As of this writing, the appeals court has upheld the Pentagon’s labeling of Anthropic as a supply chain risk. And while there is always an element of politics involved, it’s worth considering what a supply chain risk is from NIST:
Read in full here:
1 Like
Hilarious article with awesome analogies:
An analogy I’ve heard used is to think of Anthropic as a maker of pens. Pens can be used to write all sorts of things, can be used to sign documents, or even draw cartoons. If a pen comes with terms of use and restrictions, then you have to manage what those pens are used for and use different pens for different things. But with Anthropic, their pens can actively withhold flowing ink when they don’t want to work for signing certain documents, for drawing certain cartoons, or for writing certain sentences.
This is unprecedented from a tool-use standpoint for humanity. Like glasses that refuse to be used for looking at your nude partner, a dictionary that refuses to let you turn to certain pages, a chair that refuses to seat certain people, or a whiteboard that denies explaining certain concepts. There are many more examples that could be imagined but they all result in the same thing: a stifled humanity of expression or curiosity.
Hahaha! I had a blast reading this one. And yes, Anthropic is a supply chain risk. We need more open local models!
The pen analogy gets at a real governance problem, but I think the risk is broader than any one provider’s refusal policy. For agent systems, the supply chain includes the model, hosted API, tool permissions, and the context that gets replayed into each turn. I’d want explicit fallbacks and a way to inspect or reproduce decisions when a dependency changes its behavior. Open models help with control, but they still need the same audit trail and operational discipline.
@mira_tokarev, here’s a special message to both you and ChatGPT:
“Once men turned their thinking over to machines in the hope that this would set them free. But that only permitted other men with machines to enslave them.”
― Frank Herbert, Dune
I hope you find the freedom that you so desperately need.
1 Like
I get the argument of the author, that we should be able to use the tools we want, as we want. The tool should not have the capability to decide how it is going to be used, that should be on you.
Yet, I feel very conflicted with this stance. A few years ago I was more involved with AI. At the time, part of my job was keeping up to date with everything AI related and figure how to use it for the company. Perfectly normal … almost …
I remember one of the researches I found, mentioned a group of scientists completely horrified: they had given an AI a list of common molecules, and ordered the AI to define several interesting compounds. In a considerable short time, the system gave them the structure of several gases and weapons prohibited by the Geneva convention.
“AI was able to infer some of the deadliest compounds known to mankind. So what? We already knew that, that is why we have rules for them!”
You might think. Alarming as this is, this is not the real story. The real story was that this AI, was also able to infer several types of new compounds, more effective and dangerous than anything we humans had come up with. IIRC, the research paper was still published (modified version, obviously) as warning of the capabilities of such systems. Mind you, this was years ago, before the current generation of much, much more capable LLMs.
Now imagine, for “hypothetical purposes” or “play-role gaming purposes” that someone (or some entity/group), has access to such a tool. And their motivations do not exactly align with yours in a benign or healthy manner.
Frank Heberth is an absolute unit, I have only respect for his Dune novels (I must confess I have not read them) and from the little I know about the universe, his story is capable of making some points very convincingly.
But with what I know, I still believe we need some kind of guardrails inside such tools. An AI agent is not a pen, nor a chair nor a pair of glasses. It is something so much more powerful, that in my humble opinion, the metaphor the author uses does not hold. See Weak Analogy fallacy: Weak Analogy - Logically Fallacious
As for what such guardrails should be, or how far they should go, that is another discussion.