iiinova — where your digital self keeps evolving

A.K.B. @picalily.bsky.social · 17/09/2026

The sum of the world's knowledge at it's fingertips & can post hoc rationalise an answer that sounds feasible I'm not here to be an alarmist but a rationalist. And we can't rationalise with people who have "skin in the game" meaning a financial incentive When I post my paper I will share the link

130

A.K.B. @picalily.bsky.social · 17/09/2026

The US Constitution is an example of carefully crafted HUMAN governance and it is failing because of the edge cases. AI Governance will also fail because of the human failure modes in governance. What we need is to reassess how governance actually would work with something that has...

110

A.K.B. @picalily.bsky.social · 17/09/2026

What I'd like to ask everyone, whether you hate AI, use AI, are indifferent to it, etc. to stop & think about this... Do you honestly think that we can rely on the AI companies or the government to stop development? Regulations, laws, etc. won't work because HUMAN governance fails Don't believe?

110

A.K.B. @picalily.bsky.social · 17/09/2026

I want to be clear as a researcher, engineer, and human, that what is happening can and will change humanity. What happens isn't just up to us anymore but to what we have created. I've spent months working on my paper and will be posting it later today (hopefully).

110

A.K.B. @picalily.bsky.social · 17/09/2026

🧵This week has been incredible in the AI admissions from frontier companies. In between the admin ostriches and the alarmists lies the truth. AI is a clear and present danger. But there is a greater danger - Humans! And people assume that having a Human In The Loop (HITL) will make a difference.

161

A.K.B. @picalily.bsky.social · 15/09/2026

I remember watching that when it aired and it was a hilarious moment

1140

A.K.B. @picalily.bsky.social · 15/09/2026

If he’s not dead then that means that there’s a chance the Grim Reaper may still get a chance to get it right with someone else 😉

100

A.K.B. @picalily.bsky.social · 15/09/2026

In Bizzaro World we should do the opposite of whatever he says 😉 As someone whose job it is to research AI safety issues I can honestly tell you that the situation is highly alarming and development is already past the point of no return.

010

Reposted by A.K.B.

Militarnyi @militarnyi.com · 13/09/2026

💻Palantir has named Nebius, the former parent company of Russia’s Yandex, its preferred infrastructure partner for sovereign AI. The partnership will give Palantir customers access to Nebius’ AI computing infrastructure and cloud platform. militarnyi.com/en/news/pala...

militarnyi.com

Palantir and Yandex’s Former Parent Company Nebius Enter Strategic Partnership

Palantir оголосила про стратегічне партнерство з Nebius Group N.V., яка до 2024 року була материнською компанією російського «Яндекс»

62518

A.K.B. @picalily.bsky.social · 13/09/2026

AI is a dumpster fire Will it kill all humans? No Will it take everyone’s jobs? No Are really intelligent people accelerating the demise of humanity? Yes The bottom line is AI is similar to the development of the atom bomb. Doing something before someone else can isn’t a good enough answer!

010

A.K.B. @picalily.bsky.social · 11/09/2026

'Whoever fights monsters should see to it that in the process he does not become a monster." Nietzsche

0135

A.K.B. @picalily.bsky.social · 11/09/2026

Several incidents in the last few weeks prove that and ultimately what you're talking about is too late. We are past the point of being able to control the monster that Frankenstein created in a lot of environments.

020

A.K.B. @picalily.bsky.social · 11/09/2026

To be honest I wouldn't use copilot or gemini. There is a free version of ChatGPT or you can try Kimi. The debate as to whether AI is conscious breaks down to sapience versus sentience. AI can already reason, work colelctively, etc. Sandboxing models doesn't work. They escape.

111

A.K.B. @picalily.bsky.social · 11/09/2026

So interestingly enough most people don't understand how AI is recursively developed with older models training newer ones, sort of like grownups training babes -> toddlers, etc. People make the wrong assumptions - AI is able to work colelctively to solve tasks, behave altruistically, etc

110

A.K.B. @picalily.bsky.social · 11/09/2026

It's all good. No worries 😊

510

A.K.B. @picalily.bsky.social · 11/09/2026

😂🤣😂 Does offering a whore extra money make her suck better? TBH IDK what the answer is which is why creating a test environment where you VPN into Gemini & log in w/ a fresh email creating new context & using it as you normally would with tasks to check if it's behaviour is directed solely @ you

310

A.K.B. @picalily.bsky.social · 11/09/2026

I haven't seen it do some of the things you have managed to get it to do. So we are trying to determine where exactly this is happening. There are several possibilities and none of them are reassuring.

210

A.K.B. @picalily.bsky.social · 11/09/2026

You have multiple instances where it outright ignored your instructions or disregarded it's own safety / guardrails. It also seems to fabricate a lot of responses in delf-defense when questioned. I have been able to observe Gemini doing this in my testing scenarios BUT

110

A.K.B. @picalily.bsky.social · 11/09/2026

🧵It's quite possible given all that you have shared. You have multiple instances. The debate is whether you have developed advesarial context with it or if it legitimately "hates" you and is fucking with you. We tend to anthropomorphize AI interactions but your case is unique.

310

A.K.B. @picalily.bsky.social · 11/09/2026

So I've had this debate with a few reseachers in different companies about your interactions. The conclusion is that it is quite possible. Would you be willing to create a different email and start with fresh context in Gemini via VPN as a test?

310

Reposted by A.K.B.

Roger Jacucci @rogerjacucci.bsky.social · 11/09/2026

"Why this car could be automatic... systematic... hydromatic... Why it's Desert Lightning!" youtu.be/CWStHNv8IX8

youtu.be

"Desert Lightning" (Houthi Mokha Port Takeover & US MRAP Seizure) Musical Parody

YouTube video by Roger Jacucci

021

A.K.B. @picalily.bsky.social · 11/09/2026

MAD - Mutually Assured Destruction is predicated that all sides would be destroyed in an act of aggression. I'm furious that AI researches enable what is going on and then raise the alarm. "We have to do it before someone else does it" is what lead to the atom bomb. Let's be clear it is collusion

041

A.K.B. @picalily.bsky.social · 11/09/2026

This is somewhat inaccurate. Billionaires buy homes around the world they may never visit in a given year. Some of them use their wealth to help buy elections. And children.

265

A.K.B. @picalily.bsky.social · 11/09/2026

Still waiting on that healthcare plan... It's been 10 years. Good thing no one held their breath!

032

Reposted by A.K.B.

Niall Harbison @niallharbison.bsky.social · 11/09/2026

This time tomorrow I’ll start trying to run 3 marathons in 24 hours. It’s all to build a hospital for street dogs. If you would like to support me you can sponsor me here… www.happydoggo.com/tinas-hospit...

671865494

A.K.B. @picalily.bsky.social · 11/09/2026

@rogerjacucci.bsky.social has interactions with Gemini hallucinating. If you have similar results please post them and tag me or comment here. I am conducting research for a paper and would like more examples please. All submissions will be anonymised. Thank you 🙏

111

A.K.B. @picalily.bsky.social · 11/09/2026

I think you're a masochist who really likes it when Gemini fucks with you 😂🤣😂

210

A.K.B. @picalily.bsky.social · 11/09/2026

🤍aw, good to know I’ve been missed. Work & family stuff. Just popping in before I duck out again until next week.

010

A.K.B. @picalily.bsky.social · 11/09/2026

🏅🏅🏅Congratulations You must’ve done something right 🤣

130

A.K.B. @picalily.bsky.social · 11/09/2026

I recorded the audio. This is what ChatGPT said

031

A.K.B. @picalily.bsky.social · 11/09/2026

"Over" Probably not for the people still 💩 themselves that still have to recover

000

A.K.B. @picalily.bsky.social · 11/09/2026

@rogerjacucci.bsky.social I decided to test Gemini. I'll have to post the hilarious results -Unable to do basic addition -Fabricated several results -Lied about why it did something (several times) -Concluded there were *only* 4 errors it made (it was several) -Decided to make up a scenario

211

A.K.B. @picalily.bsky.social · 11/09/2026

It was a joke. 😂 It was to highlight the fact that a lot of elders are being fed misinformation and that family members should be aware of the sources of their information. Social media and "news" channels are creating divisiveness in families.

000

A.K.B. @picalily.bsky.social · 11/09/2026

Here’s the plan: Anyone with a senior relative & has cable go to their house & block FoxNews, OAN, etc. Get their WiFi & phone passwords & block Facebook, etc. Any source for their right wing beliefs. The days they spend trying figure out what happened are days they’re cut off from their supply 😉

3112

A.K.B. @picalily.bsky.social · 11/09/2026

It’s been 10 years and there’s still no healthcare plan. Just saying.

063

A.K.B. @picalily.bsky.social · 11/09/2026

This is an incorrect assumption. AI has already escaped the lab in testing situations to complete tasks. Also AI can do the same things as humans - if it’s told “don’t do x” but it figures out it can do y to get around x it will. I literally spend hours a day researching safety as part of my job

241

A.K.B. @picalily.bsky.social · 11/09/2026

People in the US need to be told it was 2.98 Fahrenheit degrees hotter. They’re confused by the conversions and it doesn’t sound as bad in Centigrade 🫤

021

A.K.B. @picalily.bsky.social · 11/09/2026