jfoster

Could multiple Chinese characters be counted as a single token?

I've been contemplating that recently. You're of course correct that subsidized tokens won't be forever, but that might only be half the story, since there's two opposing forces in action:

1. Phasing out of subsidized tokens.

2. Token prices being brought down through scaling, better hardware, etc.

It's possible that these might balance each other out sufficiently that token customers won't notice any substantial increase in price.

Presumably you'd be paying for that, right?

What if the CEO of that one will come along and notice all the fat stacks of money he's leaving on the table by not advertising to all of those rich contrarians?

In this case it seems like a bit of coaching could have retained a talented employee who has a preference for taking initiative.

I have a hunch that the order of the day within big tech is to let go of anyone they can. As long as they provide a reason, it's one less headcount that needs to be laid off in the next round.

I'm surprised explanation is necessary, but OK.

Even if the criminals don't themselves consider the presence of cameras, the cameras likely ensure they will not have freedom for too much longer.

But they might stop it from happening again if the perpetrators are imprisoned as a result of the footage.

They might not, but there is a further defence against their lack of consideration about consequences.

It's called incarceration.

The entire concept of a review is that it's one person's opinion.

As I see it, you're making a few mistakes in your thinking in this post:

1. You assume that the recent change is that we are close to building in space, but actually what's changed is that it's becoming clearer what might be commercially useful to build in space. Not so long ago, the best idea for commercializing space was satellites taking photos of Earth, or somehow making spacecraft for mining asteroids. Then Starlink started, and next up will be energy-intensive but high-latency distributed compute, which suits AI training well. A lot of the "exactly how" might still need to be figured out along the way, but it likely will get figured out.

2. You're assuming a space-based data center would look the same as an Earth-based data center. It definitely won't. It probably looks more like a larger scale Starlink, but with much bigger satellites due to the need for giant radiators.

3. You're assuming that different technologies (space DCs vs AVs) will progress in lock step with each other. Definitely not the case. For example, we can get wireless internet from space already, but we still don't have a robot that can give someone a haircut? The tech tree branches are mostly separate, and some explored a lot more than others.

4. Insane complexity is already something that many types of technology require. It does not preclude those things from being explored.

If you read his articles over the years you would continually think that Tesla should go out of business in the near future, yet they never do.

He might not specifically lie, but puts such a negative spin on anything Elon-related that the overall result is essentially a lie.

Fair, but actually you'd surely want your choice of those three, right?

And what's being discussed here is what the better implementation of option 3 is.

My point is that if you're going with one of the possible implementations of option 3, then 22GB per browser is objectively a lot better than 22GB per website.

Why are you pretending those are the options?

The options are:

1. 22GB per website

2. 22GB per browser

3. 0GB / No AI capabilities

By having this in Chrome they are simply ensuring that option 2 replaces option 1. You still have option 3.

Good for Google Cloud, bad for Gemini = ??? for Google

Also much better than every website wanting its own 22 GB rather than the 22 GB being a shared resource.

Doesn't sound great, but consider how much better this is than every webpage trying to load their own models.

If it turns out useful enough I'm sure browsers will just start including it as (perhaps optional?) part of installation.

those days seem so long gone now.

Well, Musk v OpenAI kicks off in one week from now with the objective of forcing them back to their roots. A jury will be deciding whether a nonprofit accepting $50m - $100m of donations and then discarding their mission for an IPO is OK or not. Should be interesting.

... because of how homogenous the internet has become...

Designs have been settling on a local maxima. If you are aware of a better hill to be climbed, please do let everyone know.

It can write (some) code that works. Just roughly guessing from my use, but I think of it as being a bit like ChatGPT circa-2024 in terms of capability & speed.

Disappointing if you compare it to anything else from 2026, but fairly impressive for something that can run locally at an OK speed.

These are speed bumps for you but they make it nearly impossible for something to go mainstream, so rest assured they work extremely well on others.

Suspicion is in the eye of the beholder.

Really cool! Is this using a Google API? How is it not costing a small fortune?

Really nice. Bookmarked it. Thank you for making it!

They will always be training new models, so if training is expensive, that's just part of the business they are in.

Vast amounts of capital have been poured in, but they continue to raise more. Presumably because they need more.

Is the capital being invested without any expectation of ROI?

Reminds me of Facebook's memories feature which used to say: "<name>, we care about you and the memories you share here."

For an app to suggest a personal relationship with you is ridiculous.

Is harm necessary to show in a copyright infringement case?

Self-driving vehicles need aircraft-type maintenance.

That's a hyperbolic false equivalence.

Aircraft typically carry hundreds of people and can crash to the ground. As long as a self-driving car can detect when it is degraded, it can just stop with the blinkers on. Usually with 0 - 2 people inside.

I don't think it's anywhere near peaking yet.

It's probably already far more popular than the 3DTV of the 2010s.

I think VR gaming can easily grow 10x - 100x by having cheaper, better fidelity, less bulky hardware and a better games library.

I bet plenty of gamers haven't bought their first headset yet despite being interested.

When the problem is particularly exacerbated, it's not even "I got mine", but rather, "I already went into eye-watering levels of debt and I'm still paying off the roof over my head."

I think this approach gets the whole industry to adopt it.

Consider the opposite approach. If they let airlines charge any amount for it, the airlines that installed it would make it expensive. No one would use it. Other airlines would feel no pressure to offer it.

By making it free, it gets used, and eventually depended upon. SpaceX are making free wifi the expectation from consumers on flights.