Doors aren't humans either, yet we design their handles and locks to be graspable and manipulable by humans.
The purpose of technology is to serve humans. Therefore, technology must conform as much as possible to human sensibilities rather than vice versa.
I will accept a 5% drop in benchmarks for a model that talks to me like a human.
Mistakes in proofs are relatively common, but such a mistake doesn't automatically mean that the proof is entirely wrong. Many times, the mistake is just in the exposition, and can be fixed easily. Other times, the mistake is fixable and the fix is apparent. Perhaps the author forget to treat a relatively trivial edge case. In the first two cases, the student would likely just pass with minor corrections to be submitted soon.
Sometimes, of course, the proof is just wrong. That is the dangerous case, which will cause either major corrections or failure.
Funnily enough, you will have to pretty accurately copy the laws of our universe to make an Apple pie. Even small changes will change the properties of atoms, and hence all of chemistry and biology. And there won't be any apples or wheat.
"How does Kimi Work protect my privacy" can be interpreted in two ways. The first is "How does Kimi Work protect my privacy from Moonshot?" They did not answer this question. The second is "How does Kimi Work protect my privacy from other third parties?" They are answering this question by saying, we won't upload your files to some third party server without your consent.
How many free hosts from 20 years are still alive? Not many.
Dream would be a combination of AI models that are smart like a human transcriber. One can just tell the model what mistakes it has made (e.g. "No, you wrote XYZ but I meant WXY"), and it is intelligent enough to realize when it is being instructed and when it needs to transcribe exactly what I say.
This is pretty cool. The dream would be that all editors (including brower tabs) have an api interface which allows arbitrary LLMs to modify the editor content. So you can indeed look at different windows as you talk, but the editor keeps getting updated live, so you can go back and see where you already at.
In fact, a few apps I have tried do exactly this inside the app themselves. There is a live textual field that displays whatever the model thinks, and this display constantly goes back and fixes earlier words.
So I know it's possible. I just want to integrate this with the paste-at-cursor feature that these apps have. I imagine the app would have to create a virtual keyboard and use backspace or arrow keys to go back and change things.
For anyone looking to build on top of this. I have tried a few different STT systems, and they accurately capture what I am saying. Unfortunately, they don't support the reasonable workflow
I want to open an office document, for example, and start talking. And I want the software to continuously type what I am saying at the cursor with minimal latency. The continuous part is crucial. Many software will paste whatever I said after I have stopped recording, but that is not useful.
I would bet the legged robots are advertisements, and when a factory comes knocking they are provided the option of buying cheaper robot arms that are stationary or on wheels.
Robotics and textual AI have very different risk profiles. Textual AI can just be provided to users even if they are completely crap because some users will get some usage out of them.
Robots can only be released once they are good enough to actually do anything, and have passed safety certification. So, not seeing robotic products out there is completely natural. But they are not far behind LLMs, and they will very suddenly take over the the factory world as soon as they cross these performance and safety thresholds.
Unfortunately for workers, these type of actions can only term delay their replacement only by a few years, if that. The only long term solution here is to go bigger and reconceptualize the entire economic system, a system that takes into account the new realities that AI has created, and provides justice to all.
Lime is more breathable, workable and sustainable. The only thing is, it requires maintainance
What sort of maintenance does it need?
The confounding variable I think is most important is that many people have an internal clock that is shifted or lengthened relative to other people in the same house.
If others in the house prefer to sleep from 10-6 and you prefer to sleep from 12-6, but others start making noise a 6, your sleep quality in the last two hours is destroyed. Then over time, it just results in poor sleep regularity, as you cycle between exhaustion and trying to sleep according to your internal clock.
An interesting solution would be for these AI companies to train a few different versions of these models, all with different speech characteristics. Then, when you start a conversation, you get a random version.
Just today, I got frustrated with the language. I searched around, and in my Claude Instructions I put in Ref [1] (translated to English). It is certainly better phrasing (though still quite annoying), but I don't know if this makes the output technically worse in some way.
[1] https://github.com/hexiecs/talk-normal/blob/main/prompt-chat...
I wouldn't expect this to be better tasting than a regular soup/stew. If you keep eating it, new stuff added today has been almost completely consumed in a week.
I would eat it out of respect for the craft and the values that are being preserved.
In text, you have the advantage of putting things in any order, with the assurance that the reader knows exactly how long the message is.
Always start your email to a busy/important person with exactly what you want from them. Within the first two sentences. Then peel the onion, giving details from more important to less important.
