bwood

I was looking forward to that as well, but Siri refuses to read ChatGPT's responses to me...which makes it substantially less useful. It's just as frustrating as being told "sorry, you'll have to unlock your phone if you want me to play something on youtube music". Let's hope they iterate, I guess?

I've been compiling a list of ways in which physical reality is similar to game engines or mechanical simulation in general. Here's my list so far:

- The observer effect, quantum entanglement and wavefunction collapse --> lazy evaluation

- Speed of light --> speed of causality to resolve the object interaction problem O(n^2)

- Quanta of energy, Planck's length --> discretization of reality to limit computational precision

- Unreasonable effectiveness of mathematics --> reality is implemented with mathematics

- Parsimonious physics --> simple physical rules are less computationally expensive to evaluate

- Entropy --> stability of the simulation and a guaranteed "wind down"

This is probably the most delightfully interesting thing I've read in weeks.

Thank you, that was a fascinating talk and I learned quite a bit.

However, it did not provide a convincing argument as to why LLMs cannot be a part of a "doomer" AI. In fact, I got the opposite vibe from Andrej explaining expected future developments. The whole section on System 2 thinking sounds like a layer constructed around dumb LLMs that would result in vastly improved and more generalizable intelligence.

I agree that just scaling the size of LLMs is probably not sufficient for AGI...but that just seems like one relatively minor piece of all the possible ways it might be achieved.

As much as I can appreciate Doctorow’s exasperation, I find his dismissal of the doomer vs accelerationist debate rather pithy. I would love to be convinced that “dumb” LLMs can never gain sentience (or be finagled into sentience with a wrapper).

What is the actual argument for why that’s true?

(I realize you could turn the question around and ask why I think it might be possible in the first place, but I feel like my expectations have been blown out of the water so regularly and so increasingly frequently that I can’t default to being a naysayer anymore)

Sure, but it would be more constructive to explain why it's wrong. GP's comment has some interesting ideas and it would be better to directly refute them than just claiming they're "factually wrong".

Kable has been super helpful for my company so far! Our billing is pretty complicated because it's usage-based and we're customizing pricing for every single customer at this point.

The integration with Kable was very simple, took like 15 minutes to get a basic implementation working. It fires "billing events" automatically and then these can be applied to custom pricing options for every customer.

Love it!

On the face of it, no, it doesn't sound revolutionary. "Just switch to renewables" is something I've been hearing for 20 years and I never believed it could work until I read this book.

The book is quite thorough in laying out all the challenges (eg, handling variable production from renewables, how to get buy-in from existing fossil fuel stakeholders, how to rearchitect the grid from hub-and-spoke to something more like the internet, how to finance the massive upfront cost in a way that will actually save money, etc) and presents realistic solutions for each.

Please don't take my word for it, just pick up a copy and read it.

It's not really that electricity is more efficient (is IS more efficient in most cases, but a key point of the book is that we can't "efficiency" our way out of the climate crisis). It's that as long as the generation of electricity is clean, the use of it is also clean. We already have clean ways of generating electricity that are cost-competitive with fossil fuels, and by scaling up production it will actually end up massively cheaper than fossil fuels. But there is a high upfront cost to switching, so financing the switch is one of the biggest challenges.

The book is quite thorough in laying out all the challenges (eg, handling variable production from renewables, how to get buy-in from existing fossil fuel stakeholders, etc) and presents realistic solutions for each. I recommend you pick up a copy and read it!