Tostino

39 ms·

961.
by Tostino 4y ago
From what i've gathered, fine tuning should be used to train the model on a task, such as: "the user asks a question, please provide an answer or follow up with more questions for the user if there are unfamiliar concepts." F
962.
by Tostino 4y ago
I don't see these as comparable at all.
963.
by Tostino 4y ago
They don't help with the memory requirements of these LLMs though.
964.
by Tostino 4y ago
The noise on the search index seems like it introduces quite a few issues with accuracy. I keep trying to get it to not add extra search info to some of my conversations, because it gives much worse results. Blog-spam really sucks.
965.
by Tostino 4y ago
That's essentially the same thing that GP said. I know I'd rather safety nets be improved for this future rather than outlawing, but don't see it happening until some major suffering and upheaval happen. Likely, they have a s
966.
by Tostino 4y ago
I've been using it daily since Docker Desktop license change and won't say I haven't run into issues from time to time (waking from sleep on Windows for example). But overall, it's been solid and worked well for me. I us
967.
by Tostino 4y ago
Good ole' Florida. It's not the first and won't be the last miscarriage of justice this state carries out.
968.
by Tostino 4y ago
Rancher Desktop has been just fine for me.
969.
by Tostino 4y ago
What is the actual server-side limit then?
970.
by Tostino 4y ago
Yeah...this is where the talk of "guardrails" sometimes gets, forgive the pun, derailed. There are good reasons to be able to put some guardrails in place on your AI model other than pure censorship. I'd really like the page
971.
by Tostino 4y ago
I honestly don't think it will be quite like that, at least not terribly soon. There is so much work being done to hook up LLMs to external sources of data, allow them to build longer term memories of interactions, etc. Each of these a
972.
by Tostino 4y ago
Though when using the browser based Bing, it's able to look at the "page" for context. I've opened text files / code in my browser and used the sidebar Bing to ask about the file without having to copy/paste in
973.
by Tostino 4y ago
You really want it integrated with an OpenAI API clone rather than directly integrated. Otherwise, interoperability will suffer greatly as new and improved models are released.
974.
by Tostino 4y ago
I've seen people saying that limiting it to 4 cores out of the 8 total can actually lead to improved performance. Have you seen that?
975.
by Tostino 4y ago
No, you misunderstand what it means in that case. Watt-hours are comparable to joules. 1 watt-hour = 3600 joules.
976.
by Tostino 4y ago
Not to mention, there are already OpenAI API clones which handle running against any of the open-source models. When they get more powerful and act as an actual competitor, it'll be very nice to have open-source components throughout t
977.
by Tostino 4y ago
I really wish this left join into union all optimization was applied automatically. There have been multiple discussions on the mailing list over the years, and no movement last I saw. It is such a performance cliff for the type of queries
978.
by Tostino 4y ago
> The last time we found something that was predicted by the Standard Model was 1983, wasn't it? The Higgs Boson was found in 2012 by the LHC and was predicted by the standard model.
979.
by Tostino 4y ago
It matters when the limited resources are spent chasing unicorns that are vanishing unlikely to actually be there. That is part of the argument. Those dollars could go other places than these giant particle accelerators that haven't sh
980.
by Tostino 4y ago
For someone who has tried to avoid hearing as much as possible about this drama (and didn't notice this was about Twitter until looking at the comments), what are the condensed "lesson learned" takeaways if you don't min
981.
by Tostino 4y ago
That is my exact recollection from the time. I know I stayed away due to that, even prior to the nightmare launch.
982.
by Tostino 4y ago
And Japan did similar levels of atrocities to Korea, China, and others in the region. We can acknowledge that things were historically pretty horrible and strive to be better in the future.
983.
by Tostino 4y ago
Here is a Last Week Tonight piece on the topic in addition: https://www.youtube.com/watch?v=wqn3gR1WTcA
984.
by Tostino 4y ago
I think we should vocalize the concept of "negative work" more often. That being work with which society as a whole would be better off if left undone.
985.
by Tostino 4y ago
This is how I grew up in the US (central FL) in the 90's. I was 6-7 when I was riding my bike down to the corner store ~0.5mi away, and ~10 I was staying home and watching my sisters when necessary. I was perfectly capable of cooking a
986.
by Tostino 4y ago
I think hierarchy of needs comes into play here. You may not spend all that much time thinking about philosophy if you're foraging or farming, and busy with physical labor. Also, messages get lost, and misinterpreted. Quickly. I do bel
987.
by Tostino 4y ago
Exactly my experience. And if you point out the errors, often it will correct itself (most of the time) and explain why it was incorrect before (sometimes).
988.
by Tostino 4y ago
Exactly, having your programs able to run on a developer's machine is really important for productivity.
989.
by Tostino 4y ago
That is fair enough, I was weighing the cost of getting it wrong very highly (probably more than I should have).
990.
by Tostino 4y ago
No. For some opensource work I am doing.

More ›