Her comments on the nature of the threat from Russia and China are well put and stand up to analysis. That she was stating these things in public suggests that she wants the politicians to stop dithering and she is correct about that.

Her comments on tech seen naive. The tech world won't take her seriously and with good reason.

I would call them a bad idea. Not just the glasses but wearing them in public. Anyone who wears them is just going to get punched and or shit on people around them.

Yes. There is clearly more than one "cause", but the proximate cause is chemicals tricking the immune system into attacking specific cells.

The wall of text you referenced has the following sections in bold text adorned with the usual arguments.

"4. We can't just "decide not to build AGI"
"5. We can't just build a very weak system"

I've been at this for a long time and have read many doomer papers and blogs. To their credit some of them actually acknowledge the obvious - science is presently unable to answer this question. There is no feasible way to test a hypothesis that involves something that doesn't exist turning everyone into paperclips or any other doomsday scenarios. There is no way to derive statistical probabilities of outcomes when you have no basis for deriving them in the first place. Hand waving and guessing is not science even if domain experts are doing that hand waving.

If you believe there is an objective basis for deriving a statistical probability of clippy the paperclip maximizer turning everyone into paperclips or any of the other doomsday scenario please do provide the objective basis for such a derivation. Don't just throw a link with a meandering list of AI safety related arguments which are not responsive to your science claim. Quote chapter and verse and make your case. If you are unable or unwilling to do that then I'm just going to ignore your "science" assertions as unfounded.

I did not refuse to read the following reference https://ancillary-proxy.atarimworker.io?url=https%3A%2F%2Fwww.lesswrong.com%2Fpost.... I refused to care about the signature list in your previous reference which was clearly polluted with abject nonsense.

Well, peak rate of extinction is not the same as "all the glaciers gone."

And, glaciers gone just means that the flow rate will equal the precipitation rate, rather than the precipitation rate delayed until the spring thaw.

Disabling RC4 is a deflection from Microsoft's refusal to deploy secure authentication and authorization technology such as ZKP.

This does NOT solve the problem it merely increases the cost of Kerberoasting roughly 1000x (No thanks to anything inherent in RC4 vs. AES) this increase means very little in terms of real world outcomes where this type of scaling is the difference between spinning up more threads on more systems or waiting hours or days instead of seconds and minutes.

I'm still writing new software in .Net8 Winforms.
Now get off my lawn!

Your claim buddy I merely called you on it. If there is a standard for who can sign what their qualifications and experience must be that would be one thing yet obviously no such standard exists so why should anyone give a flying rats ass about the number of signatures when the count is absolutely meaningless?

I simply don't care enough to look, as I said in my prior commentary I stopped reading. This wasn't a figurative statement it was literally me seeing abject nonsense and having 0 inclination to care about or trust your source from that point forward.

Science is process not a destination. Science is prediction not decision.

Even if a bunch of domain experts get together and vote on subjective feelings about possible outcomes this is merely informed guessing not the exercise of science. Further blatant overstepping by suggesting courses of action is squarely in the realm of policy and politics not science.

I fundamentally disagree. There are several basic problems.

1. Doomerism is completely unbounded. There is no objective basis for any of the doomsday outcomes with no way of even estimating a likelihood of occurrence in any objective framework that does not devolve entirely into personal feelings and opinions. All the doomers ever say is that something can happen. A meteorite can put a hole through my keyboard and make it hard for me to type. Should I buy a new keyboard in advance just in case? The answer to this question is a probability that can be quantified by following a scientific processes. Doomerism on the other hand is limited only by ones imagination and is inherently resistant to falsification and scientific inquiry.

2. Complete misunderstanding of source of the threat. If you a-priori presume AI genies are an existential problem the actual source of that problem is the underlying knowledge and industrial base that makes the production of AI genies possible. As the cost of producing an AI genie is reduced with advances in knowledge and industry it will eventually come to the point where AI genies are not just the province of governments and large corporations but of small groups, death cults and eventually rich kids with nothing better to do.

3. Promulgation of absurd and transparently pathetic solutions involving trusting large corporations with a track record of being completely corrupted by simple human greed let alone swindled by a super-intelligence.

This also requires accepting the even more absurd notion of people thinking they can control anything resembling ASI or that at some point there won't be a security breach. We don't even know how to make existing AI models in a way that are "safe" when an adversary has access to underlying weights. As it is now every single time a model has ever been released whatever safety / alignment precautions were applied are trivially in a matter of days "abliterated" using numerous software and techniques to systematically remove that alignment. If you are going to say lets all have the fate of the world depend on a bag of weights not ever getting into the wrong hands... my response to that is f*** off.

4. The outrageous persistent refusal of doomers to ever just fucking say NO. Just stop all the madness by simply outlawing the technology or even as your reference puts it slow it down. This seems like a reasonable request if one were to presume even a 1 in 100 probability of clippy the paperclip maximizer turning everyone and everything into paperclips otherwise.

Since the problem is knowledge and enabling industrial base while it is unreasonable and unrealistic to force or expect everyone to stop or impose some sort of CTBT the world could at least create a regime where countless trillions of dollars in capital and millions of people were not activly, enthusiastically working to end the world. Even if this in the distant future were to ultimately fail due to advancement of dual use hardware and software it would at least delay doomsday which is better than nothing or more importantly lead to a situation where multiple AI genies arrive at the same time instead of just one due to ease of production retaliative to a future knowledge and industrial base.

5. Misattribution of threats to AI that are actually created by the advancement of classical technology having everything to do with humans and nothing to do with AI doomerism. The most salient technological threat in this category is the availability of software and hardware to manipulate and fabricate biological systems including pathogens. What was previously the province of large institutions can now be done with small groups and small budgets and in the future by single individuals with ever decreasing knowledge and monetary inputs. The Aum Shinrikyo doomsday cult is a classic example of such threats where members were able to manufacture nerve agents. If AI genies are possible it is only a matter of time before the next Aum Shinrikyo gets their hands on one and all the careful attention to detail in the world on behalf of the AI firms isn't going to mean jack shit.

If you want me to take this absurd full steam ahead we can control ASI nonsense you are peddling seriously while concurrently being unwilling to consider a total ban then my position is you are just another industry whore who cares more about AI profits than anything else.

The most immediate threat from AI by far is other people not super intelligence. It is that threat I believe we should be prioritizing by not allowing for AI protectionism or hoarding of technology in the name of safety.

Good job, comrade!

Why specifically would a conservative not boycott Smith & Wesson if they did something to upset them? It's not like Smith & Wesson is the only gun manufacturer.

There is a concrete example from a couple years back of this very concept in action. Some conservatives love their beer yet it didn't stop them from boycotting Bud Light. Apparently you can still like beer, still drink beer and at the same time stick it to the brand that did something to annoy you.

You're not making any sense.

I stopped reading at Steve Bannon and Glenn Beck. What a fucking joke this is.

AI is apparently Susan Rice's life's work.

AI indexes are measuring capabilities of AI systems not Moore's law. You can say moores law is responsible for enabling hardware industrial base but this doesn't change the nature of the thing being measured.

The more training a model of a given size the better answers it gives. The more compute you can afford the more you can afford to train the model.

This is like saying everything is caused by god and just as useful. Humans are getting better at training AIs resulting in AIs that are more useful and more capabilities. Majority of a models capability and compute budget takes form of pretraining rather than post training where CoT et el is applied.

This is conflating knowledge with intelligence. What makes AI useful isn't what they know but rather an ability to generalize and apply their experience to new situations. LLMs for example know way more than any human does and their perplexity scores are at least an order of magnitude better than human scores yet nobody would say they are more intelligent than humans.

I don't think this is a salient issue. Either you get a useful answer or you don't. If what you ask for isn't checkable and you have no way of ever evaluating the real world performance of the answer by putting it to use in some way then what was the point of asking in the first place?

Yep as time moves forward it gets easier and easier for everyone to create their own AI genies. It is ultimately the enabling knowledge and industrial base that matters not how many compliance boxes are checked or how many people on your red team.

I can respect the rare doomer who advocates for blanket AI bans. This at least has some logic to it. While it is infeasible to detect when people are breaking the rules trillions of dollars in capital flows and large scale access to enabling knowledge can't be hidden.

The typical doomer never advocates for stopping. It is just more of the same bullshit of protectionist regulatory hurdles that stand no chance of preventing either the emergence of AI genies or the granting of wishes to different masters. AI companies have already established themselves as wholly untrustworthy power seeking whores (no offense to actual whores)

I liked the Nazi themed stickers and slogans. Swasticars, 0 to 1939 in 3 seconds, fascist thing on four wheels, Tesla stylized KKK hoods...etc.

It's 2025. There are plenty of EVs on the market that don't benefit Musk or Trump's incompetent attempt at a self coup. There is no need to decide. You can for example "sell your car", still have an EV and fuck over Musk all at the same time.

You can say "I purchased this Tesla before Musk went crazy" is a weak protest when you can sell your car and make a more powerful statement yet it was never at any point a choice between green agenda and opposition to authoritarianism, sociopathy and incompetence.

Likewise people can stop paying to use LLMs. The open source models do the same shit and cost less to run than paying for an OpenAI subscription. Instead of bitching about the trillions being funneled into this crap people have the power to simply choose to ignore it. More concretely it is in everyone's interest for the AI bubble to pop sooner rather than later.

So, not a permanent Federal employment with its guarantees? But perfect for bribery and grift. Why this doesn't surprise me?

You do it by not jumping to baseless conclusions. People are capable of thinking abstractly and recognizing their own biases.

The general answer to the corrupting influence of power is systems of governance where power is constrained by power. A states imposition of this type of gating who is too stupid or unfit to access information or communicate regardless of intention is certain to lead to further aggregation of power.

Like predictions of the future nobody has any way of predicting what ASI can do. It is very much still an open question the value higher intelligence brings to accomplishing relevant tasks relative to value of doing the required work.

LLMs of today are moored to their training and structurally can't evolve outside the confines of their limited STMs. Likewise human minds are moored to their genetic histories. The evolution of a super-intelligence which presumably have no such constraints is fundamentally unpredictable.