I have been a user of ChatGPT4 from the very beginning and like others I have found it extremely useful in multiple subjects. It is particularly good on topics that there is a lot of prior art and it can sometimes provide surprising insights.
OTOH I have found that it not as good on topics that there is not a lot of prior art in the public domain or in topics like advanced mathematics which often have a formal and unintuitive presentation.
For example, I have acquired recently an interest in probabilistic logic and when discussing results with it, I have found that although it has knowledge of the subject, it cannot really apply that knowledge in a creative way. It will often make logical mistakes and when these are pointed out, it will first apologize and then continue to make the same mistake.
So as far as I am concerned I am unconvinced that it can reason yet.
In the article, I make the case that they can reason but it is in many ways distinct from the way we do it. It is also learned from very different fundamentals. Our reasoning fundamentals come from sensing, and experiencing. LLMs reasoning fundamentals are pieces of words.
We roughly go from sensing, to object permanence, to algebra.
They go from tokens, to grammar, to abstract ideas.
Human senses, inertia, gravity, and other things that are elementary to us, are complex abstract phenomena to them. They can only think of them in very hypothetical terms.
There are some clear examples of reasoning mentioned in the article, but I believe that a large part of the perception of them lacking it, stems from them having unnaturally good writing skills in relation to their reasoning skills.
That is, no human that can write this well is this bad at reasoning.
I am using real data of course, but I am also seeking a deeper understanding and ideally one based on first principles. This is why I asked this question.
It seems to me that if we could correctly characterize the dependence of the returns we could perhaps arrive at the correct theoretical answer that also coincides with practical observation. Mandelbrot almost gets there, but his use of multifractal trading time seems brilliant but also somewhat arbitrary (at least in the papers of his I read).
HUBFS is a file system for GitHub and Git. Git repositories and their contents are represented as regular directories and files and are accessible by any application, without the application having any knowledge that it is really accessing a remote Git repository. The repositories are writable and allow editing files and running build operations.
Have you ever read this HN comment from 2007 before? [0]
> I have a few qualms with this app...
Different paths towards the same outcome should multiply, so we can increase the surface area of the bandages on the different pain points along the way.
That works well for small repos or a few repos but if you want to find all cc files, at all release branch's, in your entire company and check for some exploit it is helpful to have a VFS. Makes it so you could also support N SCMs through one API. You just need to make a new VFS.
Needing to tie into a specific API (like codesearch) couples you to the specific storage backend (Github). If you build your software to operate on a POSIX-y file system, you can support anything that shows up as a file system. For example: A local working tree of files, an NFS share, or now a remote git repository.
I had a similar problem a few years ago. While I never found the perfect machine here is how I have been introducing my now 10 year old son to the world of computers:
- I bought him a Kano computer kit when he was about 7. The kit retailed for about $250 at the time and a kid could assemble it "like LEGO". It came with educational software that introduced him to programming languages, etc. (Unfortunately I do not think they make this kit anymore.) Verdict: MINOR SUCCESS.
- I tried to introduce him to Python (around 7.5) by following an online book about game programming. He did not show much interest. Verdict: FAILURE.
- I introduced him to MakeCode (arcade.makecode.com) around 8. He got absolutely hooked and it is still his favorite platform today. I bought him some cheap hardware (Meowbit) to put his programs on and he loves showing off his games to his friends and everyone else. Verdict: MAJOR SUCCESS.
- I introduced him to Godot when he was 9. He showed strong interest, built a few games in it and even understood enough of the Python like language that Godot has. He used this series of YouTube videos to learn about Godot: https://www.youtube.com/watch?v=HvPTSZl2WCc. Verdict: MEDIUM SUCCESS.
- We are currently building Ben Eater's 8-bit computer (eater.net/8bit) and he absolutely loves it. He is able to follow along with the videos and understand the material at a good level. He has named the computer "Terry". This project does require a lot of my own time. Verdict: MAJOR SUCCESS.
We have also tried other projects (e.g. Raspberry-pi with Raspbian, Arduino, Robotics kits, etc.) although nothing that he showed major interest at the time. I think you have to try with different things to see what will capture your child's imagination.
When I was around your son's age, my parents introduced me to computers by getting me a LEGO Mindstorms Robotics Invention System 2.0 set for Christmas in 2001. My favorite thing in the world was the Exploration Mars expansion.