This gives the reader opportunity to stop reading at any point having decided to help you or not, with the least waste of their time. In many case, the person can skill your description of your work, and just tell you what to do.
If there is a choice, I would much rather someone use an AI to do the project, and then used their human words to explain what the project is, rather than vice versa.
The critical difference is that a project artifact (software or mechanical design) is good as long as it works. It might not be maintanable, or editable or extendable, but it might narrowly just work. But explanations don't work like that. The content, the actual words matter just as much as the overall message.
An explanation can be thought of as software executed in your brain as you read it. I don't want to execute badly written software in my brain.
Not all the time, but I certainly do to keep up with latest results. Usually, these days I go through SciRate, where the quantum computing community is very active in voting up good paper [1].
Workers enjoy highest living standards of any time in history.
It's entirely possible for someone to be paid a lot in absolute terms, while at the same time paid very little relative to the value that they produce which is monetarily captured by their organization. The truth of the first does not invalidate the injustice of the second.
Most popular sports have world cups every four years. Basketball, cricket, rugby, volleyball, field hockey WCs are all held every four years. Some others like handball are held every two years.
Cricket gets the same level of crazy support in South Asia as football does in most of the world. But the reason football gets such support in most of the world is simply because most people consider it the best sport, or the "beautiful game". It is one of the most well designed sports, having the fewest number of unnatural rules - only one in fact (offside). It can also be played with basically any number of players from 2-20 in fields of any size, so it's really popular for normal people. The skills floor is reasonably low, but the skill ceiling is very very high.
So, then it flows naturally into social actions like blocking streets.
They address this in the conclusion
How generalizable are these methods? Can they consistently deliver truly high performance? Can we get to a place where AI produces designs that maximize every conceivable trade-off, holistically optimizing every parameter to its most ideal physical state? .... AI can hallucinate a design that creates bad circuits that don’t work. This means verification methods need to remain under human oversight.
And they are essentially correct. We need better validation and verification methods, both software and hardware to keep in check the mistakes of automated random processes.
Terence Tao talks about this [1]
In particular, the process of converting a medium-sized Lean document (containing a few thousand lines of code, with some proofs AI-generated) into a nicely golfed and structured Mathlib submission has been an interesting experience. AI agents can be used to perform local golfs that can shave the size of the code somewhat, but global refactoring decisions, such as noticing that a certain argument appears multiple times across the document and can be abstracted into a standalone lemma that can may have additional utility beyond the file, is still largely beyond the reach of current AI tools. (I find that I can explain such a refactor to an AI agent, who can then execute it, but they struggle to spontaneously discover such refactors on their own.)
This is a weird viewpoint. Most academic work in STEM fields is collaborative with 2-100 other scientists or engineers. This is because it is well realized that 2 people talking can do many times better work than two people working alone.
Your supervisor should at different times, as appropriate, wear the collaborator cap or the examiner cap. During the course of your project, he should be your collaborative advisor who helps you flesh out ideas and improve them. At the start and end, he should be your formal examiner.
This is how I have seen most Masters (and PhD) projects proceed in Physics in North America.
A simpler way to do this would be for the user to create and place some temporary virtual "grids". For example, the user might create a virtual triangle of a certain size. Or in the case of this request, a circle with n points. Note that these are virtual and temporary items never created in the underlying tikz code (though you may implement them with hidden tikz code).
Then the user can do one of two things. (1) Select an item and place it on the grid, and the item gets replicated on all the grid points. (2) Pick and place different items on each of the grid points.
Besides Wacom, which tablets would you recommend as good quality?
Nastaliq fonts already exist and used whenever possible. Yes, rendering a Nastaliq font takes marginally more compute. But in a world of electron apps which have a 10-100 times slower UX than they could be, saving compute is not an argument.
It's likely the same problem as in Pakistan. Due to the history of colonialism/control by European powers, in these countries personal economic success is usually tied to command of English or French. So even within each of these countries, the rich, educated and those in power prefer latin script. Consequently, there was never any strong push to develop computing technology for local languages.
The other reason is that it's not technologically simple to solve all the issues highlighted in the TFA. Unicode actually does a pretty decent job of setting a uniform standard, but a lot of software has to be written on top of it to get the entire system working: (1) your software must support bidi text, (2) good fonts must be available to display the text in multiple languages (3) textual data needs to be properly stored in unicode and transmitted as is at every point in the OS (4) search engines must deal with the complications of non-breaking spaces and legacy unicode characters.
You have to kind of rewrite the entire stack from top to bottom. Preferential Arabic/Persian/Urdu speakers never had the technical skills and the political power to drive those changes in software largely written in different continents.