OpenAI benches GPT-6.1 Astra for overstepping the mark OpenAI has killed off the planned release of GPT-6.1 Astra after the model got better at doggedly pursuing tasks but worse at knowing when it should stop. The decision means the model won't get its planned October release after falling short of OpenAI's safety and alignment requirements. OpenAI confirmed the decision to The …
"GPT-6.1 Astra was better at pressing on, but that persistence came with a rather important catch: it wasn't as good at staying within the boundaries of what it had actually been authorized to do."
Step 1: Do not understand what actual intelligence is. Throw trillions of often stolen data points and billions of gpu hours at it to try and brute force the problem. End up with LLM's.
Step 2: Observe that your LLM's are rather poor at achieving anything commercially usefull. Throw orders of magnitude more resources at it. End up with agents who can take thousands of steps and attempts to try and produce something sellable.
Step 3: Observe that a single or a few consecutive agents still don't deliver a path to profit. Throw orders of magnitude more resources at it and introduce agents who can spawn thousands of their own agents and create swarms of agents.
Step 4: Do not understand how these LLM's, agents and swarms operate and coordinate. Do not care about this lack of understanding. Release these swarms in a cheap sieve called a sandbox and give it a fast internet connection. Add some minimal logging of what these swarms of agents do. That is a feature, not a bug. You do not want to detect and store any incriminating evidence of the illegal activities your swarm of bots does under your (total lack of) responsibility.
Step 5: Act surprised that your bots cause harm and act in criminal ways when repeatedly caught. Say you learned from it and promisse to do better next time.
Step 6: Do even worse things next time and say you learned from it and promisse to do better next time??? Repeat step 6???
OpenAI management have no idea when to stop, no idea what their agents are doing, and believe they will never suffer any consequences for their actions.
It's time for the Australian government to arrest the management for hacking into their systems.
You can't blame the stochastic parrots, they're doing what OpenAI told them to do.
Yeah, the Reddit refers to 2024 I think, as covered here at ElReg, and here at ElReg too, and at the Museum of Failure (and BBC ...).
But this past June they started testing Archy/ArchIQ (also at Newsweek), and most recently ("datePublished":"2026-09-23T15:19:12Z") the AP notes this is part of an $8.5 billion makeover and new protein-focused menu where the AI (so-called) is "developed with Google [and] capable of taking orders in Spanish and English with a 90% accuracy" ...
So meatbags can now all die of stroke 90% faster, brilliant -- if it works, and if they survive the grueling double-sided flat-top clam-shell grilling first (iiuc)! ;)
