Music/audio signal processing
CCRMA
Joined April 2009
- Tweets187
- Following579
- Followers456
- Likes192
Julius Smith retweeted
GPT-6 Astra has the best result yet on the Bach Benchmark. Its chorale contains no voice-leading errors, and its harmonic palette is sophisticated enough to include a Neapolitan sixth chord. More importantly, it is the first model to ever write passing tones on this benchmark, a significant leap in musical understanding.
Extra High effort. Prompt below:
“In LilyPond (version 2.24), write a 4-part chorale in the style of Bach, 3/4 time, G minor. Use two staves -- soprano and alto on the top staff, tenor and bass on the bottom staff. Respond with only the code block.”
Julius Smith retweeted
MIT published a brutally honest report on what AI is doing to students.
A committee of professors and students spent five months studying how AI changed learning on campus, and the findings read like a warning to every university on the planet.
Study groups are disappearing. Office hours are emptying out. Problem sets and take-home exams no longer prove anything, because AI can produce credible solutions to almost any written assignment in the undergraduate curriculum. Students who lean on chatbots lose mastery and confidence, and some slip into what the report calls cognitive surrender, reaching for AI at the first hint of struggle.
The numbers are rough. 46 percent of surveyed MIT undergrads use LLMs daily. 90 percent worry about their own overreliance. Undergrads who feel AI makes them replaceable now outnumber those who feel it makes them capable.
The committee's answer surprised me. They refused to fight AI with surveillance. The report calls AI detectors unreliable, says lockdown browsers feel like spying, and warns that policing students builds a classroom atmosphere of mutual distrust.
Instead, MIT wants to rebuild education around the things AI can't replace. That means oral exams, semester portfolios, in-person project work, and a required social component in every subject. The report even floats the idea of rethinking grades entirely, since without a GPA to optimize, much of the incentive to cheat with AI evaporates.
The committee warns professors against replacing undergrad research assistants with AI agents just because they're cheaper, because a university exists to grow people, not output.
The most famous tech school on earth admitted the machines broke its way of teaching. Its answer is more humans, not more software.
Julius Smith retweeted
Claude Opus 4.8 is out today. It's our strongest coding model yet: up on SWE-bench Pro (from 64.3 to 69.2) and noticeably more honest about its own work. It tells you when it's unsure and catches its own bugs instead of declaring victory early. Same price as 4.7.
Introducing Claude Opus 4.8: it builds on Opus 4.7 with sharper judgment, more honesty about its own progress, and the ability to work independently for longer than its predecessors.
Available today at the same price.
ALT Benchmark table showing how Claude Opus 4.8 compares to its predecessor and to other models on tests of coding, agentic skills, reasoning, and practical knowledge work tasks.
Julius Smith retweeted
There are generally 2 types of LLM users, those that use it to learn everything , and those that use it so they don’t have to learn anything.
Julius Smith retweeted
Marc Andreessen explains why AI coding won't replace programmers, but fundamentally change what they do.
He argues that AI coding is just the latest abstraction layer, and the job of a programmer has always evolved with each one.
Andreessen's key reframe of what's actually happening:
"AI coding actually abstracts away the process of actually writing the scripting code... This is the next layer of the task redefinition under the job of programmer."
He's clear that the best programmers aren't being replaced. They're already adapting, even if their day-to-day looks radically different now.
Their job has shifted from writing code line by line to managing dozens of AI agents working in parallel.
"The world's best programmers today will tell you, 'My job is I'm sitting there and I'm orchestrating 10 code bots running in parallel.' Their day job now is kind of arguing with the AI bots to try to get them to write the right code."
But @pmarca is adamant this doesn't make foundational knowledge obsolete — it makes it more important.
"You need to still fully understand and learn how to write and understand code, because if it doesn't work or it's not doing what you expect, you need to be able to understand the results of what the AI is giving you."
He draws a direct parallel:
Just as someone writing scripting languages still needs to understand how a microprocessor works, someone orchestrating AI bots needs to understand the code those bots produce.
"It's this upleveling of capability where you actually want the depth to go down and understand what the thing is actually doing, even if you're not spending your day doing that by hand."
The result, in his view, is transformative:
"Now programmers are going to be 10 times or 100 times or a thousand times more productive. And that is overwhelmingly a good thing."
The pattern: New abstraction layer emerges → tasks change → the job gets redefined upward → productivity explodes
It raises a question every programmer should be sitting with...
Are you building the depth to evaluate what AI gives you, or just accepting the output?
When someone forwards you a long email that you have to read backwards from the bottom, you can instead have Gemini in Chrome + gmail translate it into a play script. For extra laughs, ask for the style of Shakespeare.
Julius Smith retweeted
I just watched a great compilation of various people's views about what is coming:
xitter.cf/Scr0nkf1nkle/status/19…
We've always been told "Look before you Leap".
Now, with LLMs, it's
"Look before you Think before you Leap".
Regarding copyright, I favor a technical solution wherein we automatically measure similarity and require a certain distance in that space to avoid copyright infringement. The next step is automatic differentiation, i.e., you upload your work and some server tweaks it to be different enough from everything copyrighted. People can upload their images to be placed on a "do not resemble" list. Differentiation tweaks in a high latent-space layer should look/sound really good, but different. We could also use it to invalidate copyright claims where there is older, non-copyrighted material within the official "latent-space radius of infringement".
The two-party system is clearly broken. I think we need (1) Ranked Choice Voting, and (2) Liquid Democracy. Adapting to (1) is easy, like now but better. (2) adds finer control over one's vote assignments, if desired, and can be ignored when disinterested or too busy.