BoiledCabbage

> I am so floored that at least half of this community, usually skeptical to a fault, evangelizes LLMs so ardently. Truly blows my mind. ... > I’m open to them becoming more than a statistical token predictor, and I think it would be really neat to see that happen

I'm more shocked that so many people seem unable to come to grips with the fact that something can be a next token predictor and demonstrate intelligence. That's what blows my mind, people unable to see that something can be more than the sum of its parts. To them, if something is a token predictor clearly it can't be doing anything impressive - even while they watch it do I'm impressive things.

So you're trying to manipulate people into believing false things about you?

Is that better or worse than calling it deception?

I think that's what this thread is about. A lot of people building characters they interact with now seeing a personality change.

Wow would that be horrible

I think the argument is simpler than that. I have a PC, if I wanted to emulate an old Nintendo system well enough to play I dont have to emulate from the physics upwards.

Even though every NES in existence is a physical system, you don't physical level simulation to create and have a playable NES system via emulation.

Performance is doubling roughly every 4-7 months. That trend is continuing. That's insane.

If your expectations were any higher than that then, then it seems like you were caught up in hype. Doubling 2-3 times per year isn't leveling off my any means.

https://metr.github.io/autonomy-evals-guide/gpt-5-report/

This actual answer according the the author realized after he already liked the name.

He created it intending to be +1 of APL. Accidentally came up with BQN instead of BQM. Sat with that for 1hr, really liked the name, then realized that it should be BQM which he hated, so he stuck with BQN.

That said, it's and incredibly designed language. I honestly have never read any language (especially not designed by a single person) with the level of creative thought as he put into BQN. Some really incredible insights and deep understanding. It's amazing reading his posts / documentation about it. The focus on ergonomics, brand new constructs and the consistency/coherence of how all of his decisions fit together is really impressive.

I have little knowledge in this area, but my understanding is it's like this:

There is a pseudocode app that depends on a bunch of pseudocode libraries. They want to translate that pseudocode app into a real runnable app. They can do that, and it's a good amount of work, but reasonable. The problem is to get the app to run they also need to translate the hundreds or thousands of pseudocode libraries into actual libraries. Everything from OS APIs, networking libs, rendering libs, language standard libs all need ro be converted from specs and pseudocode to real code to actually run the app. And that's a ton of work.

It would be funny if this turned out to also apply to quantum computing. Ie while we can build a quantum computer, we can't actually find any productive problem that calculates faster than a classical algorithm.

QC would turn out to be the biggest bust in physics (after string theory of course).

> If everyone could live where they wanted to live, then ya, why wouldn't everyone want to live in the best cities?

Best is different for different people. Some people it means close to beaches, for other it means cultural institutions, for others it means lots of tech companies, for others it means wide open spaces.

Land is eventually limited, but there is tons of variety available of what people like.

> I currently believe that period of time (mid-20th century, esp. in the US) was a historical anomaly set up by a fairly unique set of circumstances

It was a period of high taxation on the highest incomes, large social welfare programs and by relative terms fairly low income inequality.

Finally productivity gains benefitted the labor class and not just the capital class.

That all halted in the 70s and 80s.

And the largest part of those salaries is again used to pay for housing.

Roughly one thirds of American incomes go to cover housing. Cheaper housing means lower wage demands.

Don't want to underestimate how disastrous this could be for other insects. Even ignoring the impact on them, the impact on our needs to maintain pollinator populations.

Interesting paper - it will fascinating to see if it pans out.

The one thing I didn't see that would be good is some validation that the architecture(s) that perform best on large models are the same architectures that perform best on small models.

Ie validation the assumption that you can use small models with sma amounts of training/compute to determine the best architecture for large models and high training budgets.

Even if it doesn't translate it would still be very cool to be able to qui kly evolve better small models (1M to 400M params), but I believe the implied goal (and what everyone wants) is that this exploration and discovery of novel architectures would be applicable for the really big models as well.

If you could only ai discover larger models by spending OpenAi/Anthropic/... budgets per exploration then we're not really gaining much in terms of novel ideas as the cost (time and budget) would be too prohibitive.

> There seems to be only downsides. Startup Fails , you loose - Gets acquired - you loose What is the motivation to perform

You get to make a nice payout for a VC. And isn't that all of our life goals?

> but it's one of those tricky real-world problems where you'll only get yourself into trouble if you only follow the model's suggestions.

What's the goal of this? What are you looking for?

> Crypto didn't do that. Investment bros did.

So why didn't crypto block or ban them from doing these scams using their technology?

Unless you're saying crypto created the problem at this scale and can do nothing to stop the problem it created...?

Not to pick on you, but you stared your pay with a series of trivia steak arguments - but all examples of things I see repeated with little consideration.

> Yes, but companies run for investors will always have to pay investors, so it would have either been a buyback or dividends.

Your argument here is that the pivot two things companies can spend money on is dividends or stock buybacks? An incredibly false choice. They can also clearly invest in staying competitive.

> When interest rates go up, companies end up competing with government bonds, and they must give dividends that keep them competitive,

Why do you think they must do this? These aren't bonds - people aren't looking to park money at a risk free rate. They are stocks - different investments with different expectations. Setting the expectation of one based on the other is false. They should be ensuring the can keep with with short term growth of more than 4% per year non adjusted, which should be easy in an inflationary environment.

And let's dig further into what you said about this "must". Must implies no other reasonable option - the consequences of anything else would be to negative. So when you say they must do buy backs or higher dividends, I ask "or else what?". Simply put, what hairband is they don't? The company doesn't fold. Consumers don't suddenly stop buying their products. What actually happens that they need to avoid?

