A Blog by Jonathan Low


Apr 21, 2019

Where Should the Line Be Drawn Between Rejecting or Accepting Black Box AI?

The reality is that even their inventors are not always sure about their capabilities. So as in most other realms of human endeavor, we will probably have to learn to live with some degree of uncertainty. JL

Edd Gent reports in Singularity Hub:

Lack of interpretability in AI is a common concern and many are trying to fix it, but is it really always necessary to know what’s going on inside these “black boxes”? Once in a while we can find information about their inner workings, but most of the time we have to accept their answers and probe around the edges to figure out how reliable and accurate they are. We need to acknowledge that we’re going to need to use black boxes. This is our opportunity to do due diligence to understand how to use them responsibly, ethically, and with benefits rather than harm.

Police Responding To 911 Call About Burglar Find Trapped Roomba Robot Vacuum

'Grime doesn't pay.' JL

Joel Shannon reports in USA Today:

An Oregon sheriff's office entered a residential bathroom with guns drawn, prepared to confront a burglary suspect. Instead, they found a trapped robotic vacuum making a ruckus. Deputies surrounded the house, called in a canine officer and repeatedly instructed the suspect to exit the bathroom. 15 minutes after first responding, police entered the bathroom with guns drawn (and) witnessed a "very thorough vacuuming job being done by a Roomba Robotic Vacuum cleaner."

What Ride Sharing IPOs Teach Us About 21st Century Economics

Lyft's stock has cratered since its IPO, but the early investors got out whole and interest in Uber's remains strong.

Eventually, these companies will raise rates - significantly - be acquired or go out of business. But in the interim, billions that could have been invested in longer term economically sustainable enterprises will have been wasted. JL

Umair Haque reports in Medium:

Uber and Lyft have been vehicles (bad pun fully intended) for a net transfer from venture capital, to consumers, taking VC money   and using to subsidize artificially cheap rides for people. As soon as the subsidy from venture capital runs out — shortly — prices are going to have to rise. Lyft’s stock a classic example of a pump-and-dump. At IPO, it was flogged to “dumb money." The stock soared  and then it crashed  because the smarter money only bet against it. When capitalism tries to solve problems it can’t — like transport for a nation — the result is a net waste and bad businesses

How Hate Groups Banned On Social Media Have Adapted

Hate groups, their followers - and interested onlookers - have become adept at workarounds designed to evade the frequently ambiguous guidelines promulgated by social media.

The larger issue is that social media ad rates continue to be driven by scale, which means they do not want to alienate too many people in order to remain dominant. Until penalties are imposed for fostering hate speech, the platforms have little incentive to change. JL

Natasha Lomas reports in Tech Crunch:

It remains trivially easy for anyone who already knows the ‘brand’ to workaround restrictions by using another mainstream technology. Hate groups are well versed in using coded language and dog whistle tactics to communicate with followers and spread messages under the mainstream radar. YouTube could step up and take a leadership position by setting robust policies against individuals who seek to weaponize hate. Instead it continues trying to fudge the issue, claiming it’s about ‘balancing’ speech and community safety.

The Internet of Thoughts Is Coming

While eliminating middlemen like Amazon, Google or Facebook sounds enticing, the prospect of having telepathic communications hacked seems likely. JL

Andrew Masterson reports in Cosmos:

Before the century is out, advances in nanotechnology, nanomedicine, AI, and computation will result in the development of a "Human Brain/Cloud Interface" (B/CI), that connects neurons and synapses in the brain to vast cloud-computing networks in real time. Instant access to information thus becomes possible without the need for external architecture such as computers and internet cables. Search and retrieval exercises will be initiated by thought patterns. “Neuralnanorobotics” describe the agglomeration of neural, nano and robotic technologies that needs to happen before the internet of thoughts can come to pass.

Why Retailers Are Welcoming Competitors' Online Shopping Returns

Because retailers believe it will drive more foot traffic to stores, which may then cause an uptick in sales.

And brands like it because it may help flatten their soaring return and delivery costs. JL

Sarah Nassauer reports in the Wall Street Journal:

Walgreens Boots Alliance and Nordstrom will let online shoppers at other brands and retailers pick up or return orders at some stores, a sign of how retailers are partnering in new ways to draw customers as more shopping shifts online.“The main thing is this will drive more footfall." (And) Brands hope adding more return and pickup locations will boost sales by offering convenience and reduce the cost of shipping. Sending more items in a single package or letting carriers like FedEx pick up more packages at one location can lower shipping costs.

Baby's First Data

People are marrying later - to the extent they are marrying at all - which means they are having their first children later. After they have been educated, often well educated, and after they have spent time in the working world, replete with access to technology and data. About everything.

So it is not surprising, albeit not always helpful, that they try to apply the data-driven approaches that helped them succeed at school and work, to their child rearing practices.

If only it were that simple. JL

Emily Oster reports in the New York Times:

In 1980, 8.6% of first births were to women over 30; by 2015 this was 31%. This means that by the time a baby arrives, many of us have been through school, spent time in the working world. In many cases data can be helpful (but)  there are some parenting decisions where the data just isn’t much help at all. There are limits to the utility of general information. Parenting is full of decisions, nearly all of which can be agonized over. You can and should learn about the risks and benefits of your parenting choices, but in the end you have to also think about what works for you.