Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

2025: The Year in LLMs

simonwillison.net

Post reply on HN
Print view View original

371–380 of 643 posts

Re: 2025: The Year in LLMs

#371

Post by andy99 »

weatherlite wrote: post #345
XenophileJKO wrote: post #265

Earlier quoted context omitted.

That is like saying training tensorflow models is just calling some APIs. Actually making a system like this work seems easy, but isn't really. (Though with the CURRENT generation or two of models it has gotten "pretty easy" I think. Before that, not so much.)

No idea about training tenserflow models - is it super complex or is it just calling a couple of APIs ? Langchain is literally calling an API. Maybe you need to get good with prompting or whatever, but I don't see where the complexity lies. Please let me know.

Having used both Tensorflow (though I expect they mean PyTorch which is way more popular, and I have also used) and langchain, they are nothing alike.

They he ML frameworks are much closer to implementing the mathematics of neural networks, with some abstractions but much closer to the linear algebra level. It requires an understanding of the underlying theory.

Langchain is a suite of convenience functions for composing prompts to LLMs. I wouldn’t consider there to be some real domain knowledge one would need to use it. There is a learning curve but it’s about learning the different components rather than learning a whole new academic discipline.

Re: 2025: The Year in LLMs

#372

Post by windexh8er »

Kerrick wrote: post #231
windexh8er wrote: post #221

Earlier quoted context omitted.

> They are helping their users create things that didn't exist before. That is a derived output. That isn't new as in: novel. It may be unique but it is derived from training data. LLMs legitimately cannot think and thus they cannot create in that way.

That is a pedantic distinction. You can create something that didn't exist by combining two things that did exist, in a way of combining things that already existed. For example, you could use a blender to combine almond butter and sawdust. While this may not be "novel", and it may be derived from existing materials and methods, you may still lay claim to having created something that didn't exist before. For a more…

> That is a pedantic distinction. You can create something that didn't exist by combining two things that did exist, in a way of combining things that already existed.

This is the definition of a derived product. Call it a derivative work if we're being pedantic and, regardless, is not any level of proof that LLMs "think".

Re: 2025: The Year in LLMs

#373

Post by arcatech »

jillesvangurp wrote: post #287
ksec wrote: post #228

All these improvement in a single year, 2025. While this may seem obvious to those who follows along the AI / LLM news. It may be worth pointing out again ChatGPT was introduced to us in November 2022. I still dont believe AGI, ASI or Whatever AI will take over human in short period of time say 10 - 20 years. But it is hard to argue against the value of current AI, which many of the vocal critics on HN seems to have…

2025 was the year of development tool using AI agents. I think we'll shift attention to non development tool using AI agents. Most business users are still stuck using chat gpt as some kind of grand oracle that will write their email or powerpoint slides. There are bits and pieces of mostly technology demo level solutions but nothing that is widely used like AI coding tools are so far. I don't think this is bottle ne…

> Some examples of random crap I deal with: figuring out contracts, their meaning/implication to situations, and deciding on a course of action

This doesn’t sound like bullshit you should hand off to an AI. It sounds like stuff you would care about.

Re: 2025: The Year in LLMs

#374

Post by windexh8er »

nl wrote: post #328
windexh8er wrote: post #221

Earlier quoted context omitted.

> They are helping their users create things that didn't exist before. That is a derived output. That isn't new as in: novel. It may be unique but it is derived from training data. LLMs legitimately cannot think and thus they cannot create in that way.

What does "think" mean? Why is that kind of thinking required to create novel works? Randomness can create novelty. Mistakes can be novel. There are many ways to create novelty. Also I think you might not know how LLMs are trained to code. Pre-training gives them some idea of the syntax etc but that only gets you to fancy autocomplete. Modern LLMs are heavily trained using reinforcement data which is custom task the…

[flagged]

Re: 2025: The Year in LLMs

#375

Post by arcatech »

ordersofmag wrote: post #357
windexh8er wrote: post #221

Earlier quoted context omitted.

> They are helping their users create things that didn't exist before. That is a derived output. That isn't new as in: novel. It may be unique but it is derived from training data. LLMs legitimately cannot think and thus they cannot create in that way.

I will find this often-repeated argument compelling only when someone can prove to me that the human mind works in a way that isn't 'combining stuff it learned in the past'. 5 years ago a typical argument against AGI was that computers would never be able to think because "real thinking" involved mastery of language which was something clearly beyond what computers would ever be able to do. The implication was that t…

> I will find this often-repeated argument compelling only when someone can prove to me that the human mind works in a way that isn't 'combining stuff it learned in the past'.

