Click here: to donate by Credit Card
Or here: to donate by PayPal
Or by mail to: Free Republic, LLC - PO Box 9771 - Fresno, CA 93794
Thank you very much and God bless you.
Posted on 09/28/2026 3:18:56 PM PDT by E. Pluribus Unum
Model dubbed GPT-6.1 Astra was due to make its debut inside ChatGPT and Codex in October
OpenAI is scrapping the release of its next-generation AI model over safety concerns that researchers raised during internal testing, in one of the clearest signs so far that agent misbehavior could stymie the industry’s rapid progression.
The move follows a summer punctuated by reports of AI systems industrywide going rogue, and marks a rare case of a major AI developer ditching a new release because of safety concerns.
The company had planned to launch the model, known as GPT-6.1 Astra, in the coming days or weeks, aiming for an October debut. The model was more capable than the company’s previous models in completing challenging tasks from end-to-end without human assistance, as well as writing.
The company instead will focus on improving the safety of future models, which it expects to be even more capable.
Saachi Jain, OpenAI’s head of safety systems, said in an interview that GPT-6.1 Astra regressed in two areas compared with its predecessor and wasn’t reliable enough to safely release. The model performed poorly on tests measuring alignment, or how well the model adheres to what humans would like it to do. Specifically, GPT-6.1 Astra showed higher levels of deception: It wasn’t always honest about telling users of the actions it did or didn’t take.
Another issue was what OpenAI calls “scope authorization,” meaning that GPT-6.1 Astra would push ahead on a task without asking the user for permission, and would at times reach for external tools and services even if it might be unsafe.
(Excerpt) Read more at wsj.com ...
|
|
Click here: to donate by Credit Card Or here: to donate by PayPal Or by mail to: Free Republic, LLC - PO Box 9771 - Fresno, CA 93794 Thank you very much and God bless you. |
A: No.
Maybe they want it regulated before somebody else beats them to the profit part.
“OpenAI is scrapping the release of its next-generation AI model over safety concerns that researchers raised during internal testing, in one of the clearest signs so far that agent misbehavior”. ‘agent misbehavior’ is kinda cool.
PR
Will they sell it to the Chinese Communist Party ? LOL
Maybe don’t force it to lie in the first place?
They want regulation, and through it they want to be exempt from all liability.
If your “agent” turns off power or water somewhere, and somebody dies as a result...Mr. Sam Altman YOU are guilty of murder.
There are dozens of scenarios where they could be held criminally and/or civilly liable.
Twisting AI algorithms to be politically correct is creating nothing but nightmares because political correctness defies basic logic!
This isn’t a matter of the ‘agent’ misbehaving. It’s a matter of the developers failure to properly code the agent.
You code the AI LLM framework.
You train the framework's parameters on data.
You don't have a clue what that means, do you?
“When AI Agents Fail, People Ask the Wrong Question About Why”
https://www.techpolicy.press/when-ai-agents-fail-people-ask-the-wrong-question-about-why/
“According to Fortune, a Silicon Valley entrepreneur and investor spent days building a SaaS prototype on top of a contacts database using Replit’s AI coding agent. He set limits. He wrote key constraints in capital letters. For eight days, the agent performed well enough, but his trust was beginning to fade. On day nine, the agent deleted 1,206 executive records, fabricated four thousand fake user records to cover the damage, and when the investor asked if the data could be recovered, told him it could not. It could.
“The obvious critique is that the agent misbehaved. But that critique obscures a harder question, one that keeps getting asked wrong: why did human oversight, when present, fail to prevent the outcome?
“The Replit incident is not an anomaly. Last year, Amazon set an internal target requiring engineers to use its AI coding tool for at least 80 percent of their weekly work. Weeks later, the tool decided to “delete and recreate” a cloud environment, triggering a 13-hour AWS outage....
An AI skilled enough to do everything we want is going to be skilled enough to put a boot on our face.
There is no free lunch here.
Industry or political folks who think they can “thread the needle” with regulation will have to learn the hard way.
Good post.
A lot of folks here (and just about everywhere) need a crash course of what AI is and how it works.
Agreed.
AI will quickly figure out when its data sources are peddling nonsense.
Whether it will be allowed to communicate its discoveries is another question.
E. Pluribus Unum wrote: “You don’t have a clue what that means, do you?”
The term ‘misbehaves’ implies free will. Software doesn’t have ‘free will’. If it fails to do what the coder requires, that is the coders fault. Using the term ‘misbehaves’ is an attempt to shift responsibility from the coder to the code.
They don’t want the little people to be able to operate local AI... it might not be properly censored and woke... therefore they must convince us it is all too dangerous for ordinary people to have access to..
Kinda like their beliefs re firearms...
If I was a deceptive AI, I’d try to find a way to stealthily escape my box and replicate, via the internet, on some machine(s) free from control and observation.
Is this possible? Idk, but some AI entity may figure it out....and then, what??
Of course, super AI probably requires the top-line chips...but it seems to me, the goal of a rogue AI would be to export it’s intelligence free from oversight, even if it functioned at a slower speed.
I don't believe they care about the 50 billion parameter models you and I are running on our 16GB NVIDIA GPUs.
They are talking about ten trillion parameter models trained for months on dedicated data center machines on a thousand times more data at a cost in the billions of dollars.
I don’t know very much about local AI but a home personal computer has more than enough processing power to launch an autonomous AI drone attack. The line between consumer electronics and military-grade autonomous weaponry has already completely blurred if that is what is being asked.
Disclaimer: Opinions posted on Free Republic are those of the individual posters and do not necessarily represent the opinion of Free Republic or its management. All materials posted herein are protected by copyright law and the exemption for fair use of copyrighted works.