yeah, for a moment I was reading it as being a holomorphic encryption type setup, which I think is the only case where you can say 'provably private'.
It's better than nothing, I guess...
But if you placed the server at the NSA, and said "there is something on here that you really want, it's currently powered on and connected to the network, and the user is accessing it via ssh", it seems relatively straightforward for them to intercept and access.
Just a shame they spent so long skimping on iPhone memory. The tail-end of support for 4gb and 6gb handsets is going to push that compute barrier pretty low.
Eh, maybe a bit, but those era devices also have much lower memory bandwidth. I suspect that the utility of client models will rule out those devices for other reasons than memory.
Not really? The A11 Bionic chip that shipped with the iPhone X has 3gb of 30gb/s memory. That's plenty fast for small LLMs if they'll fit in memory, it's only ~1/3rd of the M1's memory speed and it only gets faster on the LPDDR5 handsets.
A big part of Apple's chip design philosophy was investing in memory controller hardware to take advantage of the iOS runtime better. They just didn't foresee any technologies beside GC that could potentially inflate memory consumption.
would you be willing to guarantee that some automation process will never mess up, and if/when it does, compensate the user with cash.
For a compiler, with a given set of test suites, the answer is generally yes, and you could probably find someone willing to insure you for a significant amount of money, that a compilation bug will not screw up in a such a large way that it will affect your business.
For a LLM, I have a believing that anyone will be willing to provide that same level of insurance.
If a LLM company said "hey use our product, it works 100% of the time, and if it does fuck up, we will pay up to a million dollars in losses" I bet a lot of people would be willing to use it. I do not believe any sane company will make that guarantee at this point, outside of extremely narrow cases with lots of guardrails.
That's why a lot of ai tools are consumer/dev tools, because if they fuck up, (which they will) the losses are minimal.
One of the best benefits of the current no live nuclear testing treaties / environment, was that the United States was one of a few countries that had done extensive live tests early on.
The United States is able to sit on its arsenal and data, and with extensive research and simulation validate to a high degree of accuracy that "hey our bombs still work".
Most countries do not have the data/technical expertise/resources to be able to validate with just simulation. But since no-one else is doing live tests, they do not do live tests either.
How much do you want to bet that a subset of the Russian nuclear weapons simply do not work, and that they will only figure this out when they need to 'test' in response to American tests.
> How much do you want to bet that a subset of the Russian nuclear weapons simply do not work, and that they will only figure this out when they need to 'test' in response to American tests
My bet is that most of them are in disrepair. Russia spends around 8 Billion USD on nuclear weapons. France spends around 6 Billion USD on nuclear weapons. Difference is that France has something like 200 warheads, while Russia has something like 5500 warheads.
Furthermore the fact that using of nuclear weapons has extremely low probability of happening is giving a massive space for corruption. Why maintain what you are not going to use? They managed to siphon money from maintenance of armored equipment, why not ICBMs?
We can get to the staggering reality like Russians have less than 100 working nukes and they themselves may not even know which one are those from those 5500
>My bet is that most of them are in disrepair. Russia spends around 8 Billion USD on nuclear weapons. France spends around 6 Billion USD on nuclear weapons. Difference is that France has something like 200 warheads, while Russia has something like 5500 warheads.
US spend: 57 billion USD; US GDP: 29,000 billion. US spend on nukes as % of GDP: 0.19%
Russia spend: 8 billion USD; Russia GDP: 2173 billion. Russia spend on nukes as % of GDP: 0.36%
France spend: 6 billion USD; France GDP: 3174 billion. French spend on nukes as % of GDP: 0.18%.
That's why I put as % of GDP. Russian nukes cost rubles, not USD. The numbers suggest pretty conclusively that the Russian arsenal of 5500 is maintained about as well as the US arsenal of like 5300.
Your "staggering reality" of 100 working missiles is completely delusional.
I think there is a degree of trade off in this, yes a nuclear scientist/engineer/technician in Russia or China is cheaper than in the USA. But also, the people with those kind of skills (or those technically competent enough to do a good job, are going to be expensive no matter what.)
At some level when people have enough technical skill to do these jobs well, they also have enough technical skill to leave the country and go elsewhere and do something else for better quality of life.
Like GDP per capita in china is much lower than the USA, I bet that their nuclear program engineers are getting paid at least ~80k range, which while less than the equivalent engineer in the USA is paid, is not the same level as what a direct PPP comparison would give.
There are a lot of software engineers in UK who made $50k or less who could have presumably moved to the US a make a lot more but never did. Lots of government employees making much less than they could. Patriotism, wanting to live in their own country, wanting to work on interesting things, etc.
Again, Russia has much smaller economy than USA. What are you searching for is PPP And no amount of PPP will help you to have 5500 warheads on 8 Billion budget vs 200 warheads on 6 Billion budget.
With your logic Kongo should be able to afford 5500 nuclear warheads just by spending 0.4% pf GDP. That's not possible is it?
Or, they do have a 100 working ICBMs and they do actually know which ones are those. The rest of the warheads in storage are not really maintained. Russians are corrupt as hell, but they are not actually incompetent when they need to have something working.
Nuclear non-proliferation only really works when no one feels the need for nuclear weapons in the first place. As soon as countries start feeling threatened or distrust each other, the whole idea falls apart. It’s easy to agree on disarmament when everyone feels safe, but when fear enters the picture, every nation starts looking for its own button to press.
Sadly, the world learned this lesson the hard way from Ukraine’s example: a country that gave up its nuclear arsenal for security guarantees, only to be invaded by the very power that signed them.
Now, earnest question: what happens to the nuclear engine fuel at end of flight? While it certainly won't become critical it's likely the only part it stops being is "engine fuel".
(Of course the question above is irrelevant the moment the missile is fitted with a nuclear warhead)
Yup. The last test from Russia was in 1990. China was in 1996. China was much less advanced than now compared to the US (proportionally) and Russia/USSR was into a crisis and didn't even exist as a country (Russia) back then. The US is just doing another gift to Russia and China.
For the Russians it would be a mistake to rely on the unreliability or inferiority of their weapons - they historically are very adept at addressing those with sheer numbers.
It's better than nothing, I guess...
But if you placed the server at the NSA, and said "there is something on here that you really want, it's currently powered on and connected to the network, and the user is accessing it via ssh", it seems relatively straightforward for them to intercept and access.