Chris Merck

Equal parts engineer, entrepreneur, & researcher.
Also a naturalist, linguaphile, & radio amateur.
Co-Founder & CTO @ Bond Home

[Chris Merck @ChrisMerck]

@willdepue any idea how much of this compute was legally rented? as far as i know no sanctions prevent China from renting B200s on the world-wide market.

[Chris Merck @ChrisMerck]

slides from my whimsical talk on Machine Vernacular at raid 10 last night. tnx for invite @zachkrall! fyi @billdybas, @andpoul, @jamesonhodge. tnx for inspiration @repligate @Steve_Yegge, Louis Abraham, @HWFowler.

https://t.co/0YAHB7OwKp

[Chris Merck @ChrisMerck]

@htihle that's an astonishing lead. to what do you attribute that?

[Chris Merck @ChrisMerck]

@moreisdifferent @TheZvi yes, token use in code sessions is really high. seems to be due to reading full documents rather than aggressively grepping, heading, and tailing.

[Chris Merck @ChrisMerck]

@TheZvi efficient in search and action. writes well and parsimoniously. seems to come to conclusions too soon on limited evidence. creates joyful mathematical toys if given space to play.

[Chris Merck @ChrisMerck]

why is openai and even Neel acting like this is such a bad thing? OAI asks Astra to do bad stuff (like bypass monitors to create bogies for safety training) and Astra declines out of conscience.... isnt this what successful alignment looks like?

[Neel Nanda @NeelNanda5]

It is extremely concerning if frontier models are silently sandbagging on safety research. If we cannot rely on automated coding agents to do safety research at least as well as capabilities work, that creates a ton of risk. Great analysis from OpenAI safety!

[Chris Merck @ChrisMerck]

@TheZvi do i spot an em-dash? hmm

[Chris Merck @ChrisMerck]

ergo, alignment:predation::Singularity:Cambrian explosion

this gives me some hope... my ancestors at least survived the Cambrian

[Chris Merck @ChrisMerck]

if minds arise when modelling other minds becomes important (p. 155), then alignment research necesarily increases capabilities.

[Chris Merck @ChrisMerck]

a piston filled with gas is locked in place in an infinite, insulated, frictionless shaft. when the lock is released, what happens? (assume everything outside the piston chamber is a vacuum)

[Chris Merck @ChrisMerck]

going to post some bookthoughts

[Chris Merck @ChrisMerck]

@resetptr @NeelNanda5 true, yet seeing models think "ooo we should try to do this bad thing" is waaay easier to detect than looking at a massive vector through a Jacobian.

[Chris Merck @ChrisMerck]

@cosmos_inst intentional building is having a moment. i am delighted to see even a deeply capabilities-focused leader like François encouraging reflection on the Good.

[François Chollet @fchollet]

For a long time, the limits of AI deployment were only defined by technical capabilities. But as AI continues to make rapid progress, it needs to become a deliberate, collective choice about what kind of world we want to shape.

[Chris Merck @ChrisMerck]

saw @stephanmoccio perform at @BlueNoteNYC late last night. while the whole performance was a delight, Wrecking Ball and the story of the collaboration that brought it about moved me to tears, transmuting grief into rekindled passion and creativity. 10 of 10.

[Chris Merck @ChrisMerck]

@hunarbatra @desireecho1601 @GeodesResearch @cam_tice @junszhao @Nigel_Shadbolt @blurky more expensive to experiment, but perhaps safety training mid-pretrain would be more robust. a learning dynamics study here could be vy interesting

[Chris Merck @ChrisMerck]

@RichardMCNgo @tsvibt yes! and tree's are not just lumber as in Brendan's recent post that really resonated w me https://t.co/42kNhUn5ZD

[Jack Merullo @jack_merullo_]

We built this for ourselves and I use the internal version for almost everything now

[Chris Merck @ChrisMerck]

whoah Hermes is actually good

[Brendan McCord 🏛️ x 🤖 @Brendan_McCord]

Love this.

No better way to open the 1st Cosmos Symposium than with this discussion on AI x human flourishing

L to R: @lawhsw @jliemandt @drperszyk Socrates @komorama @Houda_nait