This is the definition of the word ‘novel’.

Re: 2025: The Year in LLMs

#376

Post by simonw »

rvz wrote: post #340
minimaxir wrote: post #211

Earlier quoted context omitted.

The operative word there is "primarily". Simon comments on a variety of topics and has far more interactions that don't link to his blog than do. Simon's posts are not engagement farming by any definition of the term. He posts good content frequently which is then upvoted by the Hacker News community, which should be the ideal for a Hacker News contributor.

Except that the "content" that reaches the top is always about AI / LLMs and nothing else and it is "all the time". Any opportunity to comment, he will link back to his own blog. He even reposted the same link (which is about AI) with one of his posts when the upvotes fell off and until the second one reached the top, with the intention of promoting his own blog. Let me simply prove my point to you on how predictable…

"He even reposted the same link (which is about AI) with one of his posts when the upvotes fell off"

Where did I do that?

> He will do a blog post this month about this paper [0]

That paper you linked to is a perfect example of where my approach can add value!

Did you read it? Do you understand what it saying? It is dense.

I would love to read an evaluation of that paper by someone who can rephrase the core ideas and conversations into a couple of paragraphs that help me understand it, and help me figure out if I should invest further effort in learning more.

I have a whole tag on my blog for that kind of content called paper-review: https://simonwillison.net/tags/paper-review/ - it's my version of the TikTok meme "I read X so you don't have to".

Honestly, your problem doesn't seem to be with me so much as it seems to be with the concept of blogging in general.

Re: 2025: The Year in LLMs

#377

Post by nrclark »

arcatech wrote: post #373
jillesvangurp wrote: post #287

Earlier quoted context omitted.

2025 was the year of development tool using AI agents. I think we'll shift attention to non development tool using AI agents. Most business users are still stuck using chat gpt as some kind of grand oracle that will write their email or powerpoint slides. There are bits and pieces of mostly technology demo level solutions but nothing that is widely used like AI coding tools are so far. I don't think this is bottle ne…

> Some examples of random crap I deal with: figuring out contracts, their meaning/implication to situations, and deciding on a course of action This doesn’t sound like bullshit you should hand off to an AI. It sounds like stuff you would care about.

Agree. Even asking it can anchor your thinking.

Re: 2025: The Year in LLMs

#378

Post by derwiki »

steveBK123 wrote: post #362
pjc50 wrote: post #289

Earlier quoted context omitted.

Investing a trillion dollars for a revenue of a billion dollars doesn't sound great yet.

Indeed, its the old Uber playbook at nearly two extra orders of magnitude. It is a large enough number to simply run out of private capital to consume before it turns cash flow positive. Lots of things sell well if sold at such a loss. I’d take a new Ferrari for $2500 if it was on offer.

Uber’s playbook worked for Uber

Re: 2025: The Year in LLMs

#379

Post by windexh8er »

closewith wrote: post #343
windexh8er wrote: post #221

Earlier quoted context omitted.

> They are helping their users create things that didn't exist before. That is a derived output. That isn't new as in: novel. It may be unique but it is derived from training data. LLMs legitimately cannot think and thus they cannot create in that way.

By that definition, nearly all commercial software development (and nearly all human output in general) is derived output.

Wow.

You’re using ‘derived’ to imply ‘therefore equivalent.’ That’s a category error. A cookbook is derived from food culture. Does an LLM taste food? Can it think about how good that cookie tastes?

A flight simulator is derived from aerodynamics - yet it doesn’t fly.

Likewise, text that resembles reasoning isn’t the same thing as a system that has beliefs, intentions, or understanding. Humans do. LLMs don't.

Also... Ask an LLM what's the difference between a human brain and an LLM. If an LLM could "think" it wouldn't give you the answer it just did.

Re: 2025: The Year in LLMs

#380

Post by wpietri »

apexalpha wrote: post #356
wpietri wrote: post #354

Earlier quoted context omitted.

This is not a great argument: > But it is hard to argue against the value of current AI [...] it is getting $1B dollar runway already. The psychic services industry makes over $2 billion a year in the US [1], with about a quarter of the population being actual believers. [2]. [1] The https://www.ibisworld.com/united-states/industry/psychic-ser... [2] https://news.gallup.com/poll/692738/paranormal-phenomena-met...

What if these provide actual value through placebo-effect?

I think we have different definitions of "actual value". But even if I pick the flaccid definition, that isn't proof of value of the thing itself, but of any placebo. In which case we can focus on the cheapest/least harmful placebo. Or, better, solving the underlying problem that the placebo "helps".
Post reply on HN