ttul

60 ms·

931.
by ttul 4y ago
The thing is, machine learning sorta requires a few math prerequisites: linear algebra, differential equations, and to some degree vector calculus. Most web developers don’t have this background.
932.
by ttul 4y ago
Denmark came to mind as I was writing my comment, but I am not well enough informed. Thanks.
933.
by ttul 4y ago
America operates on the principal that it is better to fail quickly and rebuild than it is to defend what’s not working to preserve the status quo for as long as possible. In a century marked by rapid technological change, it is probably be
934.
by ttul 4y ago
TL;DR America owes its success to a relatively young population with high immigration, flexible labour laws, deep capital markets, and a swift and efficient bankruptcy process. But sclerotic politics threatens to derail this progress, parti
935.
by ttul 4y ago
GPT-4 doesn’t get the humor: “The paper aims to apply the Fairness, Accountability, and Transparency (FAT) framework to this algorithmic system to ensure better ethical outcomes. However, given the nature of the proposed solution, researche
936.
by ttul 4y ago
Thank you so much for setting me straight. CodeEmporium’s 10-part explainer is so great.
937.
by ttul 4y ago
As with most tutorials on Transformers, this one leaves out some essential details: - how are the input encodings generated? - what is in those position vectors? - how are the attention vectors learned? The answer is that these things are a
938.
by ttul 4y ago
Although it is sad that Germany has chosen the path of fear, consider that its neighbors are free to step in with additional nuclear power capacity and seem to be doing so. Since the European grid is connected across borders, Germans will c
939.
by ttul 4y ago
Yes. I recall that someone in the FastAI course mentioned an amazing speed up but then I heard nothing further. I suppose maybe it wasn’t quite ready for prime time yet.
940.
by ttul 4y ago
You’re not supposed to do this, but GPT-4 can generate RLHF data.
941.
by ttul 4y ago
Honestly if you can train a decent model for $1,000, a 15 person company could afford to train it up monthly.
942.
by ttul 4y ago
Also, they will definitely be making some upside on the OpenAI stock…
943.
by ttul 4y ago
It's one thing to come up with the technique; quite another to source the dataset of images on which the model will be trained, train it up, and then run it as a service for millions of users.
944.
by ttul 4y ago
This is a very significant paper. The first diffusion paper required 1,000 steps to generate an image. By last November, they were down to 50 steps. This paper takes us to one or two steps. What we're talking about here is nothing shor
945.
by ttul 4y ago
Anything that drives the dopamine cycle is of interest to humans. Sex in all its forms is pretty motivating.
946.
by ttul 4y ago
Nvidia: just as the sun starts setting on crypto mining, the foundation model boom begins. And in the background of it all, gaming grows without end.
947.
by ttul 4y ago
Sort of, but only if you consider that the states are massive. It’s not just “what’s the next word after this one.” It’s, “what’s the next word after this one, considering the other several hundred other words that came first.” So, the inpu
948.
by ttul 4y ago
Idea: Invite the police in from day one.
949.
by ttul 4y ago
Cloudflare’s S2 browser isolation tech should get a solid nod here. S2 was a Seattle startup that figured out how to stream the DOM from a remote headless browser to the client, delivering a very smooth experience that is far superior to RD
950.
by ttul 4y ago
If your goal is to know the risk-adjusted value of an equity grant, you need more than just the cap table. You need the bylaws and articles of the corporation, all the shareholder agreements, and all the loan agreements the company might be
951.
by ttul 4y ago
I think that tool would have broad appeal, but I wonder if the most likely case is that it would be buried inside other, higher-level systems, such as customer support automation SaaS.
952.
by ttul 4y ago
You could of course use your own question and answer data to refine the model using the same process. I wonder if anyone has tried that yet to, for instance, fine tune LlaMa to answer support queries for their company?
953.
by ttul 4y ago
There is plenty of low hanging fruit to harvest with these language models in the next year, let alone ten years. And the models are improving rapidly. GPT is a sequence to sequence model. Many jobs are effectively also sequence to sequence
954.
Baize, an open-source chat model trained with LoRA(github.com)
6 points by ttul 4y ago|1 comments
955.
by ttul 4y ago
“ Baize is an open-source chat model trained with LoRA. It uses 100k dialogs generated by letting ChatGPT chat with itself. We also use Alpaca's data to improve its performance. We have released 7B, 13B and 30B models. Please refer to
956.
by ttul 4y ago
> Develop a platform that offers a subscription-based service for farmers to rent autonomous tractors equipped with vision AI technology and automated spraying capabilities. Partner with leading tractor manufacturers and agricultural tec
957.
by ttul 4y ago
> Develop a wearable device that monitors an individual's body temperature and alerts them when it reaches a dangerously high level, potentially preventing instances of spontaneous human combustion. Incorporate a feature that tracks
958.
by ttul 4y ago
> Create a subscription-based tampon delivery service that offers a personalized selection of tampons, pads, and other menstrual products based on each customer's individual needs and preferences. The service can also include additi
959.
by ttul 4y ago
I didn't take part 1, so I can't comment intelligently on this.
960.
by ttul 4y ago
I audited this course - I say “audited” because I didn’t have time to do the homework; I just watched the lectures and spent what time I could playing around in Colab building some toy models. This is very much a full time university level

More ›