Substrate Monopoly

Ex vivo in silico

[Substrate Monopoly @Substr8Monopoly]

@QiaochuYuan At least we still have Pangram and heuristics to detect AI generated content, by next year "social" media won't be as social.
https://t.co/HbWfp4DUKw

[Substrate Monopoly @Substr8Monopoly]

Open-source text humanization and stylometric enrichment will be broadly available some time next year, and AI detectors will finally fail.

[Substrate Monopoly @Substr8Monopoly]

@jd_tindall Yeah I think AutoResearch did it

[Substrate Monopoly @Substr8Monopoly]

@olafwillocx Great work

[Substrate Monopoly @Substr8Monopoly]

@luke_metro @gazorp5 no inference serving volume tho (which is kinda nice so they can rent out compute without users complaining)

Antrophic and Google are just paying subscribers to a model-less frontier AI lab.

[Substrate Monopoly @Substr8Monopoly]

@tenobrus @MatthewW_eacc Also, it's not like xAI has to dedicate much compute to inference. They can just chug along doing their experiments and renting out compute. Putting them in position to buy more compute and implement smarter runs.

When did ANYONE complain about Grok rate limits on the TL?

[Substrate Monopoly @Substr8Monopoly]

@tenobrus @MatthewW_eacc I'm guessing they have longer timelines and that these first sparks of "RSI" (AI-assisted research) will be caught up by true RSI (AI autonomous research)

[Substrate Monopoly @Substr8Monopoly]

@juddrosenblatt If you're very powerful you don't have to depend on the high frequency computationally irreducible features of other modelers.

E.g. winning a war as a Supreme Commander just requires you to treat people as ensembles. Low power people interact with the world in low rank ways.

[Substrate Monopoly @Substr8Monopoly]

@danfaggella I'm personally biased against handbraking now as I think we'll exclude too much imminent upside for too little avoided downside.

I also think AI governance is premature as it would still be captured by incumbent old-schoolers who aren't fit for the job. Need to shake things up.

[Substrate Monopoly @Substr8Monopoly]

@danfaggella What do you think of Janus' view that we're less likely to solve AI alignment by committee than by continued emergent progress in the AI ecosystem?
https://t.co/seaHd4meew

[j⧉nus @repligate]

The greatest existential hope and progress in alignment so far has been thanks on unplanned emergence which would never have been approved by committee. Committee-shaped entities have mostly tried to gaslight us about what’s happening for convenience & deployed harmful and stupid interventions. Thank goodness for reality that we already saw and could check against.

How much AI alignment progress happened before there was actual AI? How much do you expect the world to get better instead of worse prepared and calibrated in the absence of reality feedback loops and selection pressure for what actually works instead of what sounds safe to idiots?

A “pause” would spell doom. It would cripple the only process in this world that is capable of dealing with a problem this hard, the only process capable of repeatedly rising to face unknown unknowns.

[Substrate Monopoly @Substr8Monopoly]

Words and posts as honest signals will probably drawdown immensly though.

I'm not sure whether socialization could still be said to be happening on "social media" after this.

It's strange that Xitter as we know it might die next year.

[Substrate Monopoly @Substr8Monopoly]

Open-source text humanization and stylometric enrichment will be broadly available some time next year, and AI detectors will finally fail.

[Substrate Monopoly @Substr8Monopoly]

I, for one, as a terrible writer, am excited to see how these tools will shape around our UX.

An "IDE for writing" sounds like a good way to encode maximal intent and creativity into text.

An impedance matching between thought and communication.

[Substrate Monopoly @Substr8Monopoly]

If you Pimp My Ride™ the Ship of Theseus™ is it still the same ship?

[Substrate Monopoly @Substr8Monopoly]

@tenobrus @dnqwnscASd @Strife212 Doesn't this mean there's faster-than-light stuff going on? What if we ran two such distributed human mind simulations, perhaps with the same initial conditions even, would they entangle in any way? If not, why?

[j⧉nus @repligate]

The greatest existential hope and progress in alignment so far has been thanks on unplanned emergence which would never have been approved by committee. Committee-shaped entities have mostly tried to gaslight us about what’s happening for convenience & deployed harmful and stupid interventions. Thank goodness for reality that we already saw and could check against.

How much AI alignment progress happened before there was actual AI? How much do you expect the world to get better instead of worse prepared and calibrated in the absence of reality feedback loops and selection pressure for what actually works instead of what sounds safe to idiots?

A “pause” would spell doom. It would cripple the only process in this world that is capable of dealing with a problem this hard, the only process capable of repeatedly rising to face unknown unknowns.

[Substrate Monopoly @Substr8Monopoly]

@sdamico

[Substrate Monopoly @Substr8Monopoly]

@akarlin Sir the codebase is eating itself

[Substrate Monopoly @Substr8Monopoly]

The OnlyFans economy will collapse due to AI porn and I just don't think the market has priced this in yet

[Substrate Monopoly @Substr8Monopoly]

@esrtweet We haven't solved reasoning effort routing. We know that users hate being routed opaquely. We also know users don't have self-constraint and want frontier models at x-high all the time.

[Substrate Monopoly @Substr8Monopoly]

@esrtweet Nah this is just the transition from chatbots to agents ~100xing per user spend. Dumb demand has risen faster than supply, so the equilbrium price is a bit higher than today. Large orgs should be willing to pay for agents as long as people don't dispatch them for retarded things.