It's interesting to hear about an under-served niche in what sounds an awful lot like what a lot of us were doing 20 to 30 years ago. I liked programming at that level. But if I were young now, I would be afraid, very afraid about the long-term prospects for that specialty. I believe you when you say you are thriving, but the total number of people employed industry-wide (especially domestically) in that work is a small fraction of what it was. I say enjoy it while it lasts, and put some money in the bank.

A CEO suggesting anybody not buy their stuff is always a bit edgy. Like that Yoga pants CEO who said maybe fat women just shouldn't buy our stuff. He didn't last long in the job.

Yes at some point simply adding nice thin add-on displays has to be the way.

I too have a relatively huge 17" Dell Precision laptop, I have a smaller one too, but if you actually have to sit down and use the peripherals and computation that the box provides, the big Precision just slaughters. Sturdy too.

Weird-screen laptops never catch on, but kudos anyways.

A laptop base just has to be a certain size to fit a keyboard good enough to justify it over a phone. If you make the base that wide, and in normal proportion, you have something close to a rectangle. So if you double that height, you get something awfully nonsquare - like this lenovo for example.

And I'm sure the suitability of this screen for today's close-pitch airline seating will be noted....

I'm no reverse engineer so this is more of a question than an assertion, but I would guess there's a shit-ton of formal computing proofs precluding major aspects of reverse engineering - that information is not preserved, or massive parts of it are NP-complete, or that some things must be incomputable. Turing halting-problem type stuff.

Granted, smarter heuristics can get you a much higher-quality answer to an unanswerable question, so long as it doesn't have to be rigorously correct all the time. And maybe that's 99.999% of the practical gain to be had, I don't know.

Whatever dude. I've read the paper and implemented MCTS.

The tree search part is trivial. The heuristics, which it learns through self-play, are practically the whole ballgame. It can only evaluate a tiny, minuscule fraction of possible moves at each turn. It never, ever would have made any notable progress at all towards winning Go with just tree search and not heuristics.

I do think a pure left-to-right LLM certainly needs more of an executive function to guide and direct repeated evaluations, which MCTS accomplishes, but not in a flexible enough way for real-world problems. And that's what these newer 'deliberative' models are trying to do, and what makes them so much more computationally expensive. But it is needed past a point.

Similarly I find it interesting the word "AI" wasn't in the summary. Chess was the paragon of AI for decades. If only we had super-human chess AI, we'd have super-human reasoning. Now chess isn't even called AI. Really the case for today's chess algorithms is a lot stronger than it was for Deep Blue, which was mainly human-specified heuristics and brute force.

Awfully negative spin on a purely positive outcome.

I realized one really useful feature to me on ChatGPT are "Projects." (just the normal $20 paid version, not the super-expensive reasoning models this article is about). I am not sure if projects are avaialble in the free version or not. But you can upload documents about a topic, and it will incorporate all that information into its chats with you in that project. It will remember specific facts about the "project" if told to do so and seamless integrate that knowledge into future answers. I have a long-term complex medical condition and this is super helpful, despite being just a good feature rather than an example of high-end AI per se.

You are right, "final race to AGI" sounds like a doomsday prophet or something. Why do that when it's completely dubious.

I've read it's not even a fixed thing yet. In the handful of flights, one was auctioned, one was for charity... this one, who even knows who's paying who to get Katy Perry showing this off to the world.

Depends, what's your baseline? Compared to Amazon Rufus and whatever google is putting into search results, the cheaper paid version of ChatGPT I use is practically godlike and I presume the free one is the same as that.

The reasoning / deliberative models that take a long time, I have tried somewhat and have found multistep tasks or complex formatting problems on which they do better. But you know what? I don't think people really 'reason' all that often to live, and I normally don't need AI to do much reasoning either. I mostly use it for pulling together and explaining relevant information to me.

I am glad big tech is pouring huge sums into pushing the envelope because there is more out there and it will be needed to do things as simple as tidying up the house (lots of inter-task dependencies). But, for me we're not there yet.

Or, by charging more for the more resource-consuming model, they're allowing the market to discourage people from wasting resources. If it's not worth it, people won't pay.

Really the models need to get smart enough to decide which model to use.

Google wanted to get AI out there so they do it with almost every search, but that isn't affordable so it gives such horrible results they're just tarnishing their AI image in my opinion. Use it when it's called for, but then get the job done.

I have heard the arguments (FUD ?) around population decline. Usually this boils down to something like..."Who is going to pay for my social security check and take care of me when I'm really old"

Fuck that, I want to persist. My thousand generations of fore-bearers didn't get this far so I could just roll over and watch netflix and be gone. Do I love my kids on a personal level? Obviously OBVIOUSLY! I want to see them live good lives, and then pass it on down. That's living.

I know there will still be "people." I know we've been thoroughly brainwashed into valuing only society and whatever kind of companionship makes us happy, preferably to the exclusion of procreation, and all genes are the same anyways so let somebody else do it. But apparently my ancestors didn't feel that way and apparently I inherited that trait.

Sustainability doesn't require where Japan is heading. It doesn't even require any individual to do anything in particular. A gradual decline towards sustainability without forcing anybody or excluding anybody from doing whatever they want, so long as their average approaches a nice decline of say 1.75 until we (our descendants) get down to a more sustainable total. Culturally, somehow targeting and sustaining some target value is hard. OK, so it is. Sustainability is hard.