The "hacker" community around Mindstorms was huge in the 00's. After the RCX was reverse engineered, all sorts of language support appeared. BrickOS for C and C++, lejos for Java, NQC, etc. I bumped into lejos in a section on Mindstorms toolchain options for Macintosh in one of my dad's MacWorld magazines. He helped me set it up and that's where I started learning Java.
It was a life altering gift for me if there ever was one. The first large event that set me on my current path.
I honestly kind of miss the silly hacker/spy/matrix-y aesthetic of computing in the early 00's. It was fun when I realized those LEGO Spybotics kits could be targeted by some of the Mindstorms tools. Everyone in elementary school thought I was going to be an evil genius. Now I build storage virtualization systems. Oh well I guess.
I was in middle school around that same time, my parents got me the educational equivalent of Mindstorms (with RCX 1.0). We didn't have a PC and the educational edition came with a alternative LabView-based UI that ran on classic Mac OS. I remember seeing that article in Macworld and wanting to learn some of the other languages, but I never had much success. I can't say it was totally life altering, I was already headed down the path of an engineering career.
My college was still using the official Mindstorms sets for their intro to engineering classes, and so I recall finally getting my hands on the official Mindstorms software and finding it lacking. A classmate was using NQC to do it all, and his team let him do the programming. My team wasn't a fan of that approach so I stuck to the hardware.
> alternative LabView-based UI that ran on classic Mac OS
Was that Robolab? I remember running that on an G4 PowerMac.
I always thought that the "Robotics Invention System" sets were the consumer oriented ones and the educational ones were a separate product line distributed by a company called Pitsco if I'm recalling correctly. Came in a green tub.
I remember being a kid, and really wanting one of those for xmas - I think it was the first versions they had made of them. Unfortunately it was too expensive, but man did it look exciting, and the idea of having your own robots was so cooooool.
Oh well, years later I got access to computers and ended up doing programming, but now that I could afford one, I cannot find them anymore :(
I got it as a kid but after diving into it I lost interest once I got everything working. I guess for me assembling things and getting them to work was always the most interesting part, after that I moved on to something else. Today I'm a DevOps Engineer :D
The parts should still be somewhere in my father's basement but getting them would take time. I found a new Mindstorms Robotics Invention System 1.5 on eBay (auction number 324365658011) if that is the one you are looking for.
I do miss the heyday of LEGO Mindstorms. Everything seems to have moved to low-cost limited use sets you can get on Amazon. Lowering the bar of access is a good thing, but I do miss the feeling of having a big tub of LEGOs you could build anything you want from.
I guess the cool kids learn CAD and just 3D print the parts they want, which is also awesome (I wonder if there is any research in recycling failed designs or prints back into filament)
I certainly do that with PCBs these days. The cost went down so far I just throw a design together in kicad and have them made for me and shipped over in a week or two. It's only a few bucks more than ordering protoboards and doing all of the wiring myself. I'll exchange a lunch out for that convenience.
Thanks for reminding me of the more "individual challenge" aesthetics of early coding. But I have to ask - you say, "oh well," like that's not something cool. I would have no idea where to even start with a storage virtualization system.
Maybe it's not evil genius, but it sure sounds cool.
+1 for Ben Eater's 8-bit computer, the videos are very well put together and easy to understand. It's even a bit addicting to watch them because you always want to get to the next step. I binged it in a matter of 3 days. His 6502 is also very good, I wonder if that can be a good computing platform OP is talking about.
Godot is a dead end. I tried it and found it very difficult to grok and then immediately ran into their hacked to hell threading model that you must know about (and sprinkle yields around in just the right spots) to get anything done.
I see it being revised significantly which will be a breaking change. If not, then it is what it is.
c'mon, there have been millions of human hours played on games developed on engines or systems with not even 1/1000th of what Godot is able to achieve.
I introduced my daughter to ScratchJr when she was 5 and Scratch when she was 7.
I taught her class Scratch. It does make a big difference which language you start with. I think how much you are involved with your child’s learning make also is a big part of the success.
I introduced her to MakeCode at 8 and we used the microbit to build a robot with legos and servos.
This project was such a hit that her principal asked me to teach it to the middle school kids.
