cubefox

8 ms·

151.
Pfizer, Valneva Lyme vaccine starts review in Europe(pharmaphorum.com)
1 points by cubefox 2mo ago|0 comments
152.
by cubefox 2mo ago
> How would that work? Claude appending " written by AI" to each of its messages? That would both be impractical and useless.
153.
by cubefox 2mo ago
We were talking about text.
154.
by cubefox 2mo ago
Usually when people say "not finitely axiomatizable" they just mean first-order logic. For example, Peano arithmetic is finitely axiomatizable using second-order logic, which allows quantification of predicate variables, which is
155.
by cubefox 2mo ago
> The EU regulation, from my understanding, requires AI content to be labeled for human viewers. How would that work? Claude appending " written by AI" to each of its messages? That would both be impractical and useless.
156.
by cubefox 2mo ago
This comment disagrees with you: https://news.ycombinator.com/item?id=49324387
157.
by cubefox 2mo ago
This is just wrong. There is a huge difference between a book that doesn't explicitly tell you how to create a deadly virus, and an infinitely patient future LLM that leads you through every step and every unexpected problem and every
158.
by cubefox 2mo ago
Not finitely axiomatizable using first-order logic, I assume.
159.
by cubefox 2mo ago
By the way, Google Translate is really good nowadays. It would certainly not make such mistakes anymore. In my experience, for long text it is even outperforming conventional LLMs in terms of faithfulness: Some LLMs tend to produce overly l
160.
by cubefox 2mo ago
The park orienteering course at the end reminds me of a similar development in game design: Even before most people had GPS phones in real life, many video games already had a form of virtual GPS: A map with a marker of where the player is
161.
by cubefox 2mo ago
In the past, people were also heavily relying on road signage or stopping and asking locals, especially when not carrying a map.
162.
by cubefox 2mo ago
I'm pretty sure there is sufficient general research which says that if you don't train a skill, it gets weaker.
163.
by cubefox 2mo ago
Which seems to contradict the usual consensus that purely linear architectures are not sufficiently capable and unsuited for frontier models.
164.
by cubefox 2mo ago
Unless they are using a linear architecture, the compute cost still scales O(n²) for n tokens, and nemory cost scales O(n).
165.
by cubefox 2mo ago
I would expect this only to be true for linear architectures like Mamba or Gated DeltaNet. Transformers and hybrid architectures do not have constant compute cost per token.
166.
by cubefox 2mo ago
> Data and content related to "biological weapons" already exist on the internet, in books, etc. The real issue is access to facilities and tools. No, I think tools are easy to come by (unlike in nuclear research), the real iss
167.
by cubefox 2mo ago
There is potentially a new Lyme vaccine coming out in about a year. The review process in Europe just started. https://pharmaphorum.com/news/pfizer-valneva-lyme-vaccine-st...
168.
by cubefox 2mo ago
> To give a couple of examples, using this would be a bit like locking the glove box in front of traffic cops, or going around the house locking rooms when you have a guest. That's a good point. Doing this basically says "I dis
169.
by cubefox 2mo ago
> Due to the intent system on Android allowing a pinned application to open another one (for example by sharing something), I tested it and it blocks opening other apps when pinned. Maybe blocking app opening doesn't work for apps u
170.
by cubefox 2mo ago
I don't understand why this isn't activated by default. It just adds an extra "pin" option when you tap the app icon.
171.
by cubefox 2mo ago
This is a silly article. Both Android and iOS have exactly this feature.
172.
by cubefox 2mo ago
> Open Source: We will release the weights in two weeks after launch, once safety evaluation and hardening are complete. What safety evaluation? What safety hardening? They already evaluated it and found it to be highly capable at exploi
173.
by cubefox 2mo ago
You are confusing his book Begriffsschrift ("concept notation"), where he invented what became modern predicate logic, and his later work "Grundgesetze der Arithmetik" (I/II) which (unsuccessfully) tried to derive a
174.
by cubefox 2mo ago
It blows my mind that most HN users are still impressed by algebraic types but haven't heard about set theoretic types.
175.
by cubefox 2mo ago
You can use the Gemini API which is independent of the more complex Vertex AI API. Not sure whether you still have to visit the Google Cloud UI for some things (like billing) though.
176.
by cubefox 2mo ago
Especially not competitive at software engineering.
177.
by cubefox 2mo ago
Or a Disney animation.
178.
by cubefox 2mo ago
Probably not if you account for algorithm size.
179.
The Aaronson Oracle(aaronsonoracle.com)
1 points by cubefox 2mo ago|0 comments
180.
by cubefox 2mo ago
This shows that prediction algorithms (AI models) are also very good at compression, but compression algorithms (like the ones used in gzip) are not likewise very good at prediction. Which is evidence that compression is necessary but not s

More ›