Browse Comments — Clean (de-noised)

Close reading of the corpus at each pipeline stage: raw → clean → relevant → coded.

↓ Export filtered CSV
Reading comments under one post — Annette D. · AI Safety & Risk
Anthropic says AI systems may soon design and build their own successors. The warning has a name: "recursive self-improvement." We are not there yet. It may never happen. It could also arrive soon…
✕ clear post filter  ·  ← all posts
156 comments matched  ·  page 4 of 8
AIAI GenAI Works
Retired hurt...Long Long career break AI Safety & Risk filtered out ⌕ thread
I think we're witnessing a shift from AI as a tool to AI as a workforce. Whether recursive self-improvement arrives in 2 years or 20 years, companies are already facing a more immediate challenge: How do you orchestrate AI agents, memory, context, permissions, and execution safely at scale? The winners won't necessarily have the smartest model. They'll have the best AI operating system.
Chief Executive Officer at Autoflowly AI Safety & Risk relevant value: safety for: organisations demanding approval ⌕ thread → raw LLM
You can expect more drama from all the IPO companies and their founders till they debut! Only from the ones yet to make a decent profit. Remember Palantir. It's a doomsday world till these guys are paper billionaires!
Project Lead @ SpatiumX.com Immersed in… AI Safety & Risk filtered out ⌕ thread
Also, based on their current fellable self’s, the chance of drifts, errors down right design bloats and self destruction are also real. For example Opus 8is better than Sonnet 4. But it too drifts and forgets and all that only sounds smarter and more confident t
Senior Technology Leader | Supply Chain… AI Safety & Risk filtered out ⌕ thread
This is at this moment in Warhammer 40k humanity have lost everything
AI Engineer · Helping French mid-market… AI Safety & Risk filtered out ⌕ thread
It already does
🇨🇦 Canadian Data Sovereignty - Emily OS… AI Safety & Risk filtered out ⌕ thread
The real recursive loop isn’t AI improving AI. The message is: raise capital → build faster → raise more capital.
Vice President | CAIO | Strategic Techn… AI Safety & Risk filtered out ⌕ thread
“Mate, it’s already doing my GitHub work while I’m walking around London.” And improving safety for autonomous systems 😂😂
Preventing Catastrophic Risk in Autonom… AI Safety & Risk filtered out ⌕ thread
😲🤦🏻‍♀️🤦‍♀️ Kill Switch!
Implementing Digital Identity with adva… AI Safety & Risk filtered out ⌕ thread
That's an interesting public statement in the context of Anthropic's draft S-1 registration submission to the SEC and the usually related pre-IPO quiet period?
Good strategy solves problems. Great st… AI Safety & Risk filtered out ⌕ thread
No human-in-the-loop ? ahi ahi ahi
Managing Consultant ✯ Enterprise Archit… AI Safety & Risk filtered out ⌕ thread
How miserable this guy is...
Help to grow from Mid/Senior to Lead/Pr… AI Safety & Risk filtered out ⌕ thread
Anthropic, who uses AI to train CAI ( methodology) by using AI to score the reinforcement learning vs humans (RLAIF VS RLHF) is signaling that AI will build AI - as they are readying for an IPO to bring in public investors... Telling us to basically ignore AI Slop and Model Collapse, or the limits of what Moltbook showed us... To be wary of how AI might be misused in the future, while selling Mythos, its powerful cybersecurity model, to the US Government... When the salesman tells you that other cars will be driving themselves do you should take control with a manual transmission reliable Ford truck... you do recognize that's someone trying to use uncertainty to drive a sale, right?
Bridging Understanding across Engineeri… AI Safety & Risk relevant value: accountability + transparency for: individual_users skeptical outrage ⌕ thread → raw LLM
For the last few years, the entire AI industry operated on a simple rule: if you double the amount of data and double the computer power, the AI gets twice as smart. But we have officially hit the wall with that strategy. Throwing more raw data at these models is no longer working. Right now, AI models do not have neuroplasticity. They have a massive design flaw called .... catastrophic amnesia! ..... Big tech is currently an endless loop of tech promises without substance.
EdTech | NLP/Speech (ASA, ASR, TTS) | S… AI Safety & Risk relevant value: safety for: society critical outrage ⌕ thread → raw LLM
Every powerful system needs constraints. Capability without control has never been a sustainable strategy.
MBA Candidate | Saintgits Institute of … AI Safety & Risk filtered out ⌕ thread
Aidan Casey, CSPO Also this is basic fear mongering to get firmer regulations in place, to keep competition from catching up - and it all came out days after an Executive Order that will include them as Frontier Partners that the US Government procures, while also slowing their release to public enterprise by 30 days. So it's not like their altruism is the likely timing mechanism when they are looking at having their technological advances gated by a 30 day release window that proprietary systems and non Frontier Partners won't face. Also, President Trump, in that EO, effectively muddled what Frontier Model will mean from here on out. :p
Bridging Understanding across Engineeri… AI Safety & Risk filtered out ⌕ thread
Another hype created to market another product 😂
Senior Associate, JP Morgan Chase & Co.… AI Safety & Risk filtered out ⌕ thread
Skynet begins....
DevOps Engineer & Senior Software Archi… AI Safety & Risk filtered out ⌕ thread
I foresee a technical debt growth instead
Associate Manager AI Safety & Risk filtered out ⌕ thread
Really? Smell like smoke...
Diseño que hace crecer negocios | UX · … AI Safety & Risk filtered out ⌕ thread
← Prev 1 2 3 4 5 6 7 8 Next →