Connection Problem S03E18: Humans in the Loop

Sitrep: I'm typing this on a laptop which might or might not have a working X key. If you see a lack (or the occasional double) X, don't be surprised. The reasons will become clear as you read on.

×


×

As always, a shout out to tinyletter.com/pbihr or a forward is always appreciated!

×

Learning about machine learning

I've decided to get a little more hands-on with machine learning, so I signed up to Andrew Ng's seminal Stanford machine learning course on Coursera. Let's see how far my bits of rusty algebra will get me; I have a hunch that maybe not super far, but also expect to still learn a lot. With a dozen or so classes it seems doable over a few weeks; I'll report back, hopefully.

×

Personal rant / feel free to skip

Gotta drop a bit of a personal rant about issues with Apple's Macbook Pros and their customer service. Feel free to skip.

Whatever happened to Apple laptops, and has their customer service always be so useless? My Macbook Pro from late 2016 with it's over-engineered "butterfly keyboard" has, within just over a year, had three instances of keys not working properly. I've owned computers since I was about 6 years old, so in over 30 years I've never before had a computer which had a regularly failing keyboard. Yet this is a €2.000+ machine, which puts it pretty much at the top of the so-called "premium" segment. This makes for an interesting conundrum: Because I want the best, most reliable tools I've been shelling out for what's supposed to be the top-of-the-line tool out there—and it's been failing spectacularly in ways so basic it's unbelievable: I'd expect all kinds of failures in high end products, but a keyboard that's not working reliably... that's just sad.
So you'd expect this premium brand to make up for messing up in the first place through exceptional customer service, but oh boy, you'd be wrong. My first trip to the Genius bar for this (even typing the name makes me want to take a shower), they ran software diagnostics and couldn't find anything so determined everything was fine, despite the fact that a key had not worked for several week between me booking the appointment and having it. (Convenient, huh?) Then the Genius (ugh) went on to condescendingly told me one shouldn't eat over the laptop like his girlfriend does (wtf?); so there's a good way to insult a customer and blame a victim. They said they see no problem, thanksbye. The second time they asked me to come in again to run diagnostics; again, within a few weeks I had gotten the keys to work again (for now) using their "compressed air" method that strikes me as something straight out of a DIY garage hack video, not something you want to do to a €2K machine. (If you're curious, check it out here.) Now their phone hotline scheduled a call with me, and at the appointed time slot I received a robo call that asked me to hold—for 10 minutes. They called me and had me wait for 10 minutes.

This whole system is broken in the way that system only break if they're either a) optimized for cost too rigorously or b) optimized to shield against legal actions, or both. The rhetoric is one of "we won't leave you alone with this" (to which I had to reply "well so far you have left me alone"), and one of "we'll make sure this is resolved" (to which I'm thinking "I very much doubt it as no action you've been initiating points towards a solution"). So this is bad, and I'm not sure if this should simply be the last Mac I've bought, or if there's any reason to trust they'll fix things; so far it seems their approach is one of "well look dust motes whatchagonnadoaboutit" just give us more money because we build the best products—when really this sets a new rock bottom base line of shoddy customer service and dysfunctionality.

I can't tell if the hardware is broken, or if it's working perfectly to a broken design; for me as a user it doesn't matter, I have an expensive laptop that doesn't work as advertised. Not sure. Not convinced. Not happy.

End of rant.

On to greener pastures.

×


×

Aspen AI 2018

I was kindly invited to an event by Aspen Institute Germany about the impact of AI on society and humanity. It was great to see a different mix of people discuss these issues than I usually hang out with: There were some technologists but also lots of policy people and some military folk and some artists and researchers. So that's good. I wanted to share a few observations I found noteworthy, especially from a panel about AI in warfare which I found pretty fascinating (it's an area I know practically nothing about otherwise):

(1) Humans in the loop. There was surprising consensus among the panelists that autonomous weapons systems absolutely require a human in the loop. (One researcher made sure he thought that even better would be a global ban on autonomous weapons systems.)

