Is it? I've never seen a single OCR that would replace a human just typing it by hand.
What if the goal is something actually useful, such as converting scientific paper PDF back to LaTeX that renders into a pixel-perfect copy? What about converting tables from electronics datasheets into computer-readable form? I wouldn't even expect it in the next decade.
How does it compare against Finereader? Comparisons against transformer-based OCRs don't really tell anything. The last time I checked, neither of them were of "OCR this legal document" quality.
Yeah, I've first seen it over 15 years ago. Usually you use operator of the same priority as you'd like, and also #define xor &xor_i& to get all that detail out of sight.
The last time I remember, the green company did the same HTTP thing literally with their driver downloads from the website, and refused to fix it.
Makes me wonder, how much of that 4 month delay was spent deliberating with the state actor. As if there was Prism, and both companies were legally bound to allow MitM to happen, and thus don't have a bug bounty for it.
Like detecting constriction or loss of integrity of blood vessels, and doing the corresponding intervention.
The saddest thing here is not that it requires some future nanotechnology, but is achievable at the present scientific level, yet too expensive to develop, and wouldn't see FDA permission in a decade or two anyway.
It's quite easy to check responses to other customers in other threads there, and somehow I see quite a lot of "oh, go to that other support" and ghosting.
If you create support ticket on hacker news, then yes, you will probably get it waved. It's somewhat sad that HN is their support forum now.
So basically they got fined a cost of single tractor repair, and it didn't even create a legal precedent due to settlement? Someone believed that "will make available" has any consequences, given decades-long tradition to just ignore such agreements? Well, great.
The statement is a bit too strong. It's not malice, it's just plain old stupidity. In the same way as soviet nuclear reactors don't explode, nuclear submarines don't either. Nobody have thought it might happen, nobody was aware, and no special "on call" service for recovering people from sunk submarine ever existed. _Of course_, it would never happen in just 6 hours.
In the same way, if you don't have anyone on-call to recover the backend along with backups and recovery plans, the chance to have production up and running in 6 hours will be zero.
In case of a way more physical thing of "submarine sunk in polar cold waters", it'd take a good 4 hours just to get _something_ there. Not to mention extracting a person from 300ft involves a good plan on avoiding decompression sickness, and you can't really bring a 50-ton decompression bell on a helicopter and hover for half a day. I can hardly imagine what would such a disaster recovery plan even be.
Those profiling tools don't really care which features are going to be used for predictions. It's just machine learning, and it's indiscriminate. So if you have an extension that correlates with you being Muslim, it will be used for whatever ML predictions they give to other companies, and the worst case will be another "oh we didn't do this intentionally".
Of course, that's not the first time this ever happened in human history, so even if it's not "something inherently sinister", it's just "criminal negligence".
It's actually a strange situation. Tao is the only high-profile mathematician who tries to resolve the reproducibility crisis in math by popularizing proof checkers, and who used LLMs in probably the most legitimate way: to produce formal undeniable proofs with them, that cannot really suffer from LLMs being wrong all the time.
On the other hand, I still find content and arguments he produces to be quite weak, and honestly it's getting annoying to hear them that often. It's the case when he could really get some help of a ghost writer who is more experienced in popularization, otherwise this repetition might cause some serious harm instead.
In fact, the paper has an error in the argument that AI might find Fermat's theorem to be incorrect due to definition of natural numbers including a zero, because paper's version of a theorem explicitly says that the number should be greater than two, and zero cannot be greater than two.
Surprisingly, this mistake proves the author's point that human can implicitly understand what was said, and that it still has value to it, even if it's incorrect.
The word "Polymarket" can be replaced by "market" there.
Global market is just the largest gambling venue. Always has been. No amount of "monitoring" and "leagues" will keep people from going Boeing whistleblower way.
What if the goal is something actually useful, such as converting scientific paper PDF back to LaTeX that renders into a pixel-perfect copy? What about converting tables from electronics datasheets into computer-readable form? I wouldn't even expect it in the next decade.