Besides founding JPL, Qian also co-founded the university I graduated from, USTC (University of Science and Technology 中国科学技术大学), one of the most prestigious schools in China. He headed the Department of Modern Mechanics for 20+ years and was so notoriously strict that the university had to extend its graduation track from 4 years to 5, as he refused to let students graduate without a thorough understanding of science.
Legend has his infamous, multi-hour-long finals had only one question:
With no calculators/computers allowed, calculate the trajectory of a rocket launched from Earth that will orbit the Moon.
The Original Airpods Max had terrible connectivity. The bluetooth would break down for days, and all instructions on the internet would tell you needed to remove the device and re-add it. This was not just a single incident, but kept on happening to LOTS of users. Frustrated, I went to Apple service to learn that they could not fix it and could only replace it for like 70% of a new one.
So, if Apple keeps on building new features while not fixing major problems, I'm not considering buying these unreliable, expensive products at all ...
Our team doesn't use things like OpenClaw. We use Windmill, which is a workflow engine that can use AI to program scripts and workflows. 90% of our automated flows are just vanilla python or nodejs. We re-use 10% of scripts in different flows. We do have LLM nodes and other AI nodes, and although windmill totally supports AI tool calling/Agentic use, we DON'T let AI agents decide the next step. Boring? Maybe. Dependable? Yes.
Synology is actively downgrading their systems; it used to have video station and h265 hardware decoding on 7.2.1, but on 7.2.2+, Video Station is removed and so is h265 hw decoding.... and guess what, if you update, you can't downgrade.
There is an app called Piano Maestro that makes it much more fun, a large pool of pop songs, and increasing difficulty. We use multiple apps at home with our Yamaha piano that has a Bluetooth midi connected to it, including Notequest, Notevision and recently Piano Maestro.
Sorry, but as a paid Windsurf user, I think Windsurf should stop chasing shiny frontier models and focus on building better predictable & manageable workflows to build real-life products.
- How about providing a Jira/Trello-style dashboard with subtasks for our AI, instead of copy-pasting "Cline Memory Bank" to .windsurfrules?
- How about supporting TDD and regression-fixing by default?
- How about using git with branches instead of the current undo-redo system?
- How about a better way of syncing documentation vs real code?
We are paying for more "manageable" AI agents to get stuff done, not a chaotic "genius-hacker" to hack together quick prototypes.
Hi, I've been using notequest before and this really hits a string for my needs. There are some places that could be improved.
1. There is lacking logic in the notes, there should be some kind of pattern that an LLM could create that sounds better. For example for left hand usually there could be eg. walking bass/ chord progressions etc, instead of just random notes.
If you need ideas and help on that, you can contact me at (myhackernewsname_at_gmail_dot_com) I am willing to help
2. Now the game is like, press the notes as fast as you can, instead, a musician needs to practice, pressing the note at the right time, which means you can make a metronome + press the notes on time mode, which is even more useful than the present version.
Great explanation!
It's actually a kind of version of "TL;DR," which stands for "Too long, don't read."
One of AI's benefits is personalization, which means that you can jump between abstraction levels, or in other words, information density levels.
Hmm, isn't there like five lives before "ending"? So instead of doing a "perfect run" there should be chances, and feedback, such as "one missing", like a real human player?
I greatly respect how image benchmarks like MNIST and Imagenet have contributed to AI advancements, so this is my attempt at developing a benchmark. Please leave feedback.
Of course, babies and adults are different but more important is how language is learned or acquired. Most adults do not have mentally stimulating content that is basic enough. There is a dilemma where the content they need is the content they don't want to watch.
Thanks Xiaojun for FSRS, I've talked multiple times with him about FSRS and Anki.
In traditional spaced repetition systems like Anki+FSRS, learners often review knowledge always in the same context, which can lead to an understanding of knowledge as isolated units rather than part of a larger network. For example, learning words isolated does not fully prepare learners to use the language in real-life situations, where context, collocation, and usage play a significant role.
Instead of isolated repetition, a better way to learn languages is to expose learners to words in different contexts and combinations, helping them to understand how words naturally form in chunks and sentences. This varied exposure can help learners build a deep network of neural connections, enhancing their ability to understand, recall, and use language effectively.
So for learning languages, instead of Anki cards, we're building something that shows real-life short videos where each of our videos is a mini scenario, showcasing language use in real-life situations. This means students aren't just memorizing words, but seeing how they're used in natural contexts, complete with visual cues and body language. This approach allows students to understand the practical application of words and phrases, making the learning process more engaging and effective. Since the main reason that students fail to learn foreign languages is they don't keep on studying.
Another thing that Anki lacks is, real-life scenario-based quizzes instead of "do I recall this?", students need to apply what they've learned in a fun, low-pressure environment.
So, SRS algo is just part of the solution, and definitely not a silver bullet. There are a lot of things about memory, such as how emotions affect encoding - you don't need SRS to remember your first kiss or graduation day. People learn their mother tongue to fluency w/o using any kind of SRS, while most people fail to learn a foreign language to fluency even if they use SRS.
Newton's Bucket is a thought experiment proposed by Isaac Newton, which he used to demonstrate the existence of absolute space. He imagined a bucket filled with water suspended by a cord. If the bucket is twisted, and then released, it begins to spin. As the bucket spins, the water inside it climbs up the sides, forming a concave surface. Newton argued that this change in the water's shape was due to its rotation relative to absolute space.
Ernst Mach, a 19th-century physicist and philosopher, offered a different viewpoint. He suggested that instead of absolute space - the bucket can be seen as stationary while the universe is rotating, and the water would also have the same concave surface. This is known as Mach's Principle.
According to General Relativity, a rotating universe can actually "drag" the space and time in it, affecting the spacetime of the bucket, so there doesn't need to be any centrifugal forces. This omits the need for any "absolute rotation" or "absolute coordinates".
I want to share a behind the scenes story in Douyin, the Chinese version of TikTok, which relates directly to the subject of this article. Douyin decided to test out its own e-commerce platform, in an alternative to handing over its users to prominent retailers like Taobao and JD.com. They planned to test the water with one category at a time, starting with books.
Douyin worked with major publishers to offer their products on their e-commerce platform. The day they began selling books on their own platform, Taobao's book sales immediately plummeted by half. To everyone's surprise, users were reading and buying more books thanks to Douyin.
The story ends with Jack Ma, the founder of Taobao, arriving at Bytedance HQ to strike a deal on further collaboration between Douyin and Taobao.
Interesting point about SNR, social network and moderation. Ten years ago, when I interviewed for the Chinese clone of Facebook(renren 人人网), I told that the one of the major problems of social media is the SNR, and my interviewer insisted it was a feature problem. I didn't get the job, but within 3-4 years, renren goes downhill, and today nobody uses it anymore. Increasing SNR is actually a difficult question, because user behavior doesn't actually mean signal, something that is eye-catching, click-baity, is actually a good signal for the user.
Something interesting about Tiktok is it was designed for optimizing SNR, especially, only having one item(which is a video) each screen, instead of having a list of content. This, and autoplaying videos was breaking all the rules of app design guidelines. Now it is copied everywhere.
So how does Tiktok influence SNR, if user behavior does not compeletely correlate on what the real signal is. The secret recipe for Tiktok is human moderation, not volunteers, but tens of thousands of people, curating, moderating, to complement its realtime recommendation system.
Calling learning remembering, is like calling a computer a hard drive. Sorry, but, remembering is just one part, although important part of learning. There is loads of difference between remembering the words of a foreign language and fluently speaking it. The process of learning is complicated, eg. good teaching is not giving the final answers to learner, which should seem faster, but letting the learner arrive at their answers, even sometimes arriving at a wrong answer at first.
And the point about reducing forgetting, that's also a terrible way of learning. If you want to minimize forgetting, that means you have to learn in baby steps, repeat everything using some kind of spaced repetition algorithm, and endure the pain of doing all this for a considerate amount of time. Instead, learning the content rapidly (meanwhile forgetting a lot of the content), getting a whole picture, and gradually gain more understanding repeating the process using different textbooks/courses, is a much more faster way of learning. The brain is more used to BFS (breadth-first-search) than DFS (depth-first-search), and the whole process is much more stimulating. This is because new knowledge needs to be encoded with connections to pre-existing knowledge, and most knowledge is intertwined together, for example in calculus, limits, derivatives, integrals, infinite series, etc. Learning limits in isolation of the whole picture, optimizing for less percentage of forgetting, often leaves learners confused, why am I even learning this, and what use is this for. Even though the percentage of forgetting is much higher in the latter holistic BFS process, the overall content mastered in the same month or week duration is higher. So forgetting is actually normal and you shouldn't panic when you forget things- since you will gain more understanding each time you learn and re-learn the content.
For the last ten years, I've been making products for learning, and every day now and then, I'm still learning something new about learning.
Sorry but to say, TikTok for P*rn has already been made, works for iOS and Android. Probably already has millions of users, with a percentage of them uploading original content. Not giving the name or link here for obvious reasons.
We are paying for good quality content, but we have a problem in promoting our contest. Do you have any good ideas on ways of finding young people who could be interested?