EGreg

55 ms·

1. 841.▲
   by EGreg 1y ago

   Would you go further, and say that Unreasonable Effectiveness… is considered harmful?
2. 842.▲
   by EGreg 1y ago

   Can they? Yes. Should they? No.
3. 843.▲
   by EGreg 1y ago

   So can this be used to predict patterns for traffic, restaurant table availability, and your customers’ demand for things based on other customers?
4. 844.▲
   by EGreg 1y ago

   I remember the same was said when GOP lost horribly in 2008 and Dems rode Obama’s coattails. The GOP was supposed to never recover. Demographically they were in a significant minority. Then they hatched REDMAP…
5. 845.▲
   by EGreg 1y ago

   I've been a huge proponent of open source for a decade. But in the case of AI, I actually have opposed it for years. Exactly for this reason. Yes, AI models can run on GPUs under the control of many people. They can provision more GPUs
6. 846.▲
   by EGreg 1y ago

   No, why would they. You are supposed to maintain that cache. What I really want to know is about caching the large prefixes for prompts. Do they let you manage this somehow? What about llama and deepseek?
7. 847.▲
   by EGreg 1y ago

   This is super interesting and obviously someone would have tried diffusion for text. But I will ask the obvious question… how does it know how many words or even tokens to fill in, before it knows what the words will be? It would hamstring
8. 848.▲
   by EGreg 1y ago

   But seriously… the solution is often to cache / shard to a halfway point — the LLM model weights for instance — and then store that to give you a nice approximation of the real problem space! That’s basically what many AI algorithms do
9. 849.▲
   by EGreg 1y ago

   You’re not wrong Using an LLM and caching eg FAQs can save a lot of token credits AI is basically solving a search problem and the models are just approximations of the data - like linear regression or fourier transforms. The training is ba
10. 850.▲
    by EGreg 1y ago

    Can you get into more detail and share some links? Inquiring minds want to know
11. 851.▲
    by EGreg 1y ago

    This is also the problem I have with John Searle’s Chinese room
12. 852.▲
    by EGreg 1y ago

    Tell me you’re using Imagen 3 without telling me you’re using Imagen 4… or something
13. 853.▲
    by EGreg 1y ago

    Because they forgot to eliminate the time bias They were supposed to make recordings of the submissions, then play the recordings in random order to the judges. D’oh
14. 854.▲
    by EGreg 1y ago

    Is 23andMe no longer going to provide matrilineal mitochondrial DNA analysis and haplogroups? It is a very valuable service if you’re Jewish, that others like Heritage.com don’t. Where can you get it otherwise?
15. 855.▲
    by EGreg 1y ago

    Sure, but why would a sales person give me a free ticket to a conference? Am I going to come by their booth and buy something?
16. 856.▲
    by EGreg 1y ago

    How exactly do you contact them and ask? Can you give an example
17. 857.▲
    Tech giants honestly admit they'll unleash AGI even if it kills us all

    3 points by EGreg 1y ago|0 comments
18. 858.▲
    by EGreg 1y ago

    I thought this was mostly a solved problem. Fibers Green threads Coroutines Actors Queues (eg GCD) … Basically you need to reason about what your thing will do. Separate concerns. Each thing is a server (microservice?) with
19. 859.▲
    by EGreg 1y ago

    Been building something along these lines for a while. At Qbix, we call it our QBOX. Full stack, using Nix for the base, and Nitro attestation. No SSH. We have the exact same approach — cron running and only downloading signed scripts and b
20. 860.▲
    Coinbase cyberattack: hackers paid overseas employees, got scans of IDs, etc.(theguardian.com)

    1 points by EGreg 1y ago|1 comments
21. 861.▲
    by EGreg 1y ago

    don't you mean bernstainbear? https://www.cosmopolitan.com/lifestyle/a7664693/mandela-effe...
22. 862.▲
    by EGreg 1y ago

    I gotta be honest It didnt start out that way But 14 years later I would say this was also a labor of love and I’m not sure if it was entirely rational: https://github.com/Qbix and https://community.qbix.com
23. 863.▲
    by EGreg 1y ago

    Oh man I wish I knew I was being paid to be present and available In my last full time job I worked for a tech consulting company that rented us out to teams at financial instutions that managed insane amounts of money. This was in 2021 bef
24. 864.▲
    by EGreg 1y ago

    All you need to stop posting titles like that !
25. 865.▲
    by EGreg 1y ago

    We need more than that, and all you need to stop saying that!!
26. 866.▲
    by EGreg 1y ago

    I just told you how… it would show your special icon or phrase inside so you’d confirm it before you typed anything. The phrase would be managed through a system screen, like a login screen
27. 867.▲
    by EGreg 1y ago

    I once sent an email to Steve Jobs back in 2009 or so I told him that the MacOS permissions dialog could easily be spoofed, and that Macs should have a secret phrase or icon that you choose that they’d display inside these dialogs, and prev
28. 868.▲
    by EGreg 1y ago

    I think there are two separate things. The static websites and beautiful designs With a lot of iframes and widgets, running some kind of services like Disqus chatrooms were It’s far more secure, too.
29. 869.▲
    by EGreg 1y ago

    Both VB and Hypercard were amazing, for Windows and Mac, respectively Anyone thinking of doing it on the Web?
30. 870.▲
    by EGreg 1y ago

    “Detects” Probably not the amount the aLCHemists expected centuries ago… but hey. It’s something!