GitHub/quartz
1
0
Fork 2
mirror of https://github.com/jackyzha0/quartz.git synced 2025-12-19 19:04:06 -06:00
Code Issues Actions 2 Packages Projects Releases Wiki Activity

extrusion edit

Browse Source
This commit is contained in:
Courtland Leer 2024-06-06 14:49:08 -04:00
parent 51119014b0
commit 000a6dcb77
2 changed files with 83 additions and 7 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

20
content/extrusions/Extrusion 06.24.md
View File

@ -7,24 +7,30 @@ tags:
- honcho
- philosophy
---
*Extrusions is a periodic shortform synthesis of what we've been chewing on recently at Plastic Labs--you can [subscribe here](https://plasticlabs.typeform.com/extrusions)*
# Cope Is the Canary, but Context Is Key (for The End of Software)
![[Copium Meme.jpg]]
Many reactions to Chris Paik’s prescient [The End of Software](https://x.com/cpaik/status/1796633683908005988) carry a distinct signature that readers of the [Pessimist's Archive](https://pessimistsarchive.org/) will recognize instantly–cope.
Many reactions to Chris Paik’s prescient [The End of Software](https://x.com/cpaik/status/1796633683908005988) carry a distinct signature that readers of the [Pessimist's Archive](https://pessimistsarchive.org/) will recognize instantly--cope.
Cope-y outbursts like this are almost always a canary in the coal mine. As technologists, we’re quick to notice the defensive, rationalizing outcry that accompanies the eve of disruption. If there were no threat, there’d be no negative reaction. But like everyone else, it’s hard to notice when it’s coming for you. When you’ve got skin in the game.
Cope-y outbursts like this are almost always a canary in the coal mine. As technologists, we’re quick to notice the defensive, rationalizing outcry that accompanies the eve of disruption. But like everyone else, it’s hard to notice when it’s coming for you. When you’ve got skin in the game.
It’s easy for us to see that creators denouncing the quality of image generators or English teachers asserting LLMs “only produce bad writing” herald the advent of serious change. They might be right…right now, but it’s only a matter of time (and market forces). No doubt they too can laugh at the historical examples of this in other groups disparaging stuff we all love and take for granted today. 
It’s easy for us to see that creators denouncing the quality of image generators or English teachers asserting LLMs “only produce bad writing” herald the advent of serious change. If there were no threat, there’d be no negative reaction. They might be right…right now, but it’s only a matter of time (and market forces). No doubt they too can laugh at the historical examples of this in other groups disparaging stuff we all love and take for granted today. 
The key thing to notice is that both positions can be true. New technology often does suck, but it also often gets way, way better. So much better that we can fully dispense with yesterday’s drudgery for tomorrow’s opportunity. Yet the ways in which the fresh tech sucks today form the roadmap to the ways it will be awesome in the future. It’s a mistake to say the problem is solved and a mistake to say it won’t be solved.
The key thing to notice is that both positions can be true. New technology often does suck, but it also often gets way, way better. So much better that we can fully dispense with yesterday’s drudgery for tomorrow’s opportunity.
Yet the ways in which the fresh tech sucks today form the roadmap to the ways it will be awesome in the future. It’s a mistake to say the problem is solved and a mistake to say it won’t be solved.
Chris is right that AI is coming for software like the internet came for journalism[^1]. But he’s making a predictive macro argument. And he’s not saying this is a done deal. Similarly, those arguing that how they do software development is more complex than what LLMs are currently capable of are right...but again, not a done deal. If the solution was complete, we’d be on to arguing about the next thing.
So what’s missing? What roadmap can we learn from the cope that gets us to disruption? What do LLMs lack and software engineers have that’s critical to translate ideas and natural language into applications?
At [Plastic Labs](https://plasticlabs.ai), we think it’s context. Not just context on how to do a general task, like writing code, but your context. How you would write the code? Why would you write it that way? To bridge the gap, LLMs need access to a model of your identity. How you solve a technical problem is about more than just your technical knowledge. It’s about all the elements of your identity and psychology and history that inform how you synthesize unique solutions. That’s why we’re building [Honcho](https://honcho.dev).
At [Plastic](https://plasticlabs.ai), we think it’s context. Not just context on how to do a general task, like writing code, but *your* context. How *you* would write the code? *Why* would you write it that way? What made you that person?
And to realize a future replete with trusted autonomous agents working across diverse domains on your behalf reliably, as true extensions of your agency, we’ll need Honcho too.
To know this--to bridge the gap--LLMs need access to a model of your identity. How you solve a technical problem is about more than just your technical knowledge. It’s about all the elements of your identity and psychology and history that inform how you synthesize unique solutions. That’s why we’re building [Honcho](https://honcho.dev).
[^1]: There’s a distinction to be made re: cs & journalism degrees. Journalism is actually more like software engineering here, & computer science like language. Lang & cs will remain useful to study, but the journalism & engineering trade degrees built on top of those primitives need a serious refresh to be worthwhile. I.e. it’s a good idea to have aptitude with symbolic systems & abstract technical knowledge, but application & execution will change as technology evolves.
And to realize a future replete with trusted autonomous agents working across diverse domains on your behalf reliably--as true extensions of your agency--we’ll need Honcho too.
[^1]: There’s a distinction to be made re: the piece's comparison between computer science & journalism degrees. Journalism is actually more like software engineering here, & computer science like language. Lang & cs will remain useful to study, but the journalism & engineering trade degrees built on top of those primitives need a serious refresh to be worthwhile. I.e. it’s a good idea to have aptitude with symbolic systems & abstract technical knowledge, but application & execution will change as technology evolves.

70
content/notes/honcho's (secret) crypto roadmap.md Normal file
View File

@ -0,0 +1,70 @@
disclaimers:
- we don't talk about this often yet for a few reasons:
- the first priority is building honcho's core technology and turning that into a viable business
- it's hard enough to communicate a paradigm shifting AI concept, without adding a crypto narrative on top
- we need to win the AI/ML community first
- everything that follows aside, honcho as self-improving user memory and best in class theory of mind models are a massive business opportunity
but honcho's mission is fundamentally about decentralization--radically decentralizing alignment, aligning to the individual first...to do so it must:
- be credibly neutral--trustless, unbiased, transparent
- cohere to the user, building a high-fidelity, auditable simulacra
- project agency across substrates, enabling digital agents to act and transact as the user
- camouflage identity against malicious actors
so we see our crypto roadmap as one organized around progressive decentralization and it's fundamentally tethered to our identity, privacy, and security roadmaps
- we start like any startup, with a core team seeking to build new technology and monetize it
- but since we're playing in the sensitive domain of identity, we have to be thinking about trust and ownership, hence crypto
- crypto is a trust primitive and AI is an intelligence primitive--we can build on top of both to solve the problems of digital identity
- so we plan to iterate through phases of maturity for honcho that become more and more autonomous, credibly neutral, and protocol-like
- this builds trust, protects identity, and lays the groundwork for a new agent economy
- which is an even bigger prize than we typically pitch at this stage
- eventually, the identity management parts of honcho are probably an intelligent hyperstructure
- it’s ultimately the only way to be trusted with the magnitude of what we’re proposing to do--go trustless
we're currently thinking about things in terms of 5 overlapping phases:
PHASE 0: Beta (now)
- work with early partners to hone honcho's core value
- data is siloed and not shared
- users agree to terms
- as privacy/security roadmap is being built out, open core earns us credibility and trust
PHASE I: Confidential Computing (within 6mos)
- right now, we can compute-to encrypted data within secure GPU enclaves
- the result of inference is not perfectly leak-proof, we do have to trust a cloud provider and our team
- but it's currently highest-grade privacy & security paradigm available for AI that's economical right now
- any post-beta honcho service will use this technology
- UPGRADE: when it's economical, fully homomorphic encryption will replace secure enclaves to allow user data to be computed over anywhere while encrypted
- this means no party technically needs to see user data, and we're not married to secure enclave providers
PHASE II: User-Owned Data (within 18 mos)
- it's already technically feasible to allow users to hold the keys to their data and give them full custody
- when the rest of the product is ready, we'll make this switch
- that shift complete, we can start to connect data stores across applications, firing up the honcho network
- we can also then have user facing controls, so users can chat in natural language with their honcho to edit/curate/supplement the model of themselves being used to personalize their experiences across apps
- even pre-ZK here we can eliminate some need for persistent user accounts and run active obfuscation in the background to destroy statistical relationships between users and their data
- UPGRADE: when it's economical, decentralized storage and compute networks will be leveraged (optionally or by default) to eliminate nexus of trust around cloud providers
- UPGRADE: when it's economical, we can compress storage of user represenation inside per-user AI models that get trained and re-trained dynamically--your honcho is not longer just storage and insights with an agent on top, but an agent that shares your identity
PHASE III: ZK-powered Agent Economics (over 2-5 years)
- honcho is now sufficiently decentralized, with control of digital identity being restored to the individual--regen identity
- honcho becomes the logical top of funnel to the new internet of agents
- with ground truth on you, it's perfectly positioned to recruit agents to take action on the internet on your behalf
- interface to the new internet
- also logically, this happens on a crypto substrate
- zk-proving elements of your identity without giving away personal details will be critical
- these proofs can be tokenized and used down a lineage of autonomous agents
- anonymous data markets on robustly modeled identity become possible
- businesses can train on user data/representations and compensate users without revealing any personal info
- lots of research to do here, but most personalization operations can probably eventually be done in a zero knowledge way
- the need for persistent user accounts no longer exists--everything is just-in-time personalized
- in fact, zk will eventually enable persistent accounts with access to elements needing verification (e.g. credit, payments, legal function), while preserving anonymity
- well before all of this, we can use zk-ml for ultra-sensitive use cases like medical agents (as long as the compute overhead is justified for the app builder)
PHASE IV: Intelligent Hyperstructure (within 5-7 years)
- limit of this future is parts of honcho become a self-sustaining protocol that can run for free and forever with credible neutrality and anyone can use, but is still profitable to own
- it forms the substrate of intention and identity for agent-agent interaction on the web
- maintained by neutral agents running on distributed compute whose actions are recorded on distributed ledgers
- sufficiently abstracted, no need to be what we think of today as a crypto-native to use
- user can leverage the protocol's toolkit to manage their digital identity, train models of themselves, and allocate intelligence to whatever end
- businesses and agents can aquire user identity data brokered by the protocol
- perhaps plastic, having profited massively along the way, spawns a temporary nonprofit (DAO?) that oversees the hyperstructure's establishment and distributed ownership while the for profit continues to monetize everything already built on top of it, as well as new opportunities arising from protocolization of digital identity