ryeights

This is misleading. In order to read traffic you would need to swap out the end-entity TLS certs being used. Phones and browsers (at a minimum) enforce certificate transparency, so connections would fail.

We have to err on the side of humanity. Wear and tear and some bending of rules is part of the human condition and should be considered the cost of doing business in the STR space. I believe increasing neurosis, anxiety, and social dysfunction can be explained by increasing systemization, surveillance and rigidity in the world.

This is a microcosm of everything that has gone wrong in modern technological society. Most Airbnbs have occupancy rules that bar any daytime visitors if read strictly. This has been normalized under the understanding that the policy isn’t strictly enforceable, but exists as a backstop for situations where extra visitors cause damage/nuisances.

So if you want to have a (perhaps spontaneous) guest over or a small polite gathering, as long as you don't cause a disturbance, it's no harm no foul. This is beneficial to the social fabric and doesn't bring any detriment to the host or neighbors. Yet, thanks to the marvels of technology, we can now detect and deter these aberrances that would have otherwise gone unnoticed!

Meat proxy at work, meat proxy in your personal time. The utopian visions of AI futures strike me as alternative forms of hell. Humanity accomplishes everything and it means nothing, actions have no real consequence, the bumpy friction and individuality of life smoothed to a plane of perfect optimization.

As soon as we stop posting lazy LLM reductionism that has been clearly falsified by events

Zitronian dreck. It’s possible to be concerned with multiple aspects of a technology at once

International control of nuclear arms faced similarly tall odds. So why is it impossible for AI?

All of the world’s brightest mathematicians have clearly been slacking on the job—turns out solving a Millennium problem doesn’t require any thought at all!

i.e., we're entering RSI and LLMs will help to build their more capable successors under a new paradigm? Fair point, but most people saying #2 surely mean to say LLM development has been a dead end and hasn't gotten us substantially closer to AGI

Books, writing, comments, etc on cocktails are certainly in the training set. The model doesn’t need a sensation of taste in order to understand what humans enjoy. RLVR specifically on cocktails is not needed as the general intelligence of the models increases

Emergent capabilities, bitter lesson, increasing general intelligence, etc

But an LLM provider could very easily add a "Jev mode" to any existing model, right? LLMs already produce a probability distribution over arbitrary classes. Just tell e.g. 5.6 Luna “here is the user's question, you must respond ONLY with the words 'foo', 'bar', or 'baz',” run a single forward pass of the model, and report the normalized probabilities of 'foo' 'bar' and 'baz' tokens before the first output.

With such an approach you could even retain full reasoning capability

The Chinese models talk Claudlish too

Proper time asymptotically approaches zero here right?