Not much really much of note really. They say/signal "instead of giving you this money right now, we're going to invest it to remain competitive and more productive in the future".

If some investors don't believe them or doubt want to wait the stock price dips for some period of time and then rebounds as those investments become realized with increased efficiency and superior products.

> For investing in the future to be possible interest rates must essentially be low.

Finally this is wrong. It doesn't require low interest rates is you're flush with cash. If you've given away all of your cash then yes you need to borrow to invest. But if you have tons of cash you don't need to borrow to have cash.

More info on "The Expression Problem" https://en.wikipedia.org/wiki/Expression_problem

> But there are customers who are looking for tractor wheels and flywheels. Maybe not you or people you know, but there's definitely large industries for both. Just like there's a market for "vehicle to get to work". Same thing exists in software. Linux caters to all the uses, from low power embedded systems to huge high performance computing systems and everything in between.

And that's exactly the problem. Linux is so focused on ensuring it works for every possible user out there, that it never makes the tough choices to be optimized for the 80% of users.

80% of people drive cars and need simple care wheels and tires.

But they're forced to think about 100 different types of wheels before they can go anywhere. My mom has never had to consider a flywheel in her life, and yet if her car were linux she's need to think about that, and everything from low power, high power, LED, gas powered, and oil powered headlights. She'd need to consider 50 types of seats, 7 seatbelt configurations,

Linux is what you get when engineers own a product and there is no PM around to say 'No' and force you to make the hard decisions. It's good for what it is, but nobody is making it actually easy. There are just "complex" and "less complex" flavors of linux. None are actually consumer level of simple - they're just "simple" compared to the other variants of linux.

Chrome OS is actually consumer level simple.

The better a product works, the more often you opt to use it. The better it finds needles in haystacks, the more often you choose to use it. Higher engagement is really rarely bad for a product outside of particular odd cases.

> Why should I, as a homeowner, be forced to pay the power companies...?

You keep making this disingenuous argument all over this thread.

If you have solar and use their grid to distribute it you pay to use their grid. If you want solar off the grid and want no distribution you pay nothing to them.

It's really that simple. There is production and distribution. If you are using their distribution you pay them for it. Not to mention for the stability of still getting power when your solar is insufficient...

The fact that the customer just needs a vehicle to get to and from work, but they're offered tractor wheels and 50k RPM flywheels is exactly the problem.

Stop inventing wheels, stop offering esoteric wheels and start solving customer problems simply.

It's not a team vs individually, it's specifically a team/duo with similar or same model vs a team/duo with different models. The benefit is seen by having the models be different. Each finds unique things and enhances the other.

> There must be a valid determination made that a person entered illegally for expedited deportation to apply.

Are you arguing about the way it should be or making a statement about the way things currently are?

If it's that easy to separate from natural mercury then it seems like they could make a fortune just separating it and selling the separated mercury.

Something isn't adding up

> What you're describing is called: programming.

Is that the part of programming that you enjoy? Remembering logger vs logging?

For me I enjoyed the technical chalenges, the design, solving customer problems all of that.

But in the end, focus on the parts you love.

> numbers in the response are real.

OpenAI very well may have a bug, but I'm not clear on this part. How do you know the numbers are real?

I understand you know the name is the company is real, but how do you know the numbers are real?

It's way may than anyone should need to do, but the only way I can see someone knowing this is contacting the owners is the company.

> The IT industry is also full of salesmen and con men, both enjoy unrealistic exaggeration. Your statements would not be out of place 20 years ago when the iPhone dropped. Your statements would not be out of place 3 years ago before every NFT went to 0. LLMs could hit an unsolvably hard wall next year and settle into a niche of utility.

The iPhone and subsequent growth of mobile (and the associated growth of social media which is really only possible in is current form with ubiquitous mobile computing) are evidence it did change everything. Society has been reshaped by mobile/iPhone and its consequences.

NFTs were never anything, and there was never an argument they were. The were a financial speculative item, and it was clear all the hype was due to greater fools and FOMO. To equate those two is silly. That's like arguingsome movie blockbuster like Avengers Endgame was going to "change everything" because it was talked about and advertised. It was always just a single piece of entertainment.

Finally for LLMs, a better comparison for them would have been the 80's AI winter. The question should be "why will this time not be like then?" And the answer is simple: If LLMs and generative AI models never improve an ounce - If they never solve another problem, nor get more efficient, nor get cheaper - they will still drastically change society because they are already good enough today. They are doing so now.

Advertising, software engineering, video making. The tech is already for enough that it is changing all of these fields. The only thing happening now is the time it takes for idea diffusion. People learning new things and applying it are the slow part of the loop.

You could have made your argument pre-chatgpt, and possibly could have made that argument in the window of the following year or two, but at this point the tech at the level to change society exists, it just needs time to spread. All it need are two things: tech stays the same, prices roughly stay the same. (No improvements required)

Now there still is a perfectly valid argument to make against the more extreme claims we hear of: all work being replaced..., and that stuff. And I'm as poorly equipped to predict that future as you (or anyone else) so won't weigh in - but that's not the bar for huge societal change.

The tech is already bigger than the iPhone. I think it is equivalent to social media, (mainly because I think most people still really underestimate how enormous the long term impact of social media will be in society: Politics, mental health, extremism, addiction. All things they existed before but now are "frictionless" to obtain. But that's for some other post...).

The question in my mind is will it be as impactful as the internet? But it doesn't have to be. Anything between social media and internet level of impact is society changing. And the tech today is already there, it just needs time to diffuse into society.

You're looking at Facebook after introducing the algorithm for engagement. It doesn't matter that society wasn't different overnight, the groundwork had been laid.

Can you please stop shilling for this?

HN comments aren't an ad network...