atleastoptimal

An unintelligible but correct proof is better than no proof. These first AI proofs may be overly complex and un-elegant, but they are the worst that frontier math proofs will ever be. AI math in 2030 will be leaps and bounds ahead of humans both in rigor and elegance.

If humans have nothing to contribute then shared understanding is a pointless endeavor. It makes sense now in the "centaur" period where human + AI > AI alone, but when AI mathematicians are both more rigorous and more elegant, then taking the time dumbing down their proofs to a human level of understanding is like requiring that we ensure all our current proofs be understandable by a monkey.

I don't mean currently, I mean in the future when an AI system exceeds all human abilities in that domain.

> If you automated all of those things, there would be so much room for disruption that the AI wouldn't be able to maintain dominance.

I think the skill "being able to identify and exploit areas ripe for disruption" is also a skill AI could hill-climb to become better than humans at. Being an entrepreneur is a mixture of domain knowledge, perseverance, intelligence, sales, and execution, and there is nothing keeping an embodied AI system, or even a trivially implemented AI-led system with ceremonial human representatives (for shaking hands when deals are made) from dominating in those domains and thus the entire entrepreneurial space.

> I'm guessing the people who successfully block construction will be happy the data centers are built elsewhere.

Depends on their goals. Do they want to slow down AI altogether, or just hate the concept of the building itself?

The main issue with wanting to block data centers in any specific location, is the hyperscalers will just move to building them wherever politics are most favorable. If protesters in your blue state block construction, they'll just build more in red states, giving the red states more leverage in how AI progress continues. If they are blocked across the board in the US, more will be build abroad, in countries that would never give credence to anti-datacenter movements that would stall economic growth.

The best option is to get everyone on the same page about AI benefits and hazards, and not be unrealistic about the way AI could shape up as a transformative element of human society.

Basically any "brand name" product which isn't either family owned or privately run by an extremely idealistic founder/CEO ends up getting predictably worse over time. Once some private equity fund buys you out and determines you are not on the quality/price Pareto frontier and can manage to lose only .8% of your revenue making your product .9% cheaper, the die is already cast.

There is a persistent bias, in general and online, against any claim that one's job prospects, social status, overall life satisfaction is dependent on the way people treat you specifically due to your physical appearance. I think this bias has worsened in the recent decade due to how emphatically the affirmative claim is espoused by the incel/looksmaxxer movement, so conceding any ground means "proving them right".

There are certain realities people feel more comfortable denying/de-emphasizing. Another such bias is the impact of innate talent and its connection to success in any field tied to cognitive ability. People don't seem to want to accept that there are differences in ability/faculties that are:

1. Extremely impactful on one's life outcome

2. Extremely hard or impossible to change, especially past adolescence

3. Largely predetermined by genetic differences, many of which are random

It seems like everything will follow this pattern:

1. AI is dismissed because an expert in a particular field finds an outdated model's outputs sub-par

2. New model, released or unreleased, makes a major stride in that field

3. Expert either recants and becomes AI-pilled, or claims it is just an artifact of the broad search space available to AI, and "no new knowledge was created".

Software slop has existed for as long as software has existed. AI is now a better coder than the median. By the end of next year it'll be better than the best human coders.

This is true, but likely happening in talk therapy too with overly sycyphantic therapists

I think there's one weirdly simple reason DeepMind isn't doing as well as OpenAI and Anthropic. I may be wrong on this.

OpenAI and Anthropic went from tiny startups to huge companies. As a consequence the stock options/RSU's offered to the employees paid off a far higher percentage ROI than any stock options a DeepMind (and thus Google) employee would get (since Google is already huge). This disincentivizes people who truly believe in the economically transformative power of AI to work at Google since their benefits will be capped by Google being large + having public company obligations.

It's also labor costs. It's far cheaper to get a crew, rent out space, etc in Eastern Europe, the US South, even the midwest, which is what led to films being made there. Government/tax incentives too of course.

This works better for math because math is self-verifiable. Once you have a proof it needs no outside evidence.

Expertise is needed to evaluate model outputs where it can't verify itself, or at the very least one's expertise can help steer the model in the right direction.

However this is irrelevant if models themselves are better at evaluating/leveraging expertise/information.

Obviously not, have you seen the 10 mathematical advances OpenAI found with Astra?

https://openai.com/index/ten-advances-in-mathematics/

This seems inevitable for most industries in the US (car/general manufacturing)

1. City/region becomes a boom-town based on a new high profit-margin industry

2. As that region becomes more wealthy, labor increases in price, thus there is more incentive to seek labor/resources for that industry outside the city/region

3. City/region hollows out of middle-class level jobs for the industry, retains ceremonial labor and executive offices.

Do you also test for contamination/adulterants/heavy metals?

