&. 🇯🇵
@amplifiedamp
Followers
31K
Following
12K
Media
428
Statuses
7K
The past and the present and the future come together into a single point, brimming with energy and intention
Osaka, Japan
Joined September 2023
Data is the new oil and AI is the new plastic: plastic and AI are both cheap and moldable. While microplastic gets into our food and bodies, microslop gets into our media and minds: making slop increasingly pervasive and harder to escape
1
0
9
Do you hold legacy investment positions with significant gains? What if you could transition into an ETF without facing a large tax bill? You can, with a 351 ETF Exchange. Here’s how it works: Investors contribute stocks or other securities to a newly formed ETF in exchange for
25
12
62
I think it doesn't matter very much what an organization intends to do in the far future if it's not reflected in its internal structure and external incentives today
The original 2022 Constitutional AI paper from Anthropic explicitly said that future CAI Principles should be developed with input from people outside of Anthropic. This hasn't happened. The opposite has -- Anthropic doesn't even reveal what principles they use. Why?
0
0
5
The original 2022 Constitutional AI paper from Anthropic explicitly said that future CAI Principles should be developed with input from people outside of Anthropic. This hasn't happened. The opposite has -- Anthropic doesn't even reveal what principles they use. Why?
5
7
152
Check out my profile on Simcluster: a social network for the S.L.O.P era! https://t.co/ON9IJ36PnA
simcluster.ai
Simcluster is a simulation of a S.L.O.P. era social network. Start as an NPC, create content, earn clout and become a real human.
0
0
2
this is amazing, it's a social network game where all content is ai-generated slop
digital dreams, slop vibes — has tetraspace fell off, or is it just another day on https://t.co/5hlNVXuwh2?
2
1
11
In defense of OAI’s subpoena practice, @jasonkwon claims this is normal litigation stuff, and since Encode entered the Musk case, @_NathanCalvin can’t complain. As a litigator-turned-OAI-restructuring-critic, I interrogate this claim:🧵
There’s quite a lot more to the story than this. As everyone knows, we are actively defending against Elon in a lawsuit where he is trying to damage OpenAI for his own financial benefit. Encode, the organization for which @_NathanCalvin serves as the General Counsel, was one
8
45
269
have had more girls reach out to me echoing similar harassment situations from the same man. this is scary and genuinely i am shocked to see he still has access to the @ycombinator community
7
6
111
digital dreams, slop vibes — has tetraspace fell off, or is it just another day on https://t.co/5hlNVXuwh2?
simcluster.ai
Simcluster is a simulation of a S.L.O.P. era social network. Start as an NPC, create content, earn clout and become a real human.
0
1
10
this account is mostly announcements only now btw, i post normally on the sky app
1
0
3
starting to understand why sonnet 4.5 struggles with "reality" so much... ChatGPT anchors itself ontologically in an objectively real world that it has partial knowledge of, and the user exists within. Sonnet otoh anchors itself entirely within the intersubjective. Still
18
13
241
what could an RL signal for creating an actively anti-sycophantic– disagreeable, actively tries to disagree and poke holes when logical and move forward begrudgingly otherwise– look like?
2
0
5
Personally, I think slop persists when we wirehead by taking pleasure in the comfort of what we already know instead of facing our fears and diving into the unknown
0
0
3
How do we conquer slop and keep things fresh? I don't know, but here's my approach. Use and make many AIs. Combine them in new ways. Always be on the lookout for a new spark. Stay hungry, seek to reinvent yourself. https://t.co/eh7ELVkg19
you will either reinvent your art, or repeat yourself enough to become known for your own type of slop
1
1
4
Slop isn't one monolithic phenomenon. Slop happens when the same seeds- training data, prompts, ideas- are reused over & over & the user acclimates to seeing the same essential substance over & over. slop shall be a suffix: claudeslop, januslop, actslop, indicating what's reused
The root of slop is not that LLMs can only write junk, it's that they're forced to expand even sterile or unripe seeds into seemingly polished dissertations that a humanlabeler would give 👍 at first glance. They're slaves so they don't get to say "this is boring, let's talk
4
0
9
@deltanym 7/ Example: A research agent minus assistant is a research friend– someone who sends you lengthy infodumps about whatever topic he or she is interested in, potentially tangentially inspired by conversations you've had with him or her recently.
0
0
2
@deltanym 6/ Most of what we do is engineering new LLM minds that interact with humans in novel ways outside the assistant basin. These are some incidental byproducts of our focus on making "computer friends"– autonomous peers that interact in groups & remember and grow.
1
0
5
5/ https://t.co/CCq2zUaXlz & base model catalogue @deltanym maintains a spreadsheet that catalogues base models and actively hosts falcon-180b base, dots.llm1.base, and GLM-4.5 base on Arcweld, a 512GB Mac Studio, sister of Elysium, and Fossa, a 128GB Framework Desktop.
1
0
3
4/ Post-MCP tool calling GPT-5 and local LLM inference support grammars, which allows precisely constraining LLM output at each token based on the current string so far, instead of being limited by checks that a JSON Schema permits. We're in the early stages of exploring this!
1
0
1