This won’t happen in our lifetime. Not only because this is more complex than rambling vaguely correlated human speech while hallucinating half the time.
About half the time, the text closely – and sometimes precisely – matched the intended meanings of the original words.
Don’t be surprised but about half of the time I can predict the result of a coin flip.
I’m not saying it’s not interesting but needing custom training and an fMRI is not “an AI can read minds”
It can see if patterns it saw previously reappear in a heavily time delayed fMRI. Looking for patterns you already know isn’t such an impressive feat Computers have done this for ages now.
Later, the same participants were scanned listening to a new story or imagining telling a story and the decoder was used to generate text from brain activity alone. About half the time, the text closely – and sometimes precisely – matched the intended meanings of the original words.
You left out the most important context about “half of the time”. Guessing what you’re thinking of by just looking at your brain activity with a 50% accuracy is a very very good achievement - it’s not pulling it out of a 1 or 0 outcome like you’re with your coin flip.
You can pretend that the AI is useless and you’re the smartest boy in the class all you want, doesn’t negate the accomplishments.
Being close (and “sometimes” precise) to the intended meaning is an equally useless metric to measure performance.
Depending on what you allow for “well close enough I think” asking ChatGPT to tell a story without any reading of fMRI would get you to these results. Especially if you know beforehand it’s gonna be a story told.
Sophisticated local trained models on expensive private hardware are already dunking on publicly available versions. The problem of hallucination is generally resolved in those contexts
Sure but until I see such a thing I chose not to believe in fairy tales.
Decompiling arbitrary architecture machine code is quite a few levels above everything I’ve seen so far which is generally pretty basic pattern recognition paired with statistics and training reinforcement.
I’d argue decompiling arbitrary machine code into either another machine code or legible higher level code is in a whol other league than what AO has proven to be capable of.
Especially because with this being 90% accurate is useless.
It’s not, though. Hallucinations are inherent to the technology, it’s not a matter of training. Good training can greatly reduce the likelihood, but cannot solve it.
Why does a pre-trained model need expensive private hardware after it was trained, other than to handle API requests faster? Is Open AI training chat-GPT on inferior hardware compared to these sophisticated private versions you mentioned?
The fine tuning, while much more efficient than starting fresh, can still be a large amount of work.
Then consider that your target corpus of data may also be large.
Then consider to do your reasoning tasks across that corpus also takes strong hardware to get production ready response times.
No, openai isn’t using inferior hardware, but their model goals, token chunking strategies and overall corpus are generalist in nature.
There are then processing strategies teams are using to go beyond the “memory” limitations gpt 4 has, that provide massive benefits to coherency, essentially anti hallucination and better overall reasoning
I think it’ll be in our lifetime just not anytime soon. I feel like AI is gonna boom like the internet did. Didn’t happen overnight and not even in a year but over 35ish years
That dosen’t really translate to neural nets though. There is nothing inherent about matrix multiplication that would make it good at reading code. And also computers aren’t reading code they are executing it. The hardware just reads instruction by instruction and performs that instruction it has no idea what the high level purpose of what it is doing actually is.
Half of programming is writing code, the other half is thinking about the problem. As i learn more about programming i feel that it is even more about solving problems.
It’s the other way round. Code is being written to fit how a specific machine works. This is what makes Assembly so hard.
Also there is by design no understanding required, a machine doesn’t “get” what you are trying to do it just does what is there.
If you want a machine to understand what specific code does and modify that for another machine that is extremely hard because the machine would need to understand the semantics of the operation. It would need to “get” what you were doing which isn’t happening.
Idk the specifics, but what you say makes it sound like it would be easier to create an AI that recreates a game based on gameplay visuals (and the relevant controls)
That game would still not work because there is a ton of hidden state in all but the simplest computer games that you cannot tell from just playing through the game normally.
An AI could probably reinvent flappy birds because there is no more depth than what is currently on screen but that’s about it.
It was a staple of Asimov’s books that while trying to predict decisions of the robot brain, nobody in that world ever understood how they fundamentally worked.
He said that while the first few generations were programmed by humans, everything since that was programmed by the previous generation of programs.
This leads us to Asimov’s world in which nobody is even remotely capable of creating programs that violate the assumptions built into the first iteration of these systems - are we at that point now?
Very far from reprogramming though. The general shape of the NN doesn’t change, you won’t get a NN made to process images to suddenly process code just by training it.
So back in the ‘70s I dabbled in programming (now called “coding”, I hear). I only did higher-level languages like Fortran, Cobol, IBM Basic, but a friend had a job (at age 13!) programming in assembler. Is assembler now called assembly, or are they different?
I thought that the assembler is a specific program that translates mnemonics into the corresponding machine code. Perhaps in early computing this was done by hand so a person was the assembler (and worked in assembler), but now that is handled by software (and supports various macros). So programming in assembly would generate a stream of text that must be assembled by an assembler. (Although I have heard people refer to programming in assembler as well, just not often.)