Heroic Labs service looks interesting. So I decided to look at their Terms of Service:
> "You represent, warrant, and agree that You will not use the Services in a manner that is illegal..."
Fair enough.
> "... or otherwise inconsistent with these Terms. In addition, you will not use the Services in a manner that Heroic deems, in its sole discretion, objectionable."
So they can decide that I violated their TOS for any reason they see fit.
This is not the way to build a business relationship. Based on this I would never consider becoming their customer.
idk, seems like a pretty standard condition. Other platforms have something similar, allowing e.g. Amazon to cancel Parler's account on their services. Nobody is required to give you a platform, and if you already think that you intend to use their platform for something they will find objectionable, you should reconsider your moral compass.
Yep, very standard. Here’s the WordPress TOS as another example, under their e-commerce section:
> We may terminate your access to our Services or force refunds (where possible) to your subscribers and customers without notice to you if we determine (in our sole discretion) that your use violates the Agreement, or if a payment processor or regulatory authority requires it.
And, like with WordPress, users of Nakama can still use the open source version without having to ask for any sort of permission. This is much greater user freedom than the average closed-source SaaS.
What an individual may find objectional and what an organization with a founded interest in maintaining a positive public image in the todays political turbulence may find objectional are very different things. It's far too easy to stray on the wrong side of the masses for even the slightest perceived misstep.
I only have only one real tip for you: learn to get in the zone.
When I am in the zone I have an intense desire to create a beautiful masterpiece in my chosen art and revel in its elegance and beauty. When I am in the zone outside distractions are eliminated and my mind is fully focused on the task at hand. When I am truly in the zone matters such as health, family, children, work and (when I was younger) even hunger or the need to sleep disappear; all that remains is the task. My brain has maximum cycles to iterate again and again until I achieve greater understanding and/or arrive at a solution.
I do not have very good general tips on how to get in the zone. I have some things that work for me, although not always. They include:
- Real-world isolation: These days where everyone is at home, I go to a corner in my garden where my family knows not to bother me.
- Virtual-world isolation: No emails. No browsing. No notifications of any kind. Phone on silent. Absolutely no Slack, Teams, etc!
- Friction-free programming environment: If the build takes 20 minutes, then it will be hard to remain in the zone. If the Mac has stuck keys get them fixed or buy a laptop from a different brand.
- Headphones with some favorite music: For programming I like trance, the more repetitive the better. I often use this youtube channel: https://www.youtube.com/user/MrLemilica2
- Review previous work on the subject: This acts as a brain warm up for me and usually puts me in the right mood.
- Some times work better than others: When I was younger it was far easier to get in the zone at night. These days I do not have that luxury because of family.
- Work on a problem that excites you: It is much easier to get in the zone when the subject matter interests you. It is harder (although still possible) when the subject is less interesting.
- No context switching: Choose a task and stick to it for at least a few hours.
- Get the non-enjoyable tasks out of the way: (This may not work for everyone.) I like to complete tasks that I do not enjoy (e.g. administrative or legal) first. I can then continue my work care-free instead of having this niggling thought in the back of my mind: "do not forget to respond to the lawyer".
Like others I found the recording profound at first. The birds at the end cheapened the experience for me because they made me question the authenticity of the sounds heard and whether this was leaning closer to an actual representation or an artistic interpretation. Perhaps somewhere in between.
Assuming that this is leaning closer to an actual representation of the sounds heard on that morning (with the dubious addition of some birds for "dramatic effect") one thing that I found particularly interesting is the noise/buzz that can be heard immediately after the guns are silenced. Could it be that this noise represents the people cheering for the end of war captured by the crude recording device of oil drums + film?
> The birds at the end cheapened the experience for me...
I agree. There is no way that the microphone were that sensitive. I understand what the artists were trying to "say" but yes it did distract from the power of the quiet
OTOH I have found that it not as good on topics that there is not a lot of prior art in the public domain or in topics like advanced mathematics which often have a formal and unintuitive presentation.
For example, I have acquired recently an interest in probabilistic logic and when discussing results with it, I have found that although it has knowledge of the subject, it cannot really apply that knowledge in a creative way. It will often make logical mistakes and when these are pointed out, it will first apologize and then continue to make the same mistake.
So as far as I am concerned I am unconvinced that it can reason yet.