(2) Terror drones / Rogue actors and autonomous weapons. The potential of rogue actors deploying autonomous weapons came up a few times. I asked if this was a thing we had seen happen at any scale, and the answer was a resounding no. While a military person on the panel said it's important to think this through ahead of time (I agree), I find it quite surprising to think that no terrorist group or rogue state actor has started deploying autonomous or semi-autonomous weapons. In my work I watch pioneer communities to see how they innovate, like in the last 10 years the maker movement. It seems surprising to say the least that a group of highly motivated actors wouldn't learn from these communities and adapt their methods to build at least simple weapons. Between maker spaces and AliExpress it would seem easy to rig a cheap (semi)autonomous drone with explosives or whatever. I'm very happy we haven't seen this happen; but I don't understand why that is?

(3) Flash wars. Someone from the Bundeswehr university (I believe his name was Frank Sauer) brought up that algorithms (AI/ML/whichever terminology you prefer in this context) has been demonstrated to cause flash crashes in financial markets. He stipulated that without a human in the loop (see my first point above) we'd be likely to see autonomous systems to cause flash wars, too. His point was that introducing autonomous systems introduces instability. A strong point indeed. Instead, I'd argue, we need to think about increasing resilience.

(4) AI in the military backend. I asked about the role of AI in military outside of autonomous weapons, especially in the backend: operations, logistics, the figurative plumbing that makes the machine run smoothly. The answer (by one Toby Walsh, AI professor at the University of New South Wales and an active voice in the science community calling for a ban of autonomous weapons systems) was that yes, absolutely machine learning is used to optimize military logistics, and to great effect: He mentioned a quote that I'm afraid I was too slow to write down and forgot the details of, along the lines of the gains in efficiency in the (second Gulf?) war through algorithmic optimizations is what has been funding AI research ever since.

So that was pretty good.

×

Bitcoin Is Ridiculous. Blockchain Is Dangerous: Paul Ford's piece on the current state of Blockchain is worth reading:

It’s not that I want Bitcoin holders to suffer, really. As a technologist and entrepreneur, I’m sympathetic to and admiring of risk takers. But as a writer, I enjoy the sheer human-condition-revealing sport. I’m happy to watch other people play video games without playing myself.

×

Privacy isn't dead, but it's going to cost you. Cathleen Berger on privacy as a luxury good. Alas I agree: Same with ads—only the wealthy will be able to minimize ubiquitous advertisement (and now, privacy): "we are starting to realise that disconnecting comes at unbearable social, political, and economic cost. It’s a digital divide where only some have the means and knowledge to secure better privacy."

×

Who shares the spoils of robot work?

Via Boingboing, I stumbled on this quote by Steven Hawking. I never knew particularly much about Hawking, but this makes me appreciate him even more:

If machines produce everything we need, the outcome will depend on how things are distributed. Everyone can enjoy a life of luxurious leisure if the machine-produced wealth is shared, or most people can end up miserably poor if the machine-owners successfully lobby against wealth redistribution. So far, the trend seems to be toward the second option, with technology driving ever-increasing inequality.

Couldn't agree more, with both the aspiration and the depressing analysis. We've seen that in many Western countries GDP has been rising like crazy for decades, yet the wealth pools at the top in an increasingly unhealthy way that makes the system (everybody sing along now!) less resilient and more fragile.

×

The long listen. The Long Now Foundation has a beautiful bar at San Francisco's Fort Mason that you should absolutely visit if you're in SF. More importantly, they have an excellent podcast where they share their lectures, and I only just now found out about it. I've listened to an excellent presentation by archeologist Michael Frachetti about the Silk Road, framed here as an open network of exchange for goods and information, that I couldn't recommend more highly.

×

Yours truly,
Peter

PS. Please feel free to forward this to friends & colleagues, or send them to tinyletter.com/pbihr