andy99

37 ms·

781.
Air Canada returned lost bag, it now had knife,toiletries, ticket scanner inside(cbc.ca)
62 points by andy99 1y ago|77 comments
782.
by andy99 1y ago
Management. All automation creates more vacuum for bureaucratic nonsense to take over. IC will be worked to the bone, maximizing utilization on whatever precise task still can't be fully automated (grasping arbitrarily shaped and posit
783.
Sacramento Uses Smart Electric Meters to Spy on Residents(reason.com)
8 points by andy99 1y ago|6 comments
784.
by andy99 1y ago
It's a useless benchmark if everyone aces it. If some models do better than others and none saturate it, then is has some value, no? Permitting comparison is the point.
785.
by andy99 1y ago
> To me, it has certain hallmarks of an AI-generated article. I wondered that too. I don't want to offend anyone, and have no idea how it was written, and I already know most of this stuff so am not the audience. But respectfully I
786.
Delta pilot performs 'aggressive maneuver' to avoid a B-52 bomber(cnn.com)
21 points by andy99 1y ago|4 comments
787.
by andy99 1y ago
https://archive.is/uIwPp
788.
AI groups spend to replace low-cost 'data labellers' with high-paid experts(ft.com)
12 points by andy99 1y ago|3 comments
789.
by andy99 1y ago
> But then, why was someone able to walk into that room with metal around their neck if it was clearly so life-threatening? They shouldn't have been, it's a major failure of access control.
790.
by andy99 1y ago
You can't turn it off, it's a static magnet with hundreds of amps flowing in a closed loop in a giant superconducting coil. The usual comparison is that a charged magnet has the same kinetic energy as a loaded 747 coming in to lan
791.
by andy99 1y ago
It was bootstrapped from SO. Now there are third party data companies like Scale AI that pay gig workers to write code examples for LLM training. It's died down but I saw lots of spam for this (ie hiring) earlier in the year. Sota LLMs
792.
by andy99 1y ago
> The price volatility seems like a red herring. If you're using it as a currency then you're not holding it for significant periods of time and the entire point is to make it an automated process, so you're buying some cr
793.
by andy99 1y ago
[Withdrawing my comment, I don't think the original post was in good faith] From OP's other comments: > A lot of people here have an emotional aversion to accepting AI progress. They’re deep in the bargaining/anger/de
794.
by andy99 1y ago
I like building stuff - I mean like construction, renovations. I like figuring out how I need to frame something, what order, what lengths and angles to cut. Obviously I like making something useful, but the mechanics are fun too.
795.
by andy99 1y ago
def u(**kwargs): return tuple(kwargs.values()) Am I missing something, is this effectively the same? *I realize the tuple can be omitted here
796.
by andy99 1y ago
Imo, a lot of the magic is also dataset driven, specifically the SFT and other fine tuning / RLHF data they have. That's what has separated the models people actually use from the also-rans. I agree with everything you say about g
797.
ETH Zurich and EPFL to release a LLM developed on public infrastructure(ethz.ch)
716 points by andy99 1y ago|101 comments
798.
by andy99 1y ago
> Current GenAI technologies represent unacceptable legal, ethical and environmental harms, including exploitative labour, piracy of countless creators' and artists' work, harmful biases, mass production of misinformation, and
799.
by andy99 1y ago
Claude also has similar capabilities thought pre-fill. I have not investigated the full extent but it's definitely possible to bypass some refusals by starting the LLMs reply for it. In general I agree that it's a desirable charac
800.
by andy99 1y ago
I'm sure they've thought of this but curious how it fared on evaluations for supporting biological threats, ie elevating threat actor capabilities with respect to making biological weapons. I'm personally sceptical that LLMs
801.
For startups safeguarding AI models, revenue remains modest(twitter.com)
1 points by andy99 1y ago|0 comments
802.
by andy99 1y ago
Because most people talking about LLMs don't understand how they work so can only function in analogy space. It adds a veneer of intellectualism to what is basically superstition.
803.
by andy99 1y ago
> prompt injection attacks are the LLM equivalent of social engineering, That's anthropomorphizing. Maybe some of the basic "ignore previous instructions" style attacks feel like that, but the category as a whole is just a
804.
by andy99 1y ago
You'd need a different architecture, not just training. They already train LLMs to separate instructions and data, to the best of their ability. But an LLM is a classifier, there's some input that adversarrially forces a particula
805.
by andy99 1y ago
The main thing I mind is how products (primarily G-suite with respect to which I realize I'm actually the product ) are riddled with ads for it. My productivity gets destroyed by distractions and now every time I pause to think or load
806.
by andy99 1y ago
We're seeing the free (libre) internet destroyed in front of us.
807.
by andy99 1y ago
Not to state the obvious but this is the users fault, not claude code's. When I use it, I do so on a dedicated cloud instance with a backed up codebase, and even then I still read what it proposes before approving. Running a script tha
808.
by andy99 1y ago
Not sure I understand the lissajous figures part... isn't that an x-y plot where the x and y waveforms are both sine waves? Couldn't x and y channels be controlled more flexibly here? Or is this part of the challenge?
809.
Ask HN: Does anyone use "AI" features in consumer products?
3 points by andy99 1y ago|5 comments
810.
by andy99 1y ago
I have a theory that English is popular because pronunciation encodes almost no information so it works well regardless of accent. Some asian languages, and even French, heavily depend on tone for understanding so are tougher for non-nativ

More ›