falcor84

39 ms·

781.
by falcor84 1y ago
I wrote that based on my experience comparing my prose writing and code to what I can get from ChatGPT or Claude Code, which I feel are on average significantly higher quality than what I can do on a single pass. The quality still improves
782.
by falcor84 1y ago
I know you were joking, but responding in seriousness - while in general it's worthwhile asking "Quis custodiet ipsos custodes?", in this particular case, I don't see any issue with Down Detector detecting the Down Detec
783.
by falcor84 1y ago
Maybe I misunderstand, but it seems that there's nothing in your comment that contradicts any aspect of mine. Regarding image classification. As I see it, a company like Microsoft surveying researchers about the state of the art and th
784.
by falcor84 1y ago
> human oriented aspects The irony is of course that humans in general and software professionals in particular (myself definitely included) notoriously struggle with communication, whereas RLHF is literally optimizing LLMs for clear com
785.
by falcor84 1y ago
20 years ago NNs were considered toys and it was "extremely obvious" to CS professors that AI can't be made to reliably distinguish between arbitrary photos of cats and dogs. But then in 2007 Microsoft released Asirra as a ca
786.
by falcor84 1y ago
Agreed, and I'd also love to see a baseline of human performance here, both of experienced quant traders and of fresh grads who know the theory but never did this sort of trading and aren't familiar with the crypto futures market.
787.
by falcor84 1y ago
(Unless you're a marketer) It makes a lot more sense to build a benchmark before the capabilities are there.
788.
by falcor84 1y ago
I love it that there's a "Read AI-generated summary" button on their post about their new AI. I can only expect that the next step is something like "Have your AI read our AI's auto-generated summary", and so f
789.
by falcor84 1y ago
100% this. While in my professional capacity I'm all in for reliability and redundancy, as an individual, I quite like these situations when it's obvious that I won't be getting any work done and it's out of my control,
790.
by falcor84 1y ago
Your mileage may vary, but for me, working today with the latest version of Claude Code on a non-trivial python web dev project, I do absolutely feel that I can hand over to the AI coding tasks that are 10 times more complex or time consumi
791.
by falcor84 1y ago
It's not just one of many fine tunes; it's the default model used by OpenAI's official tools.
792.
by falcor84 1y ago
Any particular changes of interest?
793.
by falcor84 1y ago
"You know, you are also right"
794.
by falcor84 1y ago
> I guess improvements will be incremental from here on out. What do you mean? These coding leaderboards were at single digits about a year ago and are now in the seventies. These frontier models are arguably already better at the benchm
795.
by falcor84 1y ago
That looks impressive, but some of the are a bit out of date. On Terminal-Bench 2 for example, the leader is currently "Codex CLI (GPT-5.1-Codex)" at 57.8%, beating this new release.
796.
by falcor84 1y ago
See also discussion on https://news.ycombinator.com/item?id=44962529
797.
by falcor84 1y ago
Another good reason to support independent internet archives
798.
by falcor84 1y ago
I would just suggest that if you want your comment to be more helpful than the article that you're critiquing, you might want to actually quote the part which you believe is "The actual point". Otherwise you are likely to hav
799.
by falcor84 1y ago
Reminds me of the show Alias, where the premise is that there's a whole intelligence organization where almost everyone thinks they're working for the CIA, but they're not ...
800.
by falcor84 1y ago
I personally am extremely impressed about it reaching 15% on unseen environments. Note that just this year, we were surprised that LLMs became capable of making any progress whatsoever in GBA Pokemon games (that have significantly simpler w
801.
by falcor84 1y ago
From what I see, SIMA only focuses on games where you control a single avatar from a 1st/3rd person perspective, and would assume that switching to a non-embodied game where you need to control the whole army at once would require sign
802.
by falcor84 1y ago
I'm probably neurodiverse, so ymmv, but I really couldn't care much less about how people are doing; it's a very small part of my idea of a good conversation. What I want is to bounce ideas off of each other, and so the answe
803.
by falcor84 1y ago
As a counterpoint, I've been using my own PC since I was 6 and know reasonably well about the innards of LLMs and agentic AI, and absolutely love this ability to hold a conversation with an AI. Earlier today, procrastinating from work,
804.
by falcor84 1y ago
I'm generally ok with it wanting a conversation, but yes, I absolutely hate it that is seems to always finish with a question even when it makes zero sense.
805.
by falcor84 1y ago
Slightly off-topic, but seeing that this one is behind a paywall, I looked at Deepmind's site and was surprised that there aren't any new entries since September [0]. I wonder if there was some intentional change in their publicat
806.
by falcor84 1y ago
Why? I don't recall hearing of any airline pilot who felt insulted by an Autopilot, or a cleaner insulted by a Roomba. People who see parts of their jobs replaced often have a range of feelings, but I don't see why insult would be
807.
by falcor84 1y ago
To me it seems like an exercise in semantics. As per Carl Sagan, you could argue that fully Autonomous Discovery requires that it first "create the universe". I read "Autonomous" there as "working independently with
808.
by falcor84 1y ago
I would actually put a lot of the blame for today's VR winter (or at least expectations cooldown) on Ready Player One, and particularly its movie adaptation. Not that it was bad per-se, but for me and others I spoke too it was so outla
809.
by falcor84 1y ago
This is the original report, but not much detail here either - https://www.ft.com/content/c586eb77-a16e-4363-ab0b-e877898b7...
810.
by falcor84 1y ago
I just saw that Terminal Bench introduced a new evaluation approach based on the new Harbor tool [0] and was surprised to see that it completely reshuffled the leaderboard, with the top 4 places now held by variants of gpt-5, whereas in ter

More ›