Bill de hÓra

@finners_99 @the_transit_guy de Gaulle of you to say that

@the_transit_guy Dublin

What's a city or metro that has the worst airport relative to its economic and cultural might?

Exactly: x.com/dan_goosewin/s…

At LinkedIn, we run stateful systems on bare metal servers as a first class citizen on Kubernetes. That's why we built our own stateful workload operator. Join our #KubeCon talk to learn why StatefulSet didn't cut it for us and how we work with local storage.

Strandbeests are mesmerizing: strandbeest.com/evolution

Using an enum instead of a boolean, when applicable, isn't over-engineering. It's anti-under-engineering.

The only way to extend a boolean is to add another one.

.@ftrain's 'How to Be Polite' is just over ten years old and aging beautifully in its importance. While it can be hard even unrewarding it's worth saying—being polite, being kind, online, is both an increasingly differentiating and increasingly moral act. ftrain.medium.com/how-to-be-poli…

Fun to finally catch up with President @EmmanuelMacron IRL on his ambitious plans to 10x AI infrastructure in France and make the EU more friendly for open source AI. He’s cooking

Failing vintage. The FAA in the context of NOTAM coined what might be the perfect phrase for systems, that operatively don't get, and then effectively can't be, uplifted or replaced and which increasingly gate an organization's future potential. p14 of transportation.gov/sites/dot.gov/…

I can't think of another branch of computation receiving such resistance on the long run [1] while making progress as NNs. If this approach does not solve, it's ideally explained, at some point, why after 81 years it just won't go away. [1] Ok, maybe Relational Algebra/SQL :)

And so by way of receipts, here are the references for the first paper on neural networks, 81 years ago.

Wish more commentators and professionals would understand and clarify that neural networks fall out of the same underpinnings that gave you: types, lambda calculus, logic programming, von neumann architecture, object orientation. NNs aren't some weird other, they're complimentary

@bradneuberg And in the spirit of alternate histories, it's interesting to wonder what might have happened if the Fifth Generation Computer Systems program invested into neural networks and not just logic programming.

@bradneuberg Solid individual contribution, if not quite single handed. Got a Turing not long after Perceptrons. But progress was being made on nonlinear problems by the late 60s, backprop was emerging by the early 70s. So yeah, lost a decade or so until the PDP group really went after it.

Was Marvin Minsky single handedly responsible for slowing down the growth of neural networks by perhaps 20 years? Even though Minsky actually studied and wrote on neural nets he ultimately was one of the major evangelists for symbolic approaches, which mired the field of AI in a dead end for at least 20 years. Imagine if instead we had continued pushing on the connectionist and neural network approaches from 1968 onwards? Would we have gotten specialized hardware earlier that might have helped it (just like the symbolic AI crowd got the Symbolics machines in the 1980s for their beloved LISP?) Alternate history!

Went to manning.com to pick up a book, and saw it has subscription options. The pro level (read everything / download one thing / buy anything half price) seems bargainsome at $24.99. Any reasons to not sub?

We are as far away from The Lord of the Rings (2001) as it was from The Lord of the Rings (1978)

From the rooftops. A little, a lot, goes a long way.

@definev2 In my experience, more smaller changes are more stable than less big releases.

The concern around machine learning and artificial intelligence can be real, but law like this won't help. I'd rather see legislation on utilization and/or bad actors than technology. In the sense we care about law around music rather than law around musical instruments.

SB1047 is a poster child of legislation-as-innovation, designed on a credulous, inexpert, vision of the future rather than actual concerns. There is no credible notion of what to regulate for or against. It's also, astonishingly bad for US interests. It should be walked back.

Lovely late night track to leave you with. Live at the Shacklewell Arms tomorrow, Saturday 6pm open.spotify.com/track/3aRXCzXY…