layoric

17 ms·

211.
by layoric 3y ago
I’ve used guidance, why is it defunct? I found it was powerful at templating, really decent for generating synthetic datasets.
212.
by layoric 3y ago
In my experience, the key is that it can also provide good explanations for the basics as well. You can fairly easily overwhelm it, but with some practise of prompting, learning new language/domain basics it is pretty good, and much be
213.
by layoric 3y ago
Their approach is different, they build ML tech that runs on-device, so whatever they developed has to be able to run efficiently on iPhone/iPad etc. I don’t think we will “hear” about Apple using LLMs either way because they will no d
214.
by layoric 3y ago
It doesn't seem to be a major barrier for most (it absolutely is for me). Is there enough of a want/market for tools that integrate with common OSS LLM APIs like oobabooga/Kobold/Novel/vLLM etc? One idea I've b
215.
by layoric 3y ago
I think "narrative storytelling" I think is too broad, that can absolutely work in VR, but it means creating a different experience from the ground up. A "TV Show or movie" I agree is going to have a hard time of that, b
216.
by layoric 3y ago
GPT3 benchmark, as others have said this means it would be ~46 hours to train GPT, still impressive! Also, going by other comments here, ~3500 of these or 13-14 DGX GH200s at ~$10m each means we are talking about ~$135m worth of compute her
217.
by layoric 3y ago
I built a proof of concept recently that tries to show a generic hybrid of command and intent[0]. The UI generates form representations of API calls the AI agent has decided on making to complete the task (in this case booking a meeting). S
218.
by layoric 3y ago
Out of interest, did you find this was a problem that using optimised numpy still didn’t perform good enough? (depending on what you call pure python, yes I agree python loops etc are extremely slow)
219.
by layoric 3y ago
I agree that codegen + docs is actually where most of the value is. The problem I think is in the design. Having an intermediate spec makes all the downstream tooling for codegen + docs need to handle all the complexity. Any information los
220.
Creating ChatGPT Agents in .NET to Call Internal APIs(servicestack.net)
4 points by layoric 3y ago|0 comments
221.
by layoric 3y ago
My brother got me one of these as a present when I was a teenager. The only game I remember that had full support was the god game Black & White. You controlled everything with a hand including interacting with the creature and moving&#
222.
Creating ChatGPT Agents in .NET to Call Internal APIs(servicestack.net)
3 points by layoric 3y ago|0 comments
223.
by layoric 3y ago
I might be reading the code wrong but it looks like it crawls the schema making a generation per primitive type. While that’s a clever way to ensure valid JSON, I don’t know if I’d go as far as to describe it as efficient. Saying that if th
224.
by layoric 3y ago
I am likely wrong here, but I thought the term “prompt engineering” was also related to the act of building a system around generating a prompt dynamically based on limited input? Eg not just trial and error over single prompts but using br
225.
by layoric 3y ago
I assume they mean API access to GPT4, not the ChatGPT interface.
226.
by layoric 4y ago
I wouldn't quite go that far, but the longer I am building these models into systems, I think the good use case for LLMs/Generative AI can be categorized into a few types of systems. 1. Data collection - Obvious, creating a voic
227.
by layoric 4y ago
Yes, but it is a complete gaming system. My point is $500 is still a lot of money for a used GPU that is nearly 3 years old. Jut retailed brand new for $699. So yes, prices were crazy over that period for a variety of reasons, but that sh
228.
by layoric 4y ago
Where are these cheap as dirt 30 series? A 10gb 3080 is still over $500 usd used ($750 aud) when I’ve looked. When did secondhand GPUs that still cost the same as a brand new PS5 start to be considered cheap?
229.
by layoric 4y ago
It is sad to see even the more 'conservative' (design wise) car companies following this design. Lexus gets a lot of flak for boring/old tech interiors but give me their 2010s design over 'modern' touch screen/
230.
by layoric 4y ago
Nice to have tools like this to wrap up features, definitely makes these types of solutions more accessible, thanks! It would be nice to know from your experience if there is a kind of rule of thumb for calculating cost of fine tuning and r
231.
by layoric 4y ago
I am the same, the advice of "learn SQL" is still very important. Most ORMs have the escape hatch of just mapping results of a straight SQL query to an expected structure (thinking of typed languages like C#/Java here). 95% o
232.
by layoric 4y ago
They had a chance to meet a much better price to performance but IMO nvidia showed that the market is willing to pay a premium for GPUs and both major players are exploiting that. Nvidia relaunched the 4080 12gb as the 4070ti reducing the p
233.
by layoric 4y ago
What kind of setup do you have to avoid users noticing the drop of their websocket connection due to a deployment? Blazor Server development experience is very good, but even for an intranet application, it can have a poor user experience b
234.
by layoric 4y ago
Well spotted, the shim works now, so something that could be removed in the near future.
235.
by layoric 4y ago
This is what turned me off as well. I ended up getting older Dell 2011-3 dual socket workstation for cheap. Still heaps of power, room for upgrades but whisper quiet when idle and under load.
236.
by layoric 4y ago
I think is largely correct, I’ve been working remotely for the better part of 10 years. I used laptops for around 6. The portability is great when I was split between multiple jobs. Changed to a desktop about 3 years ago and wouldn’t go bac
237.
by layoric 4y ago
Starting to make sense how Ads will be injected into using ChatGPT or something similar when used as a search engine. Eg, Question: "You are ChatGPT, a large language model trained by OpenAI. Knowledge cutoff: 2021-09 Current date: 202
238.
by layoric 4y ago
Another one to consider is the A4000 16gb. I recently bought an ex-miner card for ~$500 usd . They are around a 3070 with a decent amount of memory for training scenarios, and are single slot cards. I believe there are a lot of these workst
239.
by layoric 4y ago
I do this, but just with straight docker compose. I’ve been using AWS for personal and work for about 8 years and realised over time it is a poor investment if you are always hosting your own stuff and experimenting with tech. So this year
240.
by layoric 4y ago
A good compromise I’ve found is using something like “FakeRelay”[0] with “GetMoarFediverse”[1] to pull in specific tags from specific larger servers. This acts as a way to get more interesting content in your search and other mastodon tabs

More ›