ianand

8 ms·

31.
by ianand 3y ago
What @bunderbunder said. ^^^
32.
by ianand 3y ago
Oh man. Was not aware. That's a bummer. Have logged it as something to look into https://github.com/ianand/spreadsheets-are-all-you-need/issu... It's another reason to potentially port this thing to the
33.
by ianand 3y ago
LOL that’s totally me. Thanks for posting.
34.
by ianand 3y ago
Happy to help and thanks! Do let me know how the material could be better. Always looking to improve it. To answer your question: EECS Major in college; 20 years of engineering and product management experience. I have given a few technical
35.
by ianand 3y ago
Strictly necessary? Maybe not. I wrote that before URIAL [1][2]. I actually haven't tried URIAL in GPT2 small but I need to give it a whirl. Might be too small a model to work? Even if URIAL works with GPT2 small, the really small cont
36.
by ianand 3y ago
Oh wow!! Thanks!!
37.
by ianand 3y ago
> Brilliant idea/URL. Thanks! There's a truth to the name beyond just the play on the transformers paper. Definitely have thought about how many SaaS apps could be a spreadsheet and vice versa and often use them to create mini-
38.
by ianand 3y ago
Creator here. Thanks for posting this! Happy to answer questions or take suggestions and hope it helps folks better understand LLMs. Next video will be on embeddings and hopefully done soon-ish.
39.
by ianand 3y ago
I'm a bonehead. I did buy saayn.ai but forgot to redirect it. Fixed now. Thanks for pointing it out.
40.
by ianand 3y ago
Author here. Yes, Jeremy Howard and fast.ai was one of the inspirations for this! I'd actually be curious what he thinks of the project if he ever sees it.
41.
by ianand 3y ago
BTW this is a good article on the mechanisms you can use to counteract goodhart’s law in organizations https://commoncog.com/goodharts-law-not-useful/
42.
by ianand 3y ago
Thanks! Each one takes a surprisingly long time to make. Even figuring out how make the explanation accessible and compelling yet still accurate takes awhile and then there’s still the actual video to do.
43.
by ianand 3y ago
Thank you. Validating to hear others feel spreadsheets are a natural and accessible way to explain LLMs. Someone asks “what’s do they mean by a parameter in a model” or “what is the attention matrix” and you can just pull it up graphically
44.
by ianand 3y ago
This is great. In a similar vein, I implemented GPT entirely in spreadsheet functions with accompanying video tutorials https://spreadsheets-are-all-you-need.ai/
45.
by ianand 3y ago
That is fair. I didn't mean to imply it was earlier, just that he had mentioned it and research continued in this area. The article you linked is definitely more comprehensive.
46.
by ianand 3y ago
Karpathy himself has raised some of these issues with BPE https://twitter.com/karpathy/status/1657949234535211009 There is research about getting rid of tokens, some of which is in that thread
47.
by ianand 3y ago
Relatedly, Karpathy had an insightful post awhile back on perils of tokenization https://twitter.com/karpathy/status/1657949234535211009 (I cited it in my Excel tutorial on BPE https://www.youtube.com&
48.
by ianand 3y ago
Peter Attia did an analysis of this study and pointed out the use of casein limits the conclusions we can draw. https://peterattiamd.com/protein-anabolic-responses/ Effectively the upper limit of protein absorption was
49.
by ianand 3y ago
Relatedly, and more important for startup founders, is that more features may not actually drive increased adoption (e.g. David Bland's product death cycle https://andrewchen.com/this-is-the-product-death-cycle-why-i...
50.
by ianand 3y ago
Andrew Mayne (ex-OpenAI prompt engineer) actually uses Memento as a great metaphor for describing LLMs https://x.com/ianand/status/1723976526436417546?s=46&t=x6aIM...
51.
by ianand 3y ago
As it happens, I'm working on an approachable tutorial on LLMs using a custom GPT2 implementation in Excel. It's still a work in progress (only two lessons so far) but would love feedback: https://spreadsheets-are-all-y
52.
by ianand 3y ago
Reminds me of “FedEx is all you need” from ICML https://x.com/quantumaephraim/status/1685358421359341568?s=4... Full disclosure: I have an AI side project that also uses “is all you need” in the title.
53.
by ianand 3y ago
Creator here. Flattered someone posted this. Just did a demo of this at the local AI Tinkerers that’s a good intro to the project https://youtu.be/AIACopCeGlQ?si=HEOlIQts_oWpZ2AG
54.
by ianand 3y ago
Thanks! Had the same issue.
55.
by ianand 3y ago
Yes the key thing here seems to be the influence of type I vs II muscle fiber composition on glucose sensitivity. That more type I would be worse is what I think most folks would expect but still interesting. Notable that we tend to lose ty
56.
by ianand 3y ago
If Bruegel painted it today it would just show an iPad.
57.
by ianand 3y ago
Yes I realize that but was responding to the comment about the creami directly, not the article itself. (Regarding the article, I’m skeptical of the claim.) Officially they say to wait 24 hours to freeze but I’ve tasted it to work at around
58.
by ianand 3y ago
Highly recommend the creami. It lets you make healthier ice cream you can’t get from a typical ice cream maker. It’s basically a disruptive innovation (in the Christensen sense) for ice cream. https://allscreamforicecream.com&#x2
59.
by ianand 4y ago
I'm biased but there is a better way. Give developers a high performance method programmatically manipulating the cache key from JavaScript. That's what we created with EdgeJS: https://docs.layer0.co/guides/ca
60.
The Evolution of Jamstack: ISR/ISG, DPR, and the Jamstack Identity Crisis(cfe.dev)
1 points by ianand 5y ago|0 comments

More ›