mycall

49 ms·

1. 781.▲
   by mycall 4y ago

   There are lots of people with misconceptions of LLMs. It will take time to adjust. I reached the same conclusion as yourself, but do see a totally different path to take regarding information propagation (how GPT works). For example, cells
2. 782.▲
   by mycall 4y ago

   > it is completely impossible to gain knowledge you do not already possess from a language model Not true. Emergent abilities is an active research area in LLMs [0]. They even have pretty graphs on the topic. [0] https://ai.g
3. 783.▲
   by mycall 4y ago

   Prompt engineering is another way to say functional specification imho.
4. 784.▲
   by mycall 4y ago

   Do you see any use cases to include WASI into your godot gaming support? Maybe midi music too.
5. 785.▲
   by mycall 4y ago

   I'm impressed on the speed of the new i3-N300/N305 at 7w or 9w/15w https://ark.intel.com/content/www/us/en/ark/products/231806/...
6. 786.▲
   by mycall 4y ago

   This sounds like the job for a linter or intellisense editor.
7. 787.▲
   by mycall 4y ago

   I would love to see someone make an image to unicode "curve fitting" algorithm or converter, similar to ANSIDRAW.
8. 788.▲
   by mycall 4y ago

   I have been impressed what my GPDWIN2 m3-8100Y processor can do at 7 watts.
9. 789.▲
   by mycall 4y ago

   A system can be more than one computer, i.e. mainframe. Airgapped systems can include multiple computers that are disconnected from external networks. They can be very useful for specialized applications.
10. 790.▲
    by mycall 4y ago

    ClippyGPT
11. 791.▲
    by mycall 4y ago

    Is it unethical to buy a "cheap" replacement bicycle if your previous one was stolen?
12. 792.▲
    by mycall 4y ago

    Sometimes it is better to do tech at a non-tech company. For example, transit is in need of highly skilled tech workers, although government itself slows down progress. There are many places to go outside of tech companies.
13. 793.▲
    by mycall 4y ago

    Oil chemistry can determine country of origin and sanction enforcement is possible at port.
14. 794.▲
    by mycall 4y ago

    GPT and transformers are not the end game for logical analysis. Pretrained LLMs cannot capture factual knowledge from text, that is true. MIT is proposing a unified model for Knowledge Embedding and Pretrained LanguagERepresentation (KEPL
15. 795.▲
    by mycall 4y ago

    My favorites are the waterproof versions.
16. 796.▲
    by mycall 4y ago

    GPUs are much less efficient than cores of the type Cerebras has made.
17. 797.▲
    by mycall 4y ago

    > everything used usb-c so iPhones must be usb-c now toooooo EU agrees with this.
18. 798.▲
    by mycall 4y ago

    Put the flags into a class and pass the class around.
19. 799.▲
    by mycall 4y ago

    Customer Service is a form of continued sales. When it comes to money, people are more serious unless they are spending rounding error money. Being real is a better attitude than cheerful these days.
20. 800.▲
    by mycall 4y ago

    This one got to me, great render: https://dittytoy.net/ditty/6f30b0885d Original version: https://www.youtube.com/watch?v=BSPuY-YLe-w
21. 801.▲
    by mycall 4y ago

    Do you think GPT4 or GPT5, being multimodal, will do something similar?
22. 802.▲
    by mycall 4y ago

    Have you considered trading books with other avid readers? That freezes the need to buy new books and opens up a whole new catalog for yourself.
23. 803.▲
    by mycall 4y ago

    Were you on the phone with IBM troubleshooting their OS/2 port? I would think it was available.
24. 804.▲
    by mycall 4y ago

    "now chose different words which would make up a valid sentence." const words = [ "All", "birds", "chirp", "during", "early", "fall", "gathered",
25. 805.▲
    by mycall 4y ago

    Don't they use word2vec?
26. 806.▲
    by mycall 4y ago

    I was able to do it. Try this: provide a javascript array of size 26 which includes 26 words with each word incrementing the first letter of the word in ascii, starting with the letter A.
27. 807.▲
    by mycall 4y ago

    Did you save and submit your corrections? That is one way to help OpenAI train it.
28. 808.▲
    by mycall 4y ago

    When comparing how the human brain works, the transformer model is not the same thing. Until convergence of the mechanics occur, there will be limitations in the efficiency of AGI. Stil, I am eager to see what a 100 trillion or 1 quadtrill
29. 809.▲
    by mycall 4y ago

    > there may be a way forward in reifying those cases where ChatGPT does get it right. There is, that is the Chat in ChatGPT -- supervised learning. Teach it the right answer, then use "Save and Submit" at the top of the chat.
30. 810.▲
    by mycall 4y ago

    I remember using the Topaz MSDOS, it was ahead of its time with Targa graphics. I'll have to take a closer look at Gigapixel.