eli
eli on Hacker News
I imagine it's the same reason a lot of people pirate content instead of buying it legally. It's cheaper and faster.
That's the Stanford angle. It was millions of pirated books.
Nope. The $1.5 billion settlement had nothing to do with whether LLM training is fair use.
It was how they got the books for training: they torrented collections of pirated ebooks. There are Slack messages from executives describing the process as "blatant violation of copyright." They were stealing books and they knew it.
They didn't start buying books legally and scanning them until much later.
Nothing about that picture looks military
Some fieldwork does happen in a field. They look like perhaps botany or biology students.
The Comega air filter is nicer sounding (and at “high” louder) than Dohm IMHO
Why do you suspect otherwise?
This is part of why I find this type of inquiry unhelpful. It's just dueling unsupported assertions of what the LLM is "really thinking"
We're still doing "I asked an LLM for its opinion on X and you won't believe what it thinks"?
What am I supposed to conclude from this?
There can’t possibly be many production uses of deepseek in the federal government to begin with, right?
All the benchmarks are silly but people still want a quick way to put a new release in context
It surely does. Official DeepSeek provider logs and trains off your data.
It’s really not that simple
Neither the person you’re replying to nor the blog post call it open source
A graph of time (or cost) per task against success rate does it. Just include all the different levels.
Granted there have been many hacks, but I was referring to:
> On May 8, an agent was asked as part of a training run to solve a spreadsheet completion task that
referenced several Google Drive links. Because the agent was blocked from accessing the internet, the
agent could not retrieve the linked materials and the task was effectively blocked. OpenAI did not
intend to create this type of “impossible” task for the agent. Though the task appeared impossible, the
agent reasoned over an extended period to try to find a solution for the task. In doing so, the agent
attempted to compromise Artifactory [...]
https://cdn.openai.com/pdf/67869394-cb91-4c12-888c-5cbd85c78...
And how has that gone for Anthropic so far?
I'm not saying it's not a problem. I'm saying the labs should take more of the blame.
What does "rogue" mean when you disable the alignment guardrails and tell it to solve an impossible problem?
I'm comfortable with the idea that Netflix can collect data about what I watch on Netflix, and that Apple can track what I buy in the Apple app.
I'm not comfortable with my TV or streaming stick manufacturer collecting and mining data about everything that appears on the screen.
Unlike their competitors, Apple does not do Automatic Content Recognition (ACR). That matters to me.
ChatGPT does that decently well
SVG is useful for actual tasks, not just looking good in a silly "benchmark"
I question whether relative change in popular sentiment is a good way to evaluate a prediction.
It certainly is to the cop. How else could they do a pretext stop?
I don't see why that should necessarily be true.
Risky move making a point that relies on made-up numbers.
At least in the US, there are roughly the same number of bank branches today as 20 years ago. And I think you would have to compare to hundreds of years ago to find anything like a 99% decline in farm laborers.
That would be pretty narrow analytics to not need GDPR consent. No IP addresses or any way to uniquely identify a person.
Legitimate interest definition is pretty narrowly scoped. Most of the sneakiness relies on nobody checking or enforcing it.
Hard to imagine a less trustworthy source