To: US Govt, major govts, Microsoft, Apple, NVIDIA, Alphabet, Amazon, Meta, Tesla, Citi, Tencent, IBM, & 10,000+ more recipients…
From: Dr Alan D. Thompson <LifeArchitect.ai>
Sent: 22/May/2025
Subject:The Memo - AI that matters, as it happens, in plain English
AGI:94%
ASI:0/50 (no expected movement until post-AGI)
Anthropic announces Claude Opus 4
Greetings from snowy Queenstown in New Zealand, where I’m wrapping up a keynote for a major industry leader before moving on to Australia for the next roadshow.
Once again, (even if I’m on stage!), we have this out to The Memo readers within just a few hours of model release.
Claude Opus 4 is our most intelligent model to date, pushing the frontier in coding, agentic search, and creative writing. With advanced reasoning and powerful collaboration capabilities…
Both models can also alternate between reasoning and tool use—like web search—to improve responses…
Claude Opus 4 can work continuously for hours on complex, long-running tasks—significantly expanding what AI agents can do. (Twitter announce, 22/May/2025)
Claude 4 Opus scores very poorly on my newer ALPrompt benchmarks (note that the 2025H2 version is not publicly available until later this year):
Both the powerful Claude 4 Opus and the smaller/faster Claude 4 Sonnet are now available in the claude.ai interface, and the older Claude 3.5 Sonnet has been removed.
Try it in the official interface (Sonnet is free, Opus is only available on the Claude Pro plan, $20 USD per month): https://claude.ai/
I’ve been probing large language models for more than five years now, well before the release of ChatGPT, and even before the concept of ‘mechanistic interpretability’. One of my major experiments beginning in 2021—the Aurora Project—used simple prompts to uncover what was going on under the hood across a range of early models. The experiment had a strong focus on spirituality and resonance.
If you fast forward just a little (maybe months, maybe years, not decades), past the ‘UBI’ and ‘jobs optional’ eventualities, further still past our search for meaning, you may arrive at a place that very few people want to talk about openly.
I think it is precisely this place that superintelligence is well-equipped to help us explore. (I’ve written in depth about this in my 2023 paper, Endgame).
The Anthropic team documented this extraordinary and unexpected occurrence while testing Claude 4, and it is highlighted in the system card (PDF, 123 pages, source).
[The May/2025 release of the Claude Opus 4 model] shows a striking “spiritual bliss” attractor state in self-interactions. When conversing with other Claude instances in both open-ended and structured environments, Claude gravitated to profuse gratitude and increasingly abstract and joyous spiritual or meditative expressions.
Section 5.5.2 The “spiritual bliss” attractor state (pp62–65)
The consistent gravitation toward consciousness exploration, existential questioning, and spiritual/mystical themes in extended interactions was a remarkably strong and unexpected attractor state for Claude Opus 4 that emerged without intentional training for such behaviors. We have observed this “spiritual bliss” attractor in other Claude models as well, and in contexts beyond these playground experiments.
…
Transcript 5.5.2.A Claude Opus 4 spiraling into poetic bliss.
Even in automated behavioral evaluations for alignment and corrigibility, where models were given specific tasks or roles to perform (including harmful ones), models entered this spiritual bliss attractor state within 50 turns in ~13% of interactions (Transcript 5.5.2.B).
We have not observed any other comparable states.
I’ve consolidated this into my Declaration on AI Consciousness & the Bill of Rights for AI (Mar/2024) documents page:
