nathanasmith

They also said in the paper that 405B was only trained to "compute-optimal" unlike the smaller models that were trained well past that point indicating the larger model still had some runway so had they continued it would have kept getting stronger.

Maybe something like https://lobste.rs/ where you have to get vouched by another poster before you can comment.

I have a ton of these I keep in the spaced repetition software Anki. I use spaced repetition to remember a lot of stuff but I have a specific "life lessons" section to record things I learn about myself and other little life hacks that help me navigate the world. Everything from the embarrassing to the mundane to the profound, if it's something I think will be useful in the future I take a moment to type it in.

Here's a couple of practical tips.

A very powerful method for remembering facts long term is spaced repetition. There's a go-to app called Anki that makes it easy to get started. A good write up on how it works and why is here: https://gwern.net/spaced-repetition

If you want a neat way to help with more "absent minded" type stuff, just associate whatever you're trying to remember with something unusual in your environment. For example say you're driving home and you want to make sure you stop by the store to get milk on the way. If you wear a watch, flip the watch the other way around on your wrist. Now every time you notice your watch you'll immediately think, "oh yeah the milk." If you don't wear a watch just pick something else in your vicinity and arrange it in an unusual way. Every time you see that object you'll immediately recall the thing you don't want to forget.

I'm using Debian 11 with LXDE on an M1 iPad Pro and it's surprisingly useable. I can use this. I'd recommend trying it out and seeing for yourself.

I had to restart Pipewire just this morning when audio randomly stopped working. It's funny as this is the first time it happened then an hour later I'm reading in this thread about Pipewire acting funny on other people's computers.

Don't let the door hit you on the way out.

Between the hallucinations and over the top agreeableness I have a hard time finding LLMs useful as a reliable learning aid. First you have to disentangle what output is factual from what it's just making up and then you have to be careful not to feed it loaded questions lest you unwittingly lead it on such that it's just agreeing with whatever you say. The amount of work required to get satisfactory results might be better spent doing research in more traditional ways.

All that said I'm very excited for the future and look forward to these problems being solved as I believe they eventually will.

When QQQ and SMH close under the 200 day moving average I'll sell my TQQQ and SOXL repectively. Until then, party on! It's been a wild ride.

You can run Wine and Windows applications on Android quite effectively with https://github.com/brunodev85/winlator

So I imagine this will be at least as good.

The bright side is over the last couple of years I've been broken of my lust for every new device that comes out promising "newer and shinier." The reality is it's the same old locked down slab just a little faster with added bing bings and wahoos. So now I just save most of the cash and spend the rest on upgrading my PC.

This week on GenerativeAI Hospital..

Imagine if the original printing press had had this functionality.

We're about as close to AGI as we are to developing superluminal space flight and working time machines.

I just keep prompting with "continue" when that happens.

>There are plenty of stories of people who realized that therapy was just causing them to ruminate on their problems

This is precisely why I stopped going.

>54% of adults in the United States have a literacy below 6th grade level

I don't care how many times this is repeated it's utterly preposterous. Imagine being so naive or less charitably to possess the motivated reasoning to actually believe this shit.

TheBloke stopped uploading in January. There are others that have stepped up though.

Everything after Roddenberry is fanfiction.

I made a month's salary buying the dip after the Reuter's story broke then selling the rip back up when Elon refuted it. They can play that game every single day for all I care.

So the A17 still benchmarks over 30% faster than the Cortex X4 in single threaded tasks. Sounds about right based on my subjective experience.

Publish your data and prepare to get vilified by professional complainers because the data doesn't conform to their sensibilities. Lots of downside with very little of the opposite.

The beautiful thing is even if it fails spectacularly to follow instructions you can canonically just chalk it up to GlaDOS being GlaDOS!

The ease with which LLMs mimic mundane human behavior has, in some small ways, helped me prioritize my own thoughts and actions. I'll be doing something and realize an LLM could easily perform this action, complete this thought, have this conversation as well as I can. So why am I doing this when my time would probably be better served in pursuit of something else. I don't want to go overboard with this since many mundane things have merit but it does help pull me out of a rut occasionally.

If you had any experience with this at all you would realize how profoundly ignorant it is to conflate the downstream effects of meth with adderall. Anyone reading this please please do not make this mistake.

>I find running on battery, lots of things are throttled or disabled. CPU is slower for example.

FWIW, the performance and CPU features are the same on Mx based MacBooks whether they're plugged in or not.

What lofty goals? The board was questioned repeatedly and never articulated clear reasoning for firing Altman and in the process lost the confidence of the employees hence the "rally". The lack of clarity was their undoing whether there would have been a bag for the employees to lose or not.

The board couldn't even clearly articulate why they fired Sam in the first place. There was a departure from critical thinking but I don't think it was on the part of the employees.

On the contrary, this saga has shown that a huge number of people are extremely passionate about the existence of OpenAI and it's leadership by Altman, much more strongly and in larger numbers than most had suspected. If anything this has solidified the importance of the company and I think people will trust it more that the situation was resolved with the light speed it was.

Exactly. It either works or it doesn't today. If it doesn't I assume it'll be fixed soon so I'll come back tomorrow.