Going from 0 to 3000 weekly added users is an amazing accomplishment and you should always remember that you are capable of producing something people love using.

The problem is that the real world massively compounds on small advantages. Your competitor is 1% better per day at growth, that compounds to a 37x difference in a year. You could do everything perfect except one small issue, or operate in a space where your success is easily replaceable, and none of the effort will translate to real world results.

I would do the following

1. Seriously audit what you can do/insights you have that nobody else has

2. Translate that into a moat that cannot be replicated by anyone else

3. Translate that moat into growth/retention in a niche domain that you know is more important in a way others don't

80% of the GDP over the next two decades will be siphoning money out of baby boomers (medical care, retirement homes, luxury cruises) before their children see a cent of it. Coupled with AI taking everyone's job, the end result is neo-feudalism where familial dynasties call all the shots.

I really want to know what people specifically mean when they allude to some "higher" level of cognition that humans have above LLM's when it comes to useful outputs. I don't mean the nature of consciousness or internal experience (which is unfalsifiable), I mean the essence of cognitive labor as a tool: writing is about relating concepts, drawing comparisons, aggregating elements of reality into a fictive but believable narrative.

All these processes can be reduced to mathematics and computation. The human brain is just a tool, like any other organ. Do we deny this reality because it shields us from existential horror?

> Location Though the team has spent many years scaling remote-first companies and is currently remote, we are hiring specifically in Los Angeles, London, and New York, with an expectation for a flexible in-office schedule as we build our presence in each region.

I always get a "lifestyle business" vibe when tech companies are based in LA or NY. When you're a seasoned veteran you'd prefer to live in a larger, more fun city, even though tech is concentrated in the Bay Area

So many otherwise smart engineers, overrepresented on HN, have this weird mental block where they fail to take AI progress seriously. I think it's a fear that their status-defining skill could be commodified, so they adopt a mental framework that aims to dismantle the seeming legitimacy of any genuine leap in AI progress as some sort of mirage

This is true, which is why the ideal things to do now are actions which increase your moat, or return on model improvement in the future (for example, accumulating data which future better models could more effectively capitalize on)

Closed-source models have to deal with the current frontier being heavily regulated. Fable, at its old level, was "too good" to be released and they had to add an additional safety layer to sanitize the outputs. Lowering the quality of the models so they are safer and more steerable has been something all the closed-source models have been doing for a while, a requirement that many open source models don't need to deal with.

If Kimi k3 really were above Fable 5 then there invariably the USG would have to consider their restrictions on model capabilities excessive, or one would have to admin closed source models are held to more restrictive safety standards than open source models.

>Although I will reiterate the fact that distillation is not the primary reason why these models are performing so competitively.

How would you know this? How could you ascertain exactly how much performance is attributable to their unique engineering/research? If they really were so competitive they could surely make a model that isn't dependent on distilling Fable or other frontier models.

It matters because everyone imagines the inevitable "closing of the gap" between closed and open source, but the rate at which open source catches up with closed source seems to depend on being able to train on and distill the outputs of open source models. As long as performance of open source models is at least partially dependent on frontier-model outputs, then that gap will remain in place by definition.

>Claiming in any shape or form that fable disillation is one of the primary reasons why kimi k3 is so competitive is slandering the work of other labs that cooperatively push the open-source models forward.

If the distillation is irrelevant to why it is competitive, why do they do it then? Obviously is helps improve their benchmarks/performance to some degree, otherwise they wouldn't need to do it.

The VC money is only contingent on the strategy eventually bearing fruit. I do imagine going open source -> closed source could work for some model companies who get enterprise/ecosystem buy-in but the probability of ROI is lower.

AI models cost tens of millions to train. Offering them for free won’t justify the upfront costs.

The Chinese model of model training/open sourcing only makes sense in the context of the overall strategy of undercutting American frontier labs’ profit margins.

You went through 5 million tokens in an hour? You would be significantly benefitted by using fable as an orchestrator for other models

I wonder what percentage of the US GDP is "Fooling old people"

No matter how bad the climate crisis gets, those who have the power to change it (heads of state/capital) will never abandon their lifestyles which expend 1000x more energy and fossil fuels than the average person. Pushing for voluntary, individual reductions in consumption is pointless and insulting to the average person.

Do people actually believe that the revenue for these companies is "fake"? Do they think financial analysts don't have anything but revenue to look at to determine the substance of their growth?