Maps” »

A four-legged robot called Spot has been deployed to wander around the ruins of ancient Pompeii, identifying structural and safety issues while delving underground to inspect tunnels dug by relic thieves.

The dog-like robot is the latest in a series of technologies used as part of a broader project to better manage the archaeological park since 2013, when Unesco threatened to add Pompeii to a list of world heritage sites in peril unless Italian authorities improved its preservation.

This point was made clear in a recent paper by David Silver, Satinder Singh, Doina Precup, and Richard Sutton from DeepMind titled “Reward is Enough.” The authors argue that “maximizing reward is enough to drive behavior that exhibits most if not all attributes of intelligence.” However, reward is not enough. The statement itself is simplistic, vague, circular, and explains little because the assertion is meaningless outside highly structured and controlled environments. Besides, humans do many things for no reward at all, like writing fatuous papers about rewards.

The point is that suppose you or your team talk about how intelligent or cognitively plausible your solution is? I see this kind of solution arguing quite a bit. If so, you are not thinking enough about a specific problem or the people impacted by that problem. Practitioners and business-minded leaders need to know about cognitive plausibility because it reflects the wrong culture. Real-world problem solving solves the problems the world presents to intelligence whose solutions are not ever cognitively plausible. While insiders want their goals to be understood and shared by their solutions, your solution does not need to understand that it is solving a problem, but you do.

If you have a problem to solve that aligns with a business goal and seek an optimal solution to accomplish that goal, then how “cognitively plausible” some solution is, is unimportant. How a problem is solved is always secondary to if a problem is solved, and if you don’t care how, you can solve just about anything. The goal itself and how optimal a solution is for a problem are more important than how the goal is accomplished, if the solution was self-referencing, or what a solution looked like after you didn’t solve the problem.