It is really, really genuinely concerning how many people think there are profound measurable differences between these things.
Like yeah tonally I guess there are. But with regard to references and information? You’re literally just using three different slot machines and claiming one is hot.
I suppose though I shouldn’t be that surprised then since Vegas and every other casino on Earth has been built on duping people in that exact way.
> You’re literally just using three different slot machines and claiming one is hot.
It's a fair point. I haven't tested many queries across them all and checked their answers, but if I want to ask one of them a question - right now its Grok just because I trust its answers more.
It's not a methodology problem, it's a test-ability problem. LLMs are not deterministic. You can ask the same question to the same LLM five times and you'll likely get at least 3 answers.
You can meaningfully test if one slot machine hits the jackpot more often than another, just that the methodology should involve a large number of repeats rather than a few anecdotes. There are some LLM leaderboard sites that do it with blind comparisons.
> Grok will absolutely do the same thing another time you try it.
True; it's just not happened yet. It will at some point though. With the Sunnypilot example it right out told me that it is not possible on that fork which I appreciated. The others all seem to hallucinate some setting.
There are dozens and dozens of cheap-looking restaurants in San Antonio with absolutely no online presence that will serve you the most delicious tex mex you’ve had in your whole life
Thanks very much for sharing these notes. I studied algebraic topology in grad school but somehow avoided knot theory entirely. Reading these has sparked that feeling I had when I first got into topology.
This is a naive take. Are there specific instances involving individuals of many nationalities/ethnicities? Yes. Is ICE then ignoring race during its operations? Absolutely not. ICE agents are arresting people based solely on their physical appearance and accents. It is band faced racism.
ICE came into Maine with almost 2000 "targets". They arrested about 200 people. They ended up bragging about 17 "bad guys", and even that list is possibly filled with lies.
Some of the 200 arrested that weren't actually immigrants include a brown man who passed a background check and flew to Texas recently to fulfill immigration requirements to work for our local Law Enforcement. It includes tons of people who are legal residents and had papers on their person to that effect. Those papers are often left behind when the person gets kidnapped, which includes an unmarked van filled with ICE nuts screeching in front of someone's SUV in city traffic, jumping out, breaking the window to the SUV, dragging the man out, and speeding off, leaving a still running SUV sitting in the middle of the street, with papers. A literal kidnapping scene from a movie, but sure, totally normal and upstanding law enforcement activity. Our own cops, not exactly liberals, are finding it hard not to publicly call them stupid assholes. These cops are mostly Trump voters.
Don't stick your head in the sand and cry when people point out how uninformed you are. Their entire operation is almost entirely false positives. They've sent people who live here legally to other countries without authority.
It would be rather nonsensical to completely ignore ethnicity in your operations when the wide majority of illegal immigrants are going to be of that ethnicity. Obviously that would not justify widespread harassment of that group, but nothing like that seems to be happening. Mostly people seem to be trying to stop them from deporting people genuinely in the country illegally, which is divisive - independent of partisanship.
If the DNC has chosen this hill to die on, I don't think they're going to do anywhere as near good as they should do in November given Trump is engaging in some extremely unpopular and foolish behavior that people, again going beyond partisan lines, could easily rally together against.
I completely agree they're a thing, but at what scale? The current administration has deported something like 600,000 illegal immigrants. What do you think their accuracy rate is carrying out those deportations? An accuracy rate of 50% would mean there'd be 600,000 errors. An accuracy rate of 70% would mean we'd expect to see around 250,000 errors. An accuracy rate of 90% would mean we'd expect to see around 67,000 errors.
A quick search [1] on this topic showed 50 people have been wrongfully detained. Even if we increase that figure substantially, it implies an extremely high success rate, which isn't really possible if you're just engaging in widespread fishing expeditions.
Stopped ≠ detained. The government doesn't release stats on who was stopped. Kavanaugh stops are literally about using race as a criterion for the stop. No other probable cause is required.
Biden deported more people then any previous president and did not needed any of that. Fun fact, he even focused on criminals, proving that in fact, it is possible to not be dumb about it.
Meanwhile, what do we have here is complete breakdown of legal process, judicial orders being ignored and agency that repeatedly provably lies about everything. Including about multiple murders. All the accuracies rates you listed are absolutely terrible for anything that wants to pretend rule if law matters.
----------
The article YOU listed shows: nearly 20 children, including two with cancer. 50 Americans detained for being latino and no other reasons. From 130 Americans detained for protesting, 50 had charges dropped or rejected by court. That is so far. These were simply abusive detentions.
These are horrible statistics. In a democratic rule of law country, a few journalists wont be able such frequent and routine abuse of power.
Agreed. But generally it very much depends on the school and the effort of those in and around it. Terrence was very fortunate to have parents who supported him and likely lobbied for his unconventional high school/primary school split education, and equally fortunate that his schools were able and willing to accommodate him.
I use a subset of omz by cloning it and manually sourcing the plugins I want myself rather than initializing the entire omz system. No themes, no checking for updates, etc. For me, it’s the best of both worlds.
Superficially, these look the same, but at least to me they feel fundamental different. Maybe it’s because if I have the ability to read the script and take the time to do so, I can be sure that it won’t cause a catastrophic outcome before running it. If I choose to run an agent in YOLO mode, this can just happen if I’m very unlucky. No way to proactively protect against it other than not use AI in this way.
I've seen many smart people make bone headed mistakes. The more I work with AI, the more I think the issue is that it acts too much like a person. We're used to computers acting like computers, not people with all their faults heh.
>> I also like how you can manage Python versions very easily with it.
>
> I still don't understand why people value this so highly, but so it goes.
Well I do need some way to install multiple python versions in parallel, and ideally the correct python version would be used in each project automatically. I used to use pyenv for this, which puts shims in your path so that it can determine which python executable to run on the fly, but I found that it was sometimes very slow, and it didn’t work seamlessly with other tools. Specifically pipenv seemed to ignore it, so I’d have to use a workaround to point pipenv to the path to the pyenv-installed python executable.
When one tool does both python installs and dependency/venv management, then it can make these work seamlessly together, and I don’t need to screw up my path to make the version selection work either.
> What I was not aware of: `venv`s need to be created with the version of python they are supposed to be run. So you need to have a downgraded Python executable first.
This is one of uv’s selling points. It will download the correct python version automatically, and create the venv using it, and ensure that venv has your dependencies installed, and ensure that venv is active whenever you run your code. I’ve also been bit by the issue you’re describing many times before, and previously had to use a mix of tools (eg pyenv + pipenv). Now uv does it all, and much better than any previous solution.
But they all do that. It just comes with the territory. Grok will absolutely do the same thing another time you try it.
reply