sdenton4

The job stability in classical music is great. It's getting a job in the first place that's a problem. My partner, in an earlier life, was one of the best young flautists in Canada. There were about eight jobs in the entire country for orchestral flautists, and about a year before graduating, a few of them turned over, meaning it would probably be another five years before any more jobs opened up. Good time for a career change...

/a smaller set of weights for the same LLM./

Distillation is alive and well... Earlier work on model printing also found that it's pretty easy to find smaller sets of parameters which can replicate the behavior of the entire network with pretty good fidelity.

Large parameter counts give space to explore, and give routes out of what would be local minima in a lower dimensional space.

In other words, there's no guarantee that any given trained model is a minimal representation of its training set.

You're missing the distinction between individual action (however abhorrent) and an actual terrorist organization. Organizations fund, cultivate, and organize attacks, as opposed to lone wolf terrorists, and therefore are much greater threats.

Timothy Mcveigh (and Terry Nichols) weren't members of a terrorist organization, but they were terrorists. Meanwhile, the ku klux klan has actual initiations, membership, and organizational structure. As do Isis and al qaida.

... But also, there's no organization called antifa, and no antifa members. It's an accusation levied against people the right wing wants to call terrorists.

To be sure, it was a couple few hours here and there, not at the scale of a job.

Blue Prince is an example of what fills are calling 'notebook games' -- instead of giving you lots of qol stuff, you just use a notebook to keep notes. The act of writing and arranging your notes is actually quite immersive. It's also fun!

Blue Prince has layers of puzzles and story... It's pretty reasonable to play to the credits, and not get into the depths of the post credits game; it's a very reasonable cutoff point for non-sickos.

Allow me to make a small recommendation for the "best related work": Ada Palmer's 'Inventing the Renaissance.'

It's a fantastic work of complex history, told with the tools of science fiction and fantasy writing. But it remains a high quality academic work: the tools are used to unravel the telling of a gay and complicated topic, but you still get the vastness.

If you like ACOUP, it should be right up your alley.

Which is too bad, because Blue Prince is the better game. :)

I'm playing through Clair Obscur now. It's above average for aaa stuff. But there are way too many cut scenes, and dodge/parry mechanics continue to be tedious. And the story is fine, but isn't really grabbing me. Sigh.

Blue Prince, on the other hand, is quite probably the best puzzle game I have ever played. My partner and I worked on it for months, and have a 50+ page Google doc of notes.

The list is chronological; as it nears the present, the predictions actually refer to the future, and are thus not yet resolved.

There's a growing literature on individual animal identification from sound!

Here's one example: https://github.com/sammlapp/ovenbird-individual-recognition

My impression is that the hardest part is knowing whether or not you're actually doing a good job. For vocal learners (like the ovenbird example linked above), it seems to be pretty easy using embeddings from existing models (like perch or birdnet). For non-learning species (possibly including your swifts), you can use timbre clues to differentiate many individuals, but may not be able to really nail down individual identity.

If Jeff Dean and Sanjay Ghemawat are wrong, I don't want to be right.

This is really the point: We build systems and processes which help ensure that we get reliable outputs (bridges) from unreliable actors (human workers). The point is not that we will one day replace all of these processes with AGI. The point is that we will still want processes and systems in place to ensure reliability.

The importance of safety process is largely a function of the risk of failure (ie, how unreliable the production process is) and the cost of failure. If you've got a very reliable production process, perhaps you need less safety process. But if the cost of failure is measured in lives, you're still going to want to have checks in place, regardless of who/what is responsible for designing the bridge.

Would you trust a bridge built by a single human? In reality, we have lots of guardrails to ensure that we don't screw up and kill a lot of people by deploying defective bridges (or cars to drive on them). Those guardrails often are written in blood, and still occasionally fail.

Let alone a sack of wet, self replicating protein! Just endless copying... How could it ever do anything more?

It was still completely encrusted in crap, however.

Here's a direct link to the video, to avoid the blogspam cruft...

https://www.youtube.com/watch?v=Co57SfcT-h0

Baseball is also non-deterministic, and yet some players are apparently worth a lot more than others.

I stated that boring bounds improvement problems exist, not that all bounds problems are boring... Sigh.

Five (maybe six?) of the results are improvements on bounds. These kinds of problems tend to have some initial advances, and then stall out as the complexity of the bound skyrockets... until some grad student is bored enough to push the boundary. The big-O complexity of matrix multiplication is a good example of how this works: yeah, it's a useful problem, but the solutions are galactic algorithms, and increasingly convoluted.

As someone with a PhD in combinatorics, I believe that I'm qualified to say that, yes, there are problems as useless as calculating more digits of pi.

Some of them...

The two places were seeing lots of movement are:

* Updates to lower/upper bounds. In many cases, these kinds of problems are the deep-math equivalent of calculating more digits of pi. Yes, if you throw time at it you'll break the record, but it may not be terribly worthwhile.

* Finding counter examples which disprove conjectures. This is really useful, and helps offset some positivity bias on the human side, often bringing together known tools from distant silos.

If you read the list of ten results, almost all fall into one of these buckets.

I got an obviously-AI voice agent spam call yesterday. Had a bit of fun getting it to answer trivia questions (when was the treaty of westphalia ratified? answer in a rhyming couplet) as a precondition to handing over the keys to my bank accounts.

I did some of the napkin math on algae a couple years ago... The fun realization is that the physical carbon has to go somewhere. So if we're trying to sequester the average american's 14 tons a year of carbon or whatever it is these days, that means you're going to have at least 14 tons of dead algae to deal with. (And a lot more weight if the algae isn't dehydrated.). That's a lot of dead algae, which demonstrates the difficulty of the problem...

In other news, have you considered ditching your car and becoming a vegetarian? Not emitting the carbon in the first place is much easier!

I doubt yoda speak is actually strict enough, but I do enjoy the idea that Yoda comes from a reverse Polish civilization.

... But are those CEOs ready to admit that those skills are needed?

"succeeding despite Elon musk" is, however, a skill set applicable at only a (strike)small number of companies (/strike) single company.

I dunno, ever used a thing you learned from a textbook in your job? Did you have to continue paying for the copy of that knowledge speed on your brain? No, because that's not what copyright is about.

Learning from and building on previous work is civilization. Copyright maximalism is a plague.

but.... the LLMs are actually all trained on approximately the same stuff, and tend to have similar quirks. In the human world, writers develop recognizable voices, which are detectable and classifiable (as in the article we have all supposedly read). Furthermore, we don't necessarily care about telling one LLM from another, just that they aren't human. That's different from trying to identify one human amongst a sea o fhumans, or one bot form within a sea of bots.

The way the US military works, throwing rolls of toilet paper at the drone from the deck of the shop is likely more expensive than the drone said toilet paper is hurling towards.

Why is that? Companies still need employees, and ai makes it more obvious than ever that workers need to organize together for their rights.

Education research is hard hard hard. Getting clean studies, especially at large sample sizes, is extremely difficult, leading to a lot of ambiguity in results.

That sort of abstraction has always existed, it's just been a matter of hiring experts or labor from other humans. Reality still has a surprising amount of detail. You deal with it by engaging directly, delegating to someone else to engage with, or just brute force your way to a crooked staircase.

When you hire someone to work on the stairs for you, you /hope/ they know what they're doing, especially if you don't have the skills yourself to judge their work. Same for an agent.