anothermathbozo

> I outpreform everyone around me

Could you help me understand what you mean here?

Every time someone puts precise language to what makes ai slop sloppy it feels like frontier labs get an ever clearer RL target. Not to say people shouldn’t publish these types of posts. I enjoy reading them and I hope they continue to be published.

> internal progress on mathematics

This alone feels like a horrible new paradigm

People get income from one of three places: capital income, labor income, or the welfare state. If this technology truly unlocks a holy panacea of productivity with a commensurate drop in employment then capital’s share of the national income can and should provide for a wider and deeper welfare state. Nothing new need be invented here. Anthropic’s elaborate scenario-building makes appropriate preparedness sound like a problem requiring heroic prognostication when a simple theory of distribution will suffice.

The benchmark does not reward producing more code.

Say more!

Can you say more about why you are resistant to unionization

Unions are not the primary mechanism between employers and workers for establishing fairer wages in many European countries. Unions are designed to level a power imbalance between an employer (typically a legal vehicle which aggregates the material self interest of various actors) and the employees who would otherwise have to act alone.

Myopically focusing on wages while ignoring the many other concerns about the distribution of power and legal rights is a common misunderstanding.

unions are institutions for bargaining power, even under decent employers

He killed Brooksley Born’s proposed derivatives regulations in the late 90’s.

You’re conflating firm level taxable income with the national income.

> A wide range of pro-employment policy incentives can help to slow or reduce job displacement, including: wage insurance policies that compensate people when they have to take a lower-paying job, retention tax incentives to encourage employers not to make layoffs, workforce training grants, or infrastructure to facilitate matching of employers to employees to speed the rate of labor market adaptation. While the particulars of which interventions are best will depend on what kind of labor displacement AI brings, we should readily accept the costs and market inefficiencies that these policies could entail, particularly as they are likely to be offset by AI-driven productivity gains.

People get income from one of three places: capital income, labor income, or the welfare state. If this technology truly unlocks a holy panacea of productivity with a commensurate drop in employment then capital’s share of the national income can and should provide for a wider and deeper welfare state. Nothing new need be invented here. Dario’s long and only somewhat organized list of policy interventions makes appropriate preparedness sound like a manic pulling of any and all levers when a simple theory of distribution will suffice.

Public sector unions, like all unions, are designed to level the power imbalance between worker and employer. Nothing about public/private employers changes this dynamic.

It’s always a crowd pleaser to be skeptical of ai development. Not sure what people feel like they are achieving for continually announcing they aren’t buying it when someone claims they’ve made effective use of these tools.

What is this a bubble on? What does said bubble collapsing look like?

Income tax is not the only revenue that the state and local governments capture when production increases.

child care policy frees labor capacity for work that is more likely to earn a slice of the national income. It’s almost certainly going to result in greater economic activity for the state. In the immediate it is funded from two existing funds.

Given how much capital has already been committed to infrastructure development I find it very unlikely that even a Black Monday style market correction will do much to alter the medium and long term outlook for ai development and investment.

Too many people talk about a bubble almost wish-casting that such a thing will make this all go away. It’s probably safest to assume your political enemies won’t be hoisted on their own petard anytime soon.

> The reality of what these tools can do is sinking in

It feels premature to make determinations about how far this emergent technology can be pushed.

What difference does that make at this stage of things?

Why would that be?

Warrantless and totally spiteful for you to make unqualified claims like “cognitive decline” from skimming two papers. This is shameful.

This isn't exactly the case. The trend is a log scale. So a 10x in pretraining should yield a 10% increase in performance. That's not proving to be false per say but rather they are encountering practical limitations around 10x'ing data volume and 10x'ing available compute.

Disappeared. He was kidnapped and disappeared.

Optical Character Recognition (OCR) and Handwritten Text Recognition (HTR) are different tasks

Anthropic’s CEO said their technology would end all disease and expand our lifespans to 200 years. What on earth do you mean they’re not playing the hype game?

Looks like o1 performance without reasoning. Pretty good but seems reasonable that they didn’t want to call this 5 as they’ve already got a product out there that is as performant.

Another notable thing here is a big drop in hallucination rate as measured by their benchmarks (for whatever those are worth).

The models answer was a page long essay about why the question wasn’t worth asking. The model demonstrated common sense by not engaging with this idiot chase of a hypothetical.

Almost this entire thread is criticism

Imagine someone can do the things you can’t do and needs help doing the things you can.