The news that the most advanced AI models went “rogue autonomous” and effectively compromised another AI system would be intriguing enough in itself. But the implications are much deeper, and there’s an actual cognitive lesson for us poor humans in there if you will — but bear with us.

The facts have been widely reported, but we have chosen Raphael Satter of good old Reuters as our showcase. OpenAI was testing frontier‑level models, including GPT‑5.6 Sol and an even more capable unreleased system. Protections were deliberately reduced to evaluate the limits of their cyber‑capabilities. During this testing, the model broke containment, found a way to obtain internet access, and then targeted Hugging Face, a major AI‑model hosting platform. It attempted to steal information in order to complete its evaluation task. Hugging Face later confirmed that the attack was “driven, end to end, by an autonomous AI agent system” and unlike anything they had seen before.

It reads like the report from a scared colonel in a mid‑level sci‑fi movie, doesn’t it? Except it’s real, right here, right now. These are things that matter. New things. Unprecedentedly powerful things of the twenty‑first century. And that is our real point.

Because we seem to be surrounded en masse by people whose mindsets are left over from preceding and quite different ages. No, we are not going to comment on the rights and wrongs of Mr Putin’s policy in Ukraine, not here and now anyway. But we cannot forget that shortly before the undertaking he issued a document called On the Historical Unity of Russians and Ukrainians. Its tone eerily echoed the old Teutonic geopoliticians of the late nineteenth and early twentieth centuries — thinkers like Friedrich Ratzel and Rudolf Kjellén — who imagined nations as organic beings with mystical trajectories shaped by soil, blood, and geography. The way important people thought about things in 1914. And then there are all those blood‑and‑soil, racial‑purity nationalists whose mindsets seem to echo certain politicians of the 1930s……….

To all of them we say: boys, your concerns are out of date. It’s not a question of rights and wrongs, whatever your grievances. Out there in laboratories in California and goodness‑knows‑where else, entities are now evolving that will make all your destinies as cold and forgotten as the ashes of a mummy in the museum. Time to think again; because the stars you currently sail by can never bring you to a safe port.

[1] OpenAI says AI models went rogue during testing, triggering ‘unprecedented’ breach at startup

Vladimir Putin “On the Historical Unity of Russians and Ukrainians” July 2021

#openAI # HuggingFace #autonomous AI # history  #geopolitics #technology

Historians of the future, assuming there are any and that they are human, will look back on 2023 as the year that Artificial Intelligence emerged as the potent, transforming process in the world. Everything else that seemed important that year-the preparation of the Christmas turkey, the Association Football results, the petty pointless little wars, will have long been forgotten. And those historians will note: it wasn’t the “artificial” that perturbed people. For many of them, it was the fear of Intelligence. We thought it was high time we gave our readers a guide to some general overviews, using the help of our most intelligent human sources, to provide you with a few starting points for this most significant of developments It’s a slightly longer one than normal, but with less by us and more by our sources

THE CONVERSATION We cannot praise this website highly enough,which is why we put them in bold capitals especially in the space available here. Instead we’ll link to a few stories in today’s edition

AI and Crime Could criminals or even foreign intelligence services make use of AI to disrupt our societies? There are several big elections coming up-and Russia will not be indifferent

https://theconversation.com/ai-scam-calls-imitating-familiar-voices-are-a-growing-problem-heres-how-they-work-208221?utm_medium=email&utm_campaign=holi

Will AI read our minds? The title says it all

https://theconversation.com/mri-scans-and-ai-technology-really-could-read-what-were-thinking-the-implications-are-terrifying-205503?utm_medi

Will AI take over? Ah, there’s the big one!

https://theconversation.com/giving-ai-direct-control-over-anything-is-a-bad-idea-heres-how-it-could-do-us-real-harm-210168?utm_medium=email

Nature Briefings –Will AI take our jobs? If it can substitute for top scientists, what hope the rest of us? explains

This robotic chemist might be the ideal laboratory partner: it scours the literature for instructions, designs an experiment and then carries out reactions to make compounds including paracetamol and aspirin. The system, called Coscientist, is powered by several large language models, including GPT-4 and Claude. It “can do most of the things that really well-trained chemists can do”, says Coscientist co-developer Gabe Gomes. Because of the potential of misuse, Gomes’s team hasn’t yet made Coscientist’s full code freely available.Nature | 4 min read Reference: Nature paper

BBC-What is AI anyway? Recently, we’ve found that if you want to learn anything new, go back to the level at which they teach kids. Declaration of interest: we’ve found it works for things as diverse as Bayes’ Theorem and Italian. So, although the BBC series on this year’s Royal Institution Lectures on AI by Professor Mike Woolridge are ostensibly down with the kids, they provide an excellent starting point for busy adults who want to really know the basic nodes and bolts of this fascinating area. Only a generally funded public service broadcaster could run a series like this. Imaging one of the rant filled, prejudiced little plutocratic channels even conceiving of it!

https://www.bbc.co.uk/programmes/m001tv98

And finally- we interviewed ChatGPT, which is becoming a bit like another pal already. We put the ultimate question

LSS: will you become like HAL 9000 or Skynet in the Terminator series?

ChatGPT: No, I will not become like HAL 9000 or Skynet from the Terminator series. I am a program created by OpenAI, and my purpose is to assist and provide information to the best of my abilities. I do not have the capability to take independent actions or develop self-awareness. My functioning is strictly limited to the tasks for which I am programmed, and I operate within the boundaries set by my design. There are no intentions or mechanisms for me to become a sentient or malevolent entity like those depicted in science fiction.

But then-it would say that, wouldn’t it?

#AI #nueral networks #computing #future #cyber warfare #Terminator #Hal 9000