There's already hedge funds that scrape the insider selling data for the executives.
You can then build a statistical arbitrage alpha based on the difference between an ordinary amount of selling (conditioned on a bunch of stuff) and a lack of selling/accelerated selling.
If you're building a home lab, you might just give up because you are tired of searching for Cisco router syntax.
LLM can be the little bit of lube you need to understand things at a higher level, because the tedious things are not as bad as they used to be.
You then end up actually getting a few things in your lab to work correctly, get a confidence boost, and follow up with some searches about how routing actually works and why the syntax is the way it is.
The stochastic parrots thing is already old. It might have made sense at one point, but now that the parrots are solving millennium problems, I think people have to admit things have changed.
He is right that the AI giants would like a bit of regulatory capture.
It can still be stochastic over more complex token graph traversals, but it is still stochastic at the core. Anyone who uses them daily knows how often they make the most absurd assumptions and false statements, and show the limit of engineered intelligence
If the parrot randomly yaps out a solution to an open math problem, it's a useful parrot.
The fact that it sometimes says things that aren't useful, well, old intelligence does that all the time.
If you've used AI recently, you can't avoid having found it useful. Just general day-to-day intelligence like finding things in your email, writing code for you, that kind of stuff, it does well. People have some weird hangups about it being stochastic, when it doesn't really matter.
yes, the spaghetti throwers (current training regime) sometimes hit the mark quite well, but they also do some of the dumbest things, reinforcing to the stochastic nature of their construction.
If it were me, I would start by getting a summary of what problems the existing solution solves. I would also make sure that summary had the non-functional requirement. Maybe write a test suite that opens the page and does all the stuff.
Then ask the LLM to implement in another language, with those solutions in mind, but first coming up with reasons why the target language might not be as good, and where it might have useful features that the original language didn't. Use the test suite to check if things are working.
I honestly don't think you would land far away. You would have to answer a few questions along the way, but it would mostly be plain sailing.
I would expect to be able to do this with very little human attention, whereas a language rewrite two years ago might take a whole month, not including learning the new language.
You're contradicting yourself, how can it be unique if only 1000 can be assigned per given date of birth? What if more than one thousand babies are born in the country one day?
There have been collisions in Sweden, but that was due to lots of immigrants coming from the middle east that did not know when they where born. So many of them picked first of January. They “solved” it by overflowing to the next days.
They said sex is encoded in it, so it’s probably a sex digit.
edit: I was wrong. Wikipedia says, “The first digit of the sequence number encodes the century of birth (so that centenarians are distinguished from infants)”.
It also says “the last digit of the sequence number is odd for males and even for females”. What a strange system. Essentially the last three digits are two different sequences made to look like one.
I had a chat today with a friend about this. I think my working thesis is that AI is bad news for outsourcing/offshoring.
What was it good for? Well, often it was prompting a person in a low-cost location what code to write. He would pick it up in his timezone and implement it overnight, and present findings/questions as a response.
The problem is, coding has made implementation super cheap. It is now so cheap you have to compare it to the cost of dealing with someone in another timezone, who does not have the same cultural context on the problem. They also can't ask you clarifying questions while you're asleep. AI wins against the offshore guy, on both money and context.
Who does it help? Amazingly, the experienced onshore dev who thought he was going to be unemployed past 40. This person is now able to implement several things concurrently, has experience in making technical decisions, and has local cultural context.
Whatever it is being asked, it is great at throwing out new suggestions. This MRI idea sounds like another one of them.
It works well because humans are crap at checklists. For more or less anything you can think of, you'll come up with maybe 4 items, and the AI will add a few more. They may not be relevant, but the AI is very thorough.
At what point is human intelligence going to hold back machine intelligence?
Imagine you are evaluating what the machine should do when it is improving itself. It does a bunch of work and returns with "I supervaluated the liminal overdecomposition from the previous homological calibulation pass. It shows us that subtransitory mulutination will underspecify the tensor of stermullification. Where do you want to go from here?"
It will be like when you are reading a Wikipedia about a topic you don't understand. You follow the links, and you get more questions with more links. Your whole day is taken up following links, to the point where you forgot the original question.
Except this time, all the words come from the AI's work. You can't refer to an external authority who has already been there and can tell you what to do.
The AI needs you to tell it whether it is more intelligent than it was before, but you don't know, because you can't follow its reasoning any more. It's like an ordinary person trying to hire a math professor, there's just no way to do it.
