lukasb

The way I would rephrase OP is "what is the most useful abstraction for the mechanism of action of these drugs". It might be that limiting yourself to neurological phenomena gives you a much longer, more unwieldy model of what's happening here than an equally predictive model that centers on subjective experience.

Even simpler - you pick, knowing nothing, so there's a 2/3 chance you're wrong.

If you're wrong, Monty points to toward the right door.

So you should switch.

edit: enable in settings, log out+in, mouse over "GPT-4" at the top

Hm, is this available in the web app?

How would this work? Say I need multiple function calls to handle a user prompt. OpenAI will only supply one function call at a time. My understanding is that "auto" just lets it choose which function to call, but you still have to know that you need to call it multiple times, and when to stop.

Well, this knows when to stop calling GPT-4, which seems non-trivial.

Can anyone comment on how well this does at coercing json output vs OpenAI function calling?

Why is it foolish? Are drivers really well hardened now?

Worth the click just for the three image sequence showing the effectivess of multispectral imaging.

I'm surprised dual aspect theory as a solution to the mind-body problem isn't better known (at least, I hasn't heard of it.)

Glad to see progress on this, hoping we find out just how badly Plato slandered the sophists.

A lot of the images aren't loading for me in the post, though.

What financial reasons?

People talk about how the grid wouldn't be able to handle everyone switching to electric cars. I wonder how residential solar (which works better in some places than others obviously) fits into that.

Had no idea it was that easy to run arbitrary JavaScript on a mobile web browser. Thanks.

Just get multiple versions, discard outliers per frame, and average.

That doesn't seem like the type of problem that can be solved with a drop-in solution.

What newer schedulers?

Looked at the staff list for a product manager, or a UX designer ... eventually found "product and business development manager" (and "product design manager" and almost everyone else is an engineer. I'm guessing the product and design people are super overstretched and don't have a lot of sway. So that would explain why the UX is what it is.

But Tokyo has accommodated growth without rent increasing significantly. No US city can say the same (edit: zip1234 proves me wrong below.)

perfect

Actually I think the impossibility of using natural language instructions to LLMs to prevent prompt injection demonstrates (or will demonstrate) that no true understanding is happening.

So base model+RLLLMF performs as well as base model+RLHF. That could mean a lot of things - it could mean the base model puts a ceiling on the total possible accuracy, so having human-level performance at the RL step doesn't matter as much. And looking at the scores on the individual tests that make up the composite accuracy metric, that looks probable.

What is the accuracy of the resulting model? Over what range of inputs?

I see people saying things like this but I have yet to see anyone show data for a non-trivial workflow with human-level accuracy over a wide range of inputs, without a human in the loop.

So yeah, good news about that ...

Could the properties of the distribution (the spread? not stats literate enough) be used to calculate a confidence metric for the answer?

Well, Elon Musk just sent Twitter employees an email saying it's worth about $20B. He paid $44B for it.

https://twitter.com/nytimes/status/1640049750966665218

I wonder if he was using a Bluetooth keyboard. I've had to restart the Bluetooth module on my M1 a few times to stop this from happening.

So? Why HASN'T anyone built Visual Basic for the web?

I've been wondering if you could use something like llama-chain's tree summarization, but modified to be aware of inter-module dependencies: https://gpt-index.readthedocs.io/en/latest/guides/index_guid...

This would be a LOT of fun on salt flats