root_axis

I think the points you raise are reasonable signals to consider, but I don't think they show the author being "consistently wrong". The overall thesis still remains plausible even though we have seen LLMs continue to improve.

> - hallucinations are dramatically less of a problem

Sure, but it remains a big enough problem that human intervention and review is still necessary for any serious work across all use cases and industries.

> - several mass market use cases have emerged, most notably coding

Coding seems to be the only one, but there are still a lot of open questions about how the market can sustain the costs, and that's without considering the market dynamics that could emerge once costs are lowered enough that open source models start to become an attractive option.

> - rate of progress has increased

Debatable.

Can you point to anything specific from the article that you'd describe as consistently wrong? Not disagreeing with you, but nothing popped out to me after skimming the article.

lol, not sure if this is a joke, but Teslas do not have anything close to the necessary hardware to run grok locally.

The use of the term "understanding" in the quote you mentioned is a claim about metaphysics, not cognitive capacity.

The phenomenological quality of subjective experience is never described as "cognitive capacity".

That term is used to describe mental aptitude or skills, like the ability to learn new languages or do math.

> They're made out of weights" describes why the weight-based construction of neural networks should impact the way that you think about them and their outputs.

How do you connect that description to "LLMs could not possibly be good models of some cognitive capacity"?

None of those arguments claim "LLMs could not possibly be good models of some cognitive capacity"

Interesting how "the worst candidate in history" is not to blame.

> Putting the total addressable market at around 67T if all of us spend USD 1.5k on tokens every month

However, that's an absurd scenario.

Please link one of these top ranked posts. Before you do, be aware that I'm going to read what it says and assess if it meets the description of the argument as claimed.

> didn't even come to $600.

That's still in the ballpark. A modest change in your usage habits or workload could easily get you there.

> In the present context, the fallacy manifests in claims that LLMs could not possibly be good models of some cognitive capacity because their operations merely consist in a collection of statistical calculations, or linear algebra operations, or next-token predictions

Nobody actually makes this argument though.

It's the opposite, "weather engineering" is a veneer of legitimacy layered over crackpot nonsense.

Yeah... Take it even one step further, why would AI companies even sell access at all? They could just run competitor firms directly themselves, and that outcome is really the only way they can ever hope to make ROI, so for sure they'll do it if it ever becomes possible.

And accelerating the schedules for insiders to dump shares.

Even if quantum computing had any clear implications for LLMs (it doesn't), there is no such thing as a "consumer quantum computer" and there won't be in our lifetimes.

It's not really a pivot though, is it? As I understand it, the company went bust and sold their branding to a different company that wants to do AI infra.

The outcome of his election was a referendum on Trump's performance among engaged Republican voters.

> me and my wife have been in the US for combined 25+ years, and paid over $100,000 in

Sounds like you may be a good candidate for Trump's gold card.

I'm being fecitious of course, but I'm just pointing out that thinking of citizenship worthiness in monetary terms is something the president has already considered.

> voters didn't want this

Yes they did. Of course they didn't want to be targeted themselves, but the rhetoric was very explicit about what would happen, and they already had a preview of it in 2016 and voted even more favorably for this regime this time around.

> The media silently accepted Trump's lie at face value when he said he knew nothing about Project 2025

Not true. The media was very vocal about it, and it was obvious that he was on board with it.

> Most Republican voters, when told about the actual policies being implemented by elected Republicans, don't believe the reports, and assume that nobody would be enacting such stupid policy.

This isn't true. The recent ouster of Thomas Massie is a clear example of this. However, even if that were true, Republican voters still overwhelmingly prefer this to the alternative (Democrats), and polls show this today.

> Yet the voters keep voting for them.

Indeed. Not sure how you can acknowledge this but somehow believe it's not what the voters want.

It's definitely on the high end. Besides the fact that most startup equity ends up being worthless, you can't wait on a four year cliff to pay your rent.

People hate to be reminded of this, but that "faction" is the voters, in record numbers for the party.

Based on the comments I think a lot of people assume the headline pertains to Bun itself.

That's just the political division. Scientists and academic types tend to lean left, the republicans are a right leaning party so they oppose them. It's not even a Trump thing, it's been like this for decades, though Trump is obviously more aggressive than previous Republican leaders.

Unfortunately, there's really no clear path to viable local models for the common folk, the hardware requirements are just too extreme. I say this as someone with a pair of A100s that is absolutely delighted by what the open source models are capable of, but even with the best harnesses, tiny quantized models are just not even close to the same league as something like Kimi-k2.

Of course, here on HN it's easy to find folks who get a lot out of tinkering with tiny models, but the masses don't want to tinker with toys, they want something fast with a large context and approximating at least Opus 4.6 level reliability and capability, which simply can't be squeezed into a quantized 60b model.

> These companies are selling subscriptions for what is basically AGI

This type of rhetoric is exactly why people think it's a scam

You spent 50k for plex hosting? Why so expensive?

A properly designed benchmark won't use tests that leave room for ambiguous interpretation.

Training a model that is larger than your competitor's.