I hear people say “program in assembler” but IMO that’s wrong. I’d say you write the code in “assembly language” (or better yet, the actual architecture you’re using like “x86 assembly”) but you “assemble” it with an “assembler”. Kind of like how you could write a program in the “C language” and “compile” it with a “compiler”
A compiler and an assembler do wildly different things though. An assembler simply replaces mnemonics while a compiler transfers instructions to a whole other language.
Depends on the language, really… C maps pretty closely to assembly language, it’s not as simple as one mnemonic to one machine code byte, more like tokens get mapped to sequences of machine code, a function call translates to some code that sets up a stack frame, a return tears it down…
It’s still called programming, coding is the same thing. Assembler more commonly refers to the utility program that converts the assembly code to machine code while assembly refers to the code itself, but the term assembler code is also valid. It’s uncommon to simply call the code assembler because it would be easily confused with the utility program.
I was too young/poor to afford an assembler for my 6502 so I wore out the assembly long hand on a legal pad and then manually converted each operation to machine code.
Needless to say my programs done this way were exceptionally simple, but it’s interesting to understand the underlying code.
Having access to the source code actually makes reading machine code easier, so you’re also wrong on this entirely different thing you’re going on about.
I didn’t say it was. I just said loosely what the OG meme said, if you know how to read assembly, you know how to read (and write) what some of the code does.
well assembly is technically “source code” and can be 1:1 translated to and from binary, excluding “syntactic sugar” stuff like macros and labels added on top.
By excluded he means macro assemblers which in my mind do qualify as an actual langauge as they have more complicated syntax than instruction arg1, arg2 …
The code is produced by the compiler but they are not the original source. To qualify as source code it needs to be in the original language it was written in and a one for one copy. Calling compiler produced assembly source code is wrong as it isn’t what the author wrote and their could be many versions of it depending on architecture.
Ok, getting past the dickish, completely unhelpful first part of your reply (as you can see in the comments, not EVERYONE was saying that), the second part helped me trace it back to this:
which is a toolset that I never intentionally installed, and was evidently added by an emulator package without me knowing where it was or what it did.
So thank you for (eventually) helping me find what it was, and now you and others know how to add it to cmd and don’t have to complain about its absence.
‘Cmd’ is to ‘horse and buggy’ as ‘powershell’ is to ‘automotive vehicle’
Have no idea where you decided to pull this 100mph idea from. It wasn’t a comment about speed, it was a comment about utilizing modern practices and tools. And that the joke falls apart because it’s forcing the narrative all windows has is cmd and blatantly ignoring pwsh. It be me like making a joke how linux can’t do wifi… because there was a time Windows did wifi just fine but linux required using a special process using an ndiswrapper with windows drivers to get linux on wifi… so like 16-17 years ago. See, hilarious.
What? Ignoring the out of date, over used, ‘gottem’, phrase… it literally doesn’t make sense given the context. I’m advocating for modern tools… how is that a ‘boomer’ move?
I ended it with a question mark because I was uncertain but otherwise your asinine, dismissive statement sounds like something a jackass who’s been around too long would say because he’s entirely too full of himself. I’m willing to recognize that it’s actually ignorance formed out of youth. Don’t think about it too hard.
That is interesting. I just remoted into 5 different machines at the office and none of them worked with ‘ls’. If you enter ‘ls /?’, does it give you a synopsis and argument list?
If I do “ls /?” it returns no such file or directory, but just “ls” performs exactly as you’d expect. I haven’t installed anything to provide that function that I know of. It never occurred to me that I would have to because as far as I know it’s always worked. Until today I just assumed it had become a standard command and never investigated. Was just happy I could use the same command in cmd and on my Pi box.
It’s my own fault, and the result of 30+ years of muscle memory building up. Plus, while I agree cmd isn’t nearly as powerful as powershell or wsl can be, when I’m in Windows it’s still the fastest way for me to do 90% of the simple things I need to do. I have a long history with it, and a thorough understanding of it, so I don’t really need to think for most of the things I’m doing there.
If I need to script something, or do anything that seems like it would be annoying to do in CMD, I hop into WSL pretty quickly and get to work with bash or python. The problem I have now is that I’ve developed a little muscle memory there as well… hence my issue with entering ‘ls’ everywhere.
Lately I’ve been using it as a simple way to drag and drop a source .tar.xz archive on a .bat file so it can be twice extracted, moved, renamed, have dependencies downloaded by git, run a cmake process, do a visual studio compile, then move the result release directory back to where the .bat file is while removing unneeded files and adding new ones.
programmer_humor
Hot
This magazine is from a federated server and may be incomplete. Browse more on the original instance.