root_axis

If you imagine the latent space as a map and the prompt as a sequence of directions towards clusters of knowledge, it makes sense that alignment can cull "pathways" through the latent space that emerged during pretraining.

> There's only a finite number of bugs and vulnerabilities.

The context of an LLM is also finite.

Vulnerabilities are perpetually being created, and this will be true no matter how good LLMs become at writing code - there's simply too many factors that can contribute to something apparently benign becoming dangerous.

Access to Kuwait's market is far more valuable than anything Elon or Bezos has to say about how meta operates its business.

Not really. If run away AI ever became a real thing, the stock market would be completely destroyed since most companies would become worthless.

Way too expensive.

I've actually seen this lead to serious issues when a zoom LLM summary attributed statements to someone who didn't say them.

Someone else who couldn't attend the meeting later read that summary and it created a major argument because the topic had been a sore subject for this person due to an ongoing debate at the company. Everyone who attended the meeting confirmed it was an error, but the coincidental timing made it hard for him to accept, because the LLMs summary presented things in a way that validated this person's concerns that had been previously minimized by some folks on that meeting.

The drama got heated to the point where management produced a policy about not trusting generative output without independent verification. Seems at least it was a lesson learned.

Hilarious to see people predicting a singularity when 40% of the u.s. economy can barley keep the LLMs online to complete mundane software tasks.

Sorry but you're just seeing what you want to see. The idea that a 31b model is anywhere even in the ballpark of something like Opus 4.5 is just absurd on its face.

Well $100k was a generous guesstimate for some time in the future where something like an Opus 4.7 is old news.

If we think about the near future, something like Kimi2.6 is within the realm of Opus 4.6 today, but requires closer to $700k in hardware to run.

That cannot run KimiK2.6 or GLM5.1 i.e models within the ballpark of anything offered by frontier companies.

> They're good enough for 95% of use cases

They're not at all, not even close. Especially when you consider the use cases for people who are paying for LLM services today.

I have two A100s and have been playing with local models for years. There's definitely moments where they are quite impressive, but small context sizes and unreliability become immediately obvious.

> For those of us a bit crazy, we are running KimiK2.6, GLM5.1

Yes, those can compare to Opus, but you can't run those unquantized for less than $400k in hardware.

> Won’t these H100s drop in price in a few years

Doubtful. The increase in demand is greatly outpacing supply, and all signs point to a continued acceleration in demand

> If I could drop $10,000 to have an effectively permanent opus 4.7 subscription today, I would.

lol well obviously, but realistically that price point is going to be closer to $100k, with a perpetual $1k a month in power costs.

I use Opus 4.6 as an example because it's the LLM that has been widely recognized by the public as being reliably capable of doing real work across many domains. However, the same logic applies to Opus 4.5 and even previous generations. These models have huge parameter counts and large context sizes, there's no training technique that can compensate for those qualities in small and quantized models.

You are greatly underestimating the hardware requirements for productive local LLMs. Research consistently shows that parameter count sets the practical ceiling for a model's reliability. Quantized models with double digit param counts will never be reliable enough to achieve results in the realm of something like Opus 4.6.

They care. They are very sensitive to their public image and most cannot tolerate criticism or disapproval.

> Also their feelings are not hurt by what you or I think, they don’t care.

Definitely not true, they tend to care more than most.

The AI argument doesn't make sense to me for layoffs. If AI is making the company more productive then there's an incredible opportunity to use the existing workforce to tackle the massive backlog of important work. A big layoff only makes sense if there is no more useful work to do or you're killing products.

These suspensions send the appropriate message. This isn't the same thing as poorly reviewed marketing copy, hallucinations in government policy papers are unacceptable.

Singularities are a sign that you have a broken model.

He's not using LOC as a metric, he's making an observation about the impact of a change in the typical volume of LOC.

They're making the bet that seniors won't be needed by then. I think it's a bad bet, but it makes sense to follow through if 40% of the economy is already being occupied by this tech.

Any confirmation that a genuine port is underway? This might just be an experiment.

Diagnosis is just a small part of a doctor's job. In this case, we're also talking about an ER, it's a very physical environment. Beyond that, a doctor is able to examine a patient in a manner that isn't feasible for machines any time in the foreseeable future.

More importantly, LLMs regularly hallucinate, so they cannot be relied upon without an expert to check for mistakes - it will be a regular occurrence that the LLM just states something that is obviously wrong, and society will not find it acceptable that their loved ones can die because of vibe medicine.

Like with software though, they are obviously a beneficial tool if used responsibly.

Actually I don't really have any problems with panpsychism. It's a pretty uncommon perspective, but when discussing conscious machines, it at least presents a consistent criteria for consciousness.

Based on that reasoning anything could be conscious. If that's a bullet you want to bite, fair enough.

They don't have senses because they don't have a body. It's just a program. Do weights on a hard drive have consciousness? Does my installation of starcraft have consciousness? It doesn't make any sense.

I agree, but I don't think determinism is a factor either way. Ultimately, if arbitrary computer programs can be conscious, then it stands to reason that many other arbitrarily complex systems in the universe should also be.

What makes the argument facile is that the singular focus on LLMs reveals an indulgence in the human tendency to anthropomorphize, rather than a reasoned perspective meant to classify the types of things in the universe which should be conscious and why LLMs should fall into that category.

That's just reductive semantics. Anything can be described as "nothing but data".

LLMs have no self, sensory experience, or experience of any kind. The idea doesn't even really make sense. Even if it did, the closest analogy to biological "experience" for an LLM would be the training process, since training at least vaguely resembles an environment where the model is receiving stimuli and reacting to it (i.e. human lived experience) - inference is just using the freeze-dried weights as a lookup table for token statistics. It's absurd to think that such a thing is conscious.