Hits and Misses

Self driving cars for all the bad drivers. Granted you demanded that only 15 months ago and GM still has six months to mass produce those fully autonomous no steering wheel cars, but I’d be interested if you’ve adjusted your take? I’m as negative on the technology and sociology of the things as ever, if not more so.

I’m toying with the idea that maybe Bitcoin will remain afloat nearly indefinitely because there is a niche for a gambling activity shaped like Bitcoin*, and economies of scale and advantages to established products in that niche mean that it’s nearly impossible for the existing dominant player to be dislodged.

*A gambling activity tied into the following themes/aspects: futurism, libertarianism, allowing insiders to feel like they understand something very important outsiders don’t, edginess. As bitcoin has shown, these things have a natural synergy, and appeal to gamblers who wish to think of themselves as something more serious than gamblers.

I hide behind a pseudonym, and in any case, have no professional standing in economics, so have nothing to lose by saying this. So I’ll say it.

‘I’ve always said that, given the irrationality of markets, no one can predict when Bitcoin will reach its true value of zero,’

FWIW (probably nothing) I don’t agree with this, mainly because I have literally have no idea what phrase “the true value of a currency” might possibly mean. If people believe Bitcoin is worth something, it’s worth something: if they don’t, it’s not. QED.

I am perfectly aware of the logical arguments against Bitcoin, and accept them to a greater or lesser extent. I just have an intuition that they are wrong, and that we are all missing something. But I couldn’t say what. But FWIW (again, probably nothing) I think Bitcoin and other crypto-currencies will probably be around for a lot longer than people think: decades, not years.

“My track record is still far better than that of the many pundits who predicted success for the Iraq war and continued claiming imminent victory years after the disaster had become evident.”

Not a high standard to hold yourself to John!

= = = “GM, is releasing a fleet of self-driving Chevy Bolts this year. These Chevy Bolts have no steering wheels, pedals, or speedometers, and they mark a confident step toward driverless cars” = = =

I have a Chevy Volt, the big sister of the Bolt, with all the latest driver assist bells and whistles. It is a great car and the driver assist is already helping me to drive at a higher level of safety as I start down the long slide to Death(tm). But the technology is at least 20, and probably 50, years away from full self driving.

TLDR: Game recognize game, Krauthammer, and you ain’t got none.

John – a couple of responses on the GM car thing.

First of all – why do we believe the GM PR department when they say the technology is on track? Particularly when other company’s (including companies who have a track record here, such as BMW and Volvo) have failed to deliver?

Secondly, regulation here seems to be the insurance industry objecting to an untested technology with no safety statistics being released onto the roads. Which is a completely reasonable objection, and rather begs the question why haven’t GM subjected this technology to independent testing.

Which rather leads me to a good reason to be skeptical of all this technology. None of it has been independently tested. All we have are the heady claims of companies doing R&D in this area. You’re a finance guy – I’m guessing you’re familiar with the phrase “talking one’s book”.

AI is notorious for being an industry where the solution is perpetually 5 years away. Not sure why autonomous cars would be any different honestly.

https://www.eschatonblog.com/2019/05/your-kids.html

You don’t have to be as pessimistic as I am about the current and future state of self-driving car technology to agree with the idea (as almost everybody in the field does!) that what they call level 2 and level 3 technologies – basically self-driving kinda works but you still have to be paying attention 100% just in case – are extremely dangerous. I think it’s worse than has usually been written about, in that the issue isn’t just that drivers are unlikely to maintain 100% attention as they should, it’s also that even a 100% alert driver is never going to be quite sure when s/he is going to have to “fight” the autopilot. At 65 MPH, at what point are you supposed to realize it’s about to kill you or someone else? This is an impossible task, even for that 100% alert driver.

The better question would be “when can i get a robot taxi”. Which is alright too. The first level 5 cars will not be a software updates for cheap assistance level hardware.

@`4

‘I now predict that, as this seeps into the blockheaded crania of the American booboisie, the Senate GOP has already realized that their best chance of keeping the WH in 2020 is by dumping Trump, and going with Pence, even if possibly knocking Pence off in the primaries’.

My counterprediction is that this will not happen.

@Feliquid – do you have any links that explain how this platform is going to solve any of these problems?

I would have thought that Professor Quiggin’s proudest prediction would have been his prediction that the ‘golden age’ of (manned) space travel is over, and that, indeed, the Golden Age of all forms of (carbon-fuelled) transport is over.

Did anyone watch the news about the A380? The largest passenger plane ever built, and, of course the overwhelming probability is that it is the largest (oil fuelled) passenger plane that will ever be built. Mothballed decades ‘before its time’ and another sign that the ‘age of flight’ is winding down. (History will record the end of Concorde and the end of the A380 as the first signs that the flying party was reaching its inevitable conclusion, just as the end of the Space Shuttle indicated that the Space Age was coming to an end).

It’d be real nice if the age of flight were winding down, but passenger miles flown have been up 5 to 8% every year for the last 10 years (the 2007/2008 recession took a bite out of said growth for two years, but it was back up immediately after), China is building over 100 new airports. The insanity remains in full force.

FWIW, as an ex-AI type (all-but-thesis, 1984), the whole self-driving car thing looks to me to be more hype than content. “Neural net” image recognition is quite fragile (easily fooled* e.g. can’t recognize cows on a beach since there’s no green), “neural nets” have no similarities to neurons whatsoever (really, they don’t**). There are probably some technological kludges that can be implemented to prevent some classes of accident (cell phone monitor that forces the driver to pull over to make a call), and these will be good things that save lives, but they won’t be AI, they’ll be ad-hoc kludges.

*: Gary Marcus discusses this. More generally, kids at a young age, long before they learn language, learn the concept of “physical object”. Watch a parent playing “peek-a-boo” with their kid. Kids figure out that physical objects actually exist real early. Current AI systems don’t/can’t do that (they can’t recognize objects*** rotated more than a few degrees, e.g. can’t recognize an upside down school bus). We AI types failed miserably at figuring out how to make the computer do anything that even vaguely looked like “reasoning”, and retreated to ad hoc kludges. There’s no “I” in AI. (My basic view of AI is that human intelligence is seriously amazing, we aren’t anywhere close, and we’re not even trying.)
**: This was covered in an article in Science recently. It’s not a secret. Real neurons in real mammalian brains are extensive in space (long projections), have large numbers of inputs (7,000 on average), and compute logic functions of subsets of their inputs. “Neural nets” are layered, locally connected (4 to 10 inputs) arrays of computing elements, and deal with weights and averages, not logic. Completely different beasts.

***: This is subtle: neural nets have no way of representing or talking about “objects” at all. They are pure pattern recognition systems, with nothing higher level than that. It sometimes looks as though they are “recognizing objects”, but they’re not.

That’s a funny way of looking at the A380, Hidari. The A380 failed because smaller cheaper planes achieved just as much range, plus Boeing got their tame regulator to make a change that favours those smaller planes (ETOPS – the distance a twin engined plane is allowed to be away from the nearest airport in case an engine fails – is now up to 7 hours, where once it was deemed dangerous for them to be more than 40 minutes away).

The result is that there are a lot more point-to-point flights to relatively minor airports, at the direct expense of the big planes doing hub-and-spoke. Which has meant far MORE planes doing long hauls and burning, net, more fuel in total. For better or worse it doesn’t look like the end of the flying age at all.