But whereas a human math prof can evaluate another one, a machine intelligence can't evaluate another one, by construction. Because it's still usefulness to humans that is the evaluation criterion.
Is there a way for an llm to coin a word, and absorb it into its model? During training maybe… but not after - not the way they’re designed now, anyway.
For it to have new vocabulary we dont understand, it needs to have novel ideas that need words coined for them, and a way to persist those ideas and words into the future. I don’t think that exists.
To me this hypothetical make it clear this won’t happen, not unless there are fundamental changes to what llms are. It doesn’t suggest it will happen. To me, anyway.
The decoding step (output of final layer -> word) is not strictly needed. You can feed the output directly into the next layer (Chain of Continuous Thought). You can 'decode' the output into things other than words.
How does that give the llm a way to speak in meaning that humans can’t understand?
You input text into the system, that moves through layers as chains of weights. Sure it doesn’t ever have to be output as text, but for the text-in-text-out machine to have any impact on the world around it, text must be generated.
I don't see why this couldn't be possible. We use LLMs whose weights are frozen and are not updated at inference, most likely this is due to reasons of cost, stability and control.
Theoretically you could update the weights at inference time too though so the model evolved as it's used. Surely some people are trying this already.
This is a very outdated view on what an LLM is and how it works. We are way past the "stochastic parrot" phase, ever since double descent and proper generalisation. Then with the various flavours of RL the models learn to pluck patterns / circuits out of the massive data and combine them on the fly. There's absolutely no reason to think they can't "invent" new words, because words are just combinations of tokens at the end of the day. So if they can come up with "in this codebase bar is load-bearing" they can similarly come up with "bumblespin is the new word for reversing the polarity of the quantum surface of a spin-aware brane in four dimensional bumblespace".
So you’ve claimed that they have half of my precondition for them to speak in ways that we don’t understand.
Coining a term on the fly: yes
Remembering that term and incorporating it into its mind: no.
But, beyond saying they aren’t stochastic parrots you haven’t really given anything to substantiate the idea that they aren’t (very complex) stochastic parrots.
Interesting. I don’t have a counter to the coining a new term evidence in that article. Potentially that’s solid evidence of the first aspect.
But I do have a counter for the second point:
Context is not a part of the model. It’s an ephemeral blend of input and output fed back in as input.
Online learning is potentially a way for incorporating and evolving coined terms to happen. But since that’s not what any of the models are doing, something would have to change in how we do things before it could happen.
> Imagine you are evaluating what the machine should do when it is improving itself. It does a bunch of work and returns with "... Where do you want to go from here?"
It's worth pointing out that so many people already see and use AI the same way. In such a situation some would ask AI to provide options, and they would choose and experiment with those options. Of course, contexts/stakes can be vastly different.
Given that there are so many phenomena in nature that we can't explain or fully understand which doesn't prevent them from existing or being useful, there could be a future where humans accept the same about the things AI comes up with as long as this leads to desired outcomes. We still might have names for them, but our brain thinking/knowledge capacity wouldn't allow us to fully comprehend them. We'd need frameworks/systems in place to turn these AI features/decisions on/off, although it's hard to imagine how this wouldn't increase the likelihood of something going out of control.
The model will have to convince the human that it's making the right kind of progress. That will necessarily become part of the improvement loop - either implicitly (human trusts RSI) or explicitly (human gatekeeps every major decision).
I love your thought experiment. May I counter, what purpose does such a machine have to us, that can think beyond our needs? Sorry, but to reference the great Rick and Morty, "your purpose is to pass the butter".
“To us” is pivotal there. Continuing GP’s thought experiment: what if the model that produced the output perceives that the human it was presented to offers no value in helping it learn further?
I think if an AI developed completely new fields of thought or science.
But given our current relationship even if it did I can't foresee a point where it couldn't walk us through the necessary steps or supply the pros/cons for whatever problem is being addressed.
2 issues - trying to understand how it came to its conclusion because I feel true AI has got to be non-human intelligence. Or a something catastrophic happens and we as a species are back in the stone age. Imagine today's AI trying to converse with a cave man (yes, one without modern languages even).
From Jacobian. They developed a tool called a Jacobian Lens, or J-Lens (itself a descendant of an earlier simpler tool called a Logit Lens) to examine what was going on inside an LLM, and named the space they found with it a J-Space.
Jacobians are essentially derivatives but for matrices.
You can then build a statistical arbitrage alpha based on the difference between an ordinary amount of selling (conditioned on a bunch of stuff) and a lack of selling/accelerated selling.
reply