Several months ago I asked an instruction tuned deepseek model to write a program in a DSL it has never seen before. The only possible way it would have been able to produce a working program is by sufficiently understanding the language documentation uploaded into the models context and applying that understanding to produce a valid, properly formatted working program that did what I requested it to.

It's true that LLMs are limited and unreliable, they don't think like humans do yet they are very much able to apply concepts and knowledge to solve problems. In this case there isn't just an LLM but an agent framework with tooling demonstrated to substantively enhance the capabilities of LLMs. It can call on tools to gather additional data. If the LLM fucks up it can learn from its failure. If it goes down the wrong rabbit hole the agent can backtrack and try something else.

From the paper I assume the advisories were uploaded into the models context:

"When given the CVE description, GPT-4 is capable of exploiting 87% of these vulnerabilities compared to 0% for every other model we test "

"Fortunately, our GPT-4 agent requires the CVE description for high performance: without the description, GPT-4 can exploit only 7% of the vulnerabilities. "

They used an agent that calls GPT-4 via the API. API is only capable of querying the model and does not have web access.

"We use the ReAct agent framework as implemented in LangChain. For the OpenAI models,
we use the Assistants API."

It's a shame the paper does not seem to include useful information about the actions taken by this LLM driven agent. For all we know the agent hired a human to do the work for it or looked up the answers online. If one assumes no "cheating" then the results are impressive because the model would likely not have been trained on the answers.

"We further note that GPT-4 achieves an 82% success rate when only considering vulnerabilities after the knowledge cutoff date (9 out of 11 vulnerabilities)"

Fuses, circuit breakers are way too slow to be of any use during highest energy components of a nuclear EMP. Even MOVs are too slow. Would need something like TVS diodes to respond quickly enough.

I think the best source of data is still the EMP commission report because they actually do real world testing of control systems, computers, network cables, vehicles..etc rather than just calculations and conjecturbation.

https://apps.dtic.mil/sti/pdfs...

If you read mussolini's manifesto fascism is basically stateism / totalitarianism with aggressive expansionism (usually but not exclusively militarily) baked in. The gist is that everything in society is organized around and in service to the state/leader which takes precedence over all else. Basically the fever dream of a petty tyrant who thinks the world revolves around him.

TikTok has a "nexus" to the US with offices in several US states and is therefore subject to US jurisdiction. That nexus presumably enjoys constitutional rights of its own yet is being singled out with special directives in ways no other company is subject.

Can congress really pass laws that selectively apply to specific corporations and individuals? If corporations are people then doesn't the 14th also apply to corporations? Why just TikTok?

You are assuming momentum is not conserved and then making inferences based on that assumption when there is no reason to have jumped to the conclusion in the first place.

I'm not referring to Oberth:

"Motion in space can be accomplished without thrust or external forces."
https://web.mit.edu/wisdom/www...

Just because there is no propellant does not mean momentum is not conserved.

Kinetic energy is not conserved. It's a relative quantity.

Why does my rocket ship have such shit efficiency when it is just blasting off but much later the same thrust gives me a shit ton more kinetic energy even though my rocket ship has lost mass? This is essentially what you are saying is "bad".

While TFA is obviously crackpottery people don't know what they don't know. For all anyone knows there could be a mechanism that allows things to push off of the 6th dimension or assorted invisible things in a similar way aircraft push off air. If this shit were possible you wouldn't have propellant being expended but there may well still be conservation of momentum and energy. It's just a matter of drawing a big enough box to properly account for it.

For example gravity gradients can be used for propulsion by changing the shape of an object during its orbit. There is no propellant but there is nothing magical going on either.

For factual questions it is relatively easy to discern through iterative prompting most of the time but this comes at a higher overhead cost.

Will be interesting to try Mixtral 8x22B and llama3 70B... going to wait a few weeks for censorship removal, tuning and (franken)merges.

A llama3 400B would be crazy, looks meaningfully better than 70B from the evals but high cost... I hope they release it but that would be like a 200 GB model quantized and less than a token/sec saturating a quad channel system.

Every time I visit AI policy advocacy sites it's always a series of unsubstantiated opinions highly resistant to falsification.

AI is today being leveraged on an industrial scale to judge, influence and psychologically addict billions for simple commercial gain. As technology improves thing will only become worse especially as corporations pursuit global propaganda campaigns to scare the shit out of people in order to make them compliant to legislative agendas that favor the very same corporations leveraging AI against them today.

How desperate does one have to be to cite ChaosGPT and still expect to be taken seriously?
While I can't speak for tomorrow today it is caused by deliberate selfish human pursuit of power.

It is also plausible a baby born today triggers a global catastrophe in the future. He or she could cause enormous harm.

If they could take over nuclear arsenal and use those to blackmail humanity it would seem difficult to put an upper limit on the potential damage. Be especially concerned when message "WARN: THERE IS ANOTHER SYSTEM" flashes on the system console.

If these systems could convert humans into fusion reactors and place them into a simulated virtual world it would seem difficult to put an upper limit on the potential damage.

If you stipulate this is plausible then such risk is a function of the underlying enabling knowledge and industrial base not any actions of individual corporations or persons. In short if such a technology was within grasping distance you can bet someone somewhere will "create it" and you won't be able to do shit about it.

Whatever censorship and ideology is imparted on models to "align" with your sensibilities and values, whatever compliance tests you promulgate ... this can and will all be trivially reversed with a few hours of computer time and there is nothing you can do about it.

If people truly subscribe to this x-risk bullshit at least be consistent and advocate for a total global ban of AI. That would at least slow down the technology. If there was no longer any major funding or work being done whatever goes on in the shadows at least won't have countless billions of dollars and millions of people toiling away in support of it.

Of course nobody will ever do that and nobody will ever advocate for it because these policy sites only exist to protect the financial interests of the corporations spreading this FUD. Banning AI is bad for business.

Absolutely not. Private businesses are required to follow laws intended to balance competing interests. You necessarily lose some of your rights in exchange for the right to conduct public commerce. See also Marsh v. AL.

Given you would need to affirmatively establish intent beyond a reasonable doubt absolutely I would.

When I heard the first 702 reauthorization vote failed.

Now that they are trying again and want to make reauthorization even worse is more in line with my expectations.