So I have a high schooler home with #covid. I am trying to figure out what sort of protocol to follow before sending him back to school. Spoiler: YOLO
Public school system says “don’t ask us, we do what the Virginia department of health says.” VDH says “don’t ask us, go look at the CDC.” CDC says “you should wash your hands.”
School goes to great pains to say that no one will be forced to wear a mask, no matter how contagious they are nor how protective it is. But no one will be prevented from wearing a #mask either.
Don’t feel sick? Good enough for us. Testing schmesting. Mask schmask.
@ryanc I don’t know whether this works for your situation, but these are called “channel drains” and they are like gutters embedded in the ground. I have one at the top of my basement stairs so the water flows elsewhere instead of down the stairs into the basement.
They can be anywhere from a couple inches wide to several inches wide.
@patrickcmiller That's the key point I'm trying to make. Sometimes people bolt on a proxy in front of a vulnerable web site to mitigate SQL injection risk. We often criticize that as insufficient on its own. And that is exactly what they're doing to prevent information leakage out of LLMs: stick some stuff in the prompt to make undesirable data less likely in the output. And while bolt-on security can be PART of a multi-prong security approach, in the case of LLMs it's the only prong.
This picture from this blog is really useful at showing how our mental models about LLMs can be wrong. On the left is what the user thinks is happening. On the right is what is happening in the LLM. The effectiveness of the bolt-on security is limited by the fundamental workings of the technology.
@patrickcmiller They left out the authorisation model. All #AI systems I have seen have a binary authorisation model: an entity is allowed to inference against the model, or not. Contrast with relational databases where you can have access to some tables and not others. We can even get to row-level and column-level access controls. Just because you can query the database doesn’t mean the whole of the dataset is available to you. Data in the database that matches your query might be missing from your response because you don’t have access to those items.
With an #LLM the entire trained model is available for inference. To put it in #RBAC terms, every distinct role with distinct access to subsets of data would need its own model, trained only on the data they’re allowed to access.
In practice no one does that. So models either include too much data, risking exposure to unauthorised users, or they omit useful data in training because they don’t want the risk. Middle ground solutions are rare and difficult.
@glyph Do they understand it better, or just care more than most that it doesn’t work? I can’t decide whether the boosters really think it works or don’t care.
@skryking I’m glad you said that. Half the time I post on here “here’s the really tedious thing I’m doing” someone posts a super easy way that’s 10 times better. I used to say “here’s the thing I just spent hours doing!” And someone would reply “you can do it in 10 minutes with this $25 tool”
Soldering continues. I have all these JST connectors to do, so I sandwich 2 boards together with a C clamp to hold them all in place. Then I just go one by one soldering them all.
@feld I think that’s overly simplistic. Businesses terminate contracts all the time. Sometimes maliciously, sometimes simply in accordance with the terms of the contract. The real answer will come from whether it is more expensive to fulfill their obligations to the contract, or to pay whatever penalties for walking away. The real people of New York will be left holding the bag if they walk away.
At this moment in history, when historical scientific quality is just down all over the tech world (eg Boeing, Tesla, IT generally) I am not excited about big, one-way door decisions like bringing on nuclear power, motivated by short term fads like AI.
I guess it could be good, though. Use AI to bring on all this nuclear power, then the fad blows over, and when we retire capacity we retire dirty carbon sources.
A “team building exercise” where they abandon a team member at the top of the mountain reminds me of how financial people refer to losing money as “negative growth.”
Cards Against Humanity suing #ElonMusk reminds me of the time The Onion filed an amicus brief to the US Supreme Court. It’s humour doing serious work that really pokes at the so-called serious people who are being ass holes.
Remember, remember on the fifth of November* Corruption, treason and plot I see no reason why corruption and treason Should ever be forgot
* if the UK is going to have its election on 4th July and the US is gonna have its election on 5th November, it’s only fair that they let us borrow their slogans
She managed to get Trump going into random directions: XL pipeline, Victor Orban, now Biden dropping out. And "he (Biden) hates her!" What a crazy tirade. #debate#uspol
Amateur professional #selfhost sysadmin. Professional amateur #cloud #security at #AWS. Also fond of #cats, #cigars, #whiskey and #pipes. I like board games and some video games. I am #covid cautious and I still #wearamask. Opinions are my own, but they can be yours too.