Hacker News new | ask | show | jobs
by jstummbillig 780 days ago
I disagree.

a) A year after GPT-4 set the bar, it's still the best model, despite everyone else not having to do it first. Just copy, and just software. And that's not for lack of trying by every other viable prime player on the planet with unprecedented acceleration.

Imagine any other piece of software, where the incumbent has a mere 2-3 year head start, in which they had to work out the entire product that everyone else, despite just having to copy and pressing the pedal through the floor is struggling just trying to catch up with.

b) The current models including GPT-4 are so bad. The few billions can be made by just by continue playing this game of improvements for a few years and getting better each year. I think people are wildly confused about how big this market is going to be when that happens. They are not squeezing hosting or compute. They are squeezing intelligence. Intelligence is the entire economy. The notion that there would ever not be room for multiple things here, maybe through size or specialisation or cost (as with all other intelligence), and that a few billion dollar are a big deal, is so strange to me.

c) The game will at some point, be mostly about infra and optimization. People come to the conclusion that's a problem for the incumbents, when our entire industry is mostly about infra and optimization. AWS is infra and optimization. I think even the average hn tinkerer understands that therein lies a proposition that's not exactly equivalent to "just rent a few servers and do it yourself".

1 comments

> A year after GPT-4 set the bar, it's still the best model

Debatable. Many people find Claude Opus superior, and I know I've found it consistently better for challenging coding questions. More importantly, the delta between GPT-4 and everything else is getting smaller and smaller. Llama 3 is basically interchangeable with GPT-4 for a huge number of tasks, despite its smaller size.

> Many people find Claude Opus superior

Many more do not, according to the LMSYS leaderboard.

> Llama 3 is basically interchangeable with GPT-4 for a huge number of tasks

Sure. I am sure the number approaches infinity, if you are willing to let the model inform the task. That's usually not what most people are looking for in a tool.

GPT-4 was released in March 2023.

Which means the research that went into it would've been finalised quite some time prior.

Meaning that you're getting close to a 2 year head start.

While they still call it GPT-4, the one topping the rankings are newer iterations of it despite still retaining the same name. The latest one is from 2024-04-09. Sure that one probably finished training a few months ago but it is by no means a 2 year head start.
Agree, the delta is getting smaller. And for majority of the tasks you can use the Claude Sonet which is better than 3.5 and also fast.

But at the same time when you actually want to solve a complicated problem, deep down you know that only GPT 4 can crack it.

Even more important, you know that GPT-4 will probably also not crack it. Which is why the SOTA is not terribly interesting. The delta between GPT-4 and the competition has been closing but why anyone would assume that this is a trend and that it would continue with GPT-4.5 to competition, or GPT-5 to competition instead of the other way around is a mystery to me.

I am not saying it could not be true. But extrapolating from differences between current bad models to a future with better models is weird, specially when everyone seems to pretty much agree that scale is the difference between the two and scale is hard and exclusive.

There’s a scatterplot that’s been circulating on Twitter. The trend lines show that since the time of GPT-2, open weights models have improved at a steeper rate than proprietary models, with the two on a path to intersect.
I would argue that's to be expected after the first generally accepted POC (GPT-3.5) was released, with it an entire industry created, and other companies actually started copying/competing in a big way.

It seems a stretch to read this as a continuing trend, when (from what I gather everyone agrees on) the way to better models seems to be ever more efficient handling of ever larger amounts of money, compute and data, with no reasonable limits in sight on any of the three.

Scaling up LLMs is only going to go so far, and it will yield diminishing marginal returns on all of that money, compute, and data. It’s a regime of exponential increases in inputs for linear gains in the outputs - barring some technological breakthroughs which could come from anywhere, not just from OpenAI.