Don't tell Avi Loeb.
4,561 karma · joined August 10, 2016
All this talk of contradictory information, sycophancy, and flattery is little more than anthropomorphic projection in cases like this.
One prompt will elicit a response based on one set of weights that are most closely related to the prompt. A different prompt will elicit a response from a different set of weights, for the same reason.
There’s no contradiction from the point of view of the model and its responses, and no need to invoke sycophancy or flattery as a reason for the behavior. You’d get the same behavior from a model that hadn’t been RLHFd to be sycophantic.
The DSM even has to include an explicit exception to prevent the clinical definition of “delusion” from applying to religious belief. Without that ad hoc exception, religious belief would be classified as clinically delusional.
You're trying to use technology to handle difficulty with minor adversity.
> Maybe the feature is worth the complexity
What feature? Security?
First, that’s not what they were called at the time.
Second, those older systems aren’t what either term actually refers to.
(Selling time on a mainframe would be closer to IaaS than PaaS anyway - in fact IBM’s VM/CMS even used the VM abstraction, which compute in most IaaS environments depend on.)
Your position only works if you ignore the actual extant definitions of what “cloud” or “PaaS” actually refer to.
"Borderline insane" is trickier. Severely cognitively challenged would be closer to the mark.
Right. My own experience was that I paid for a copy of Red Hat Linux, shipped in a pretty box and everything, and then within a year or so discovered that that was the end of the line.
Although technically it was supposed to have been possible to switch to Fedora at that point, the rollout and info around that wasn't at all clear.
My reaction was oh well, guess I'll use Debian - and later, at every company I've worked at in the intervening decades, encouraged use of either Ubuntu or Debian, or provider variants like Container-Optimized OS. No-one ever objected to that, because Red Hat was doing the "enterprise sales" thing, and ain't nobody got time for that.
It's not that GUIs are inherently worse in principle, but in practice they often are.
The point about overlapping windows is that that "desktop" model permeates the thinking about GUI design, but it's fundamentally limiting and misguided.
> we need complex (but not overlapping windows, sure!) interactivity and rich display capability.
Yep. Pity today's GUIs can't deliver that.
OPA can be a component of a secure authorization system - low-level policy definition and evaluation - but there are many aspects of an authorization subsystem that OPA doesn't even try to address.
For a start, policy languages like Rego operate at the policy decision level, they have no awareness of any authorization model. If you're implementing anything with high-level authorization semantics - think MLS, Bell-LaPadula, or pretty much anything else that defines a non-trivial authorization model - you essentially have to translate the semantic properties of that system into a low-level policy decisions.
If that translation is done manually, it's highly error prone, which is not a good feature for a security system. Rego doesn't know what a clearance, classification, security compartment, etc. means. It doesn't prevent a policy author from implementing rules that aren't compatible with an authorization model.
Also, for security to be worth anything, there are inevitable cross-cutting concerns with trusted identity and attribute sources. OPA lets you tell it anything you want, but you still have to integrate that with a larger system that doesn't depend on implicit guarantees. E.g., "This is Bob because the invoking service checked it and says so" might be fine for an e-commerce site, but not for many more secure environments. Same goes for attributes other than identity. Doing all of this in secure ways is the cause of much of the complexity I mentioned.
There's no law with any teeth at that level. Who's going to enforce it?
Of course the ICC might issue arrest warrants for US leaders, just as they have for Netanyahu, but the practical implication of that is mostly that ICC members will be subject to attacks of all kinds.
> Imagine if Putin was using those on Ukraine civilians but US uses them no problem?
Since neither of those things are true, what do you think this unlikely thought experiment might tell us?
It's interesting, and concerning, to see the exact same thinking in the US today.
The claim was that theoretical, mathematical models of physics, like special relativity, were a product of "Jewish thinking," designed to deliberately confuse the world.
There's a recent video in which Tucker Carlson interviews Jess Elofson and they mutually agree that algebra is nonsense, with Elofson calling it "fake math". They didn't go so far as to relate it to some ethnic source (from the part I watched), but "designed to deliberately confuse the world" is the big selling point. The people they hate are people smarter than them.
On the plus side, this is why these anti-intellectual movements always fail, because in the end they just can't get much useful done, other than attacking and destroying which doesn't require finesse or understanding.
Edit: I have to retract what I wrote about not relating it to an ethnic source. Elofson literally used the term "goy slaves" in that interview.
It depends what the authorization requirements are. For consumer-grade security, you're right. But when you start getting into the requirements of large, security-sensitive organizations, there's a lot of unavoidable complexity.
But those contexts often influence the development of security systems, so some of the complexity you're referring to relate to requirements most users don't have.
"Cloud" today is typically used with a more specific meaning, to refer to services such as IaaS or PaaS, which didn't exist in the 90s.
I don't remember encountering the word "cloud" in the 90s used in the way you describe. The cloud could be a symbol for a telephone network, a packet-switched network (including the internet), a company WAN, etc. Typically you'd refer to whatever the cloud symbol was supposed to represent, it wouldn't be called "the cloud" unless I suppose you were discussing the diagram itself.
Terrorist Israel’s pager attack.
It’s foolish to be afraid of nonexistent threats.
All the AI-generated UIs I've seen have been very derivative, certainly not eliminating any of the disadvantages of typical GUI interfaces.
Realistically, the whole "overlapping windows" GUI model, and everything that derives from that, was a metaphor geared towards people who'd never seen a computer before. It fit the increasing consumer focus of computing interfaces. It's no wonder that technical people often prefer TUIs.
Maybe AI will bring real advancements in GUIs, but someone's still going to have to make it happen.
I've been using Gemini on development of a DNN training pipeline, and there's no way you can describe it as "dumb as hell". That description just makes it clear that you're not talking about the technical capabilities of the models, but about some sort of fanboy comparison from a parallel hype universe.
Perhaps, but not the way Flock does it. Transparency, oversight, controls - all essentially just missing.
"Disruption" to these people means "try to sell the most legally questionable service possible until we get slapped down by the society we're trying to extract money from."