swalsh

15 ms·

211.
by swalsh 3y ago
As a user, I basically just care about a minimum baseline of competence... which most models do well enough on. But then I want the model to "just give me the code". I switched to Claude, and canceled my chatgpt subscription bec
212.
by swalsh 3y ago
Is this Mistral large?
213.
by swalsh 3y ago
Why are you so negative towards this? It's just an Open source project... who cares how good it is. It's a great way to learn, and play, and experiment.
214.
by swalsh 3y ago
Embeddings are indeed sticky, I was referring to the LLM model itself.
215.
by swalsh 3y ago
> The model requires ~264GB of RAM This feels as crazy as Grok. Was there a generation of models recently where we decided to just crank on the parameter count?
216.
by swalsh 3y ago
The models are commodities, and the API's are even similar enough that there is zero stickiness. I can swap one model for another, and usually not have to change anything about my prompts or rag pipelines. For startups, the lesson her
217.
by swalsh 3y ago
Welp, guess pack it in boys. No use in trying to compete.
218.
by swalsh 3y ago
I've been trying to figure out how LLMs learn logic/reasoning. It's just not intuitive to me how that works.
219.
by swalsh 3y ago
He's selling water in a desert, kind of doesn't matter how polished his presentation is.
220.
by swalsh 3y ago
72 P/E ratio while they have a mere monopoly on one the most valuable resource in the world. Competition WILL come. Maybe it's Groq, maybe AMD, maybe Cerebras. Maybe there's a stealth startup out there. Point is, they&#x27
221.
by swalsh 3y ago
At 2 trillion, it's all baked in already
222.
by swalsh 3y ago
We should just call it open weight models at this point.
223.
by swalsh 3y ago
I'd bet a synthetic data set could do the job effectively.
224.
by swalsh 3y ago
Considering how poor it is compared to other models, it really emphasises how important fine tuning is. Models with MUCH smaller parameter counts are outperforming it in many metrics.
225.
by swalsh 3y ago
I don't know if Claude is "smarter" in any significant way. But its harder working. I can ask it for some code, and I never get a placeholder. It dutifully gives me the code I need.
226.
by swalsh 3y ago
I use grok all the time to find tweets or ask about trends on Twitter. For that it's better than what used to exist. But its not a great model outside that narrow use case.
227.
by swalsh 3y ago
That's just a matter of fine tuning
228.
by swalsh 3y ago
In terms of my "off ramp"? I have a multi-part plan. Immediately, i'm working to get closer to the business. To be closer to the position of defining requirements, not implementing requirements. Secondary, i'm experimen
229.
by swalsh 3y ago
Software is a commodity, find something more valuable to differentiate your startup.
230.
by swalsh 3y ago
You're looking at the worst version that will ever be released. If these guys don't get better over time, someone else will.
231.
by swalsh 3y ago
Ah yes, you've entered the first stage of grief. Denial. Next you'll start bargaining, you'll get angry, and you'll become depressed, eventually you'll just accept that AI is taking over software. In my mind, I&#x
232.
by swalsh 3y ago
Seeing AI do something you do for a living results in all the stages of grief. Denial (Oh it made a mistake on this, or it's not as good as a person) Anger (People are going to lose their jobs, this destroys the meaning of what we do,
233.
by swalsh 3y ago
Why make such a low effort pessimistic comment. What happened to HN?
234.
by swalsh 3y ago
Sorry if i'm completely missing it, I noticed in the code, there is something around chat: https://github.com/tembo-io/pg_vectorize/blob/main/src/chat.... This would lead me to believe there is
235.
by swalsh 3y ago
I've also seen the opposite, where tiny little 7B models get real close to GPT4 quality results on really specifically use cases. If you're trying to scale just that use case it's significantly cheaper, and also faster to ju
236.
by swalsh 3y ago
There was a period of time where data was easily accessible, and Open AI suctioned up as much of it as possible. Places have locked the doors since then realizing someone was raiding their pantry. To get that dataset now would take signifi
237.
by swalsh 3y ago
It's a shame they couldn't find another business model. I've found a lot of value in Mistral, but none of the raw models were what worked for me. The models I've been using were derived from mistral... but it was the o
238.
by swalsh 3y ago
Yeah, I'm in a similar thought process. I've been burned multiple times by Apache, will not touch ever again.
239.
by swalsh 3y ago
"A little" That's the key
240.
by swalsh 3y ago
Technology has been quite detrimental to humanity if those of us in tech want to be honest with ourselves. It could have been amazing. We told ourselves it was going to connect us, and we were all going to be smarter.. but instead we abus

More ›