p1esk

26 ms·

541.
by p1esk 2y ago
I get maybe one hallucination per twenty chats with gpt4.
542.
by p1esk 2y ago
“ What were the UK's top exports in 2023?" "List all YC founders that worked at Google and now have an AI startup." How to check the accuracy of the answers? Is there some kind of a detailed trace of how the answer was g
543.
by p1esk 2y ago
“ The engine uses thin cooling channels that swirl around the chamber jacket, with a variable cross sections as thin as 0.8mm. The Kerosene is pressed through the channels to cool the engine and prevent it from melting.”
544.
by p1esk 2y ago
How does llama’s answers compare to answers from gpt4?
545.
by p1esk 2y ago
I really hope I’m wrong. Can you imagine the massive unemployment and social chaos if I’m not? I’m rarely wrong about AI though. For example, five years ago I predicted (in a comment here on HN) Turing test will be passed within two years.
546.
by p1esk 2y ago
Yes. It will go like that: “should we hire a senior dev and a junior dev? Nah, a senior dev plus GPT5 subscription should be enough to do the workload”. And it will be enough.
547.
by p1esk 2y ago
No, GPT5 is not out yet. As I said, it should be out within 6 months. Please be patient.
548.
by p1esk 2y ago
It will, as soon as it becomes “powered by GPT6”.
549.
by p1esk 2y ago
“Better at coding” as defined by seniority levels: GPT4 - an intern; GPT5 - a junior developer; GPT6 - a senior developer. Yes, GPT5 will do a lot of damage, as it will be the first model replacing actual human coders. But GPT6 should be ab
550.
by p1esk 2y ago
I don’t know anyone who’s looking for junior devs today. That was not the case 3 years ago. Re LLMs: most people responding here seem to assume there will be zero progress in ML after today. I don’t understand why. It’s likely we will get g
551.
by p1esk 2y ago
President Biden created a new program What new program? EDIT: found it: https://www.cbsnews.com/amp/news/biden-immigration-program-u... Key information: “An immigrant who marries a U.S. citizen is generally elig
552.
by p1esk 2y ago
I publish papers.
553.
by p1esk 2y ago
Do opinions differ whether that particular beer crate is real? Of whether the name is appropriate?
554.
by p1esk 2y ago
There's no such thing as tables or chairs What do you mean?
555.
by p1esk 2y ago
This was the book that explained everything I needed to know about Linux with very little noise: Beginning the Linux Command Line, by Sander van Vugt. Today I would heavily supplement it with answers from GTP4o (or perhaps Sonnet 3.5).
556.
by p1esk 2y ago
Nvlink 4.0 used to connect H100 GPUs today is almost as fast as PCIe-7.0 (12.5GBs vs 16GBs). By the time PCIe-7.0 is available I’m sure NVlink will be much faster. So, yeah, GPUs are currently the most bandwidth hungry devices on the market
557.
by p1esk 2y ago
It's only been a week since the million dollar prize was offered. Let's see what the SOTA is in a month.
558.
by p1esk 2y ago
GPT-4o might not have been trained on a sufficiently large amount of visual data to develop advanced spatial intelligence. Perhaps it needs to see a lot more images, or perhaps it needs to be trained differently (e.g. predict the next frame
559.
by p1esk 2y ago
Well, that didn't take long, did it? 50% on ARC public test set [1] less than a week after the announcement of the prize. Though I have to say, the solution, at least superficially, does look like what Chollet alluded to: hybrid of LLM
560.
by p1esk 2y ago
we achieved a margin of 7.6% Are you one of the original contributors of the work done at together.ai [1], or are you trying to take credit for it? [1] https://arxiv.org/abs/2406.04692
561.
by p1esk 2y ago
Has gpt5 been released? No. So how do you know if it’s no better than gpt4?
562.
by p1esk 2y ago
So don’t vote for young candidates then? Why do you need these limits?
563.
by p1esk 2y ago
I just watched the whole Dwarkesh/Chollet interview, and just like Dwarkesh was clearly not convinced by the Chollet's arguments, neither am I. I still expect decent results (>50%) on ARC benchmark soon (this year) now that the
564.
by p1esk 2y ago
we didn't have something to debug the network traffic What do you mean? You couldn't mirror traffic to a linux box and use tcpdump there?
565.
by p1esk 2y ago
We have pretty much infinite training data available on YouTube. We can scale by many orders of magnitude before we run out of data. Why do you think we hit a ceiling?
566.
by p1esk 2y ago
Not sure how valuable that data would be to train models. I’m guessing it’s millions of routine phone conversations and text messages. Maybe a ton of satellite imagery. I wouldn’t consider that to be particularly useful to go from gpt5 to
567.
by p1esk 2y ago
I’m also 30 years into my career. I love collaborating. Yet at progress update meetings my boss wants to know not how much I “collaborated”, not how busy I was, not what problems I solved. He wants to see results, especially results that ma
568.
by p1esk 2y ago
Got any evidence to support your statement?
569.
by p1esk 2y ago
Thank you for the thoughtful answer. I also wish more people were asking such questions. Let's look at some of your points. the difference between AGI and ML I've seen the recent discussions on ARC benchmark. It's not clear
570.
by p1esk 2y ago
Off topic, but I would like to hear your opinion on the impact of the AI progress on jobs, globally. Let’s assume, for simplicity, that GPT5 will actually be significantly better (eg similar improvement as what we had with 3.5 —> 